1 2018-06-27 Jeff Law <law@redhat.com>
2 Austin Law <austinklaw@gmail.com>
4 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
5 * config/v850/v850.c (notice_update_cc): Remove.
6 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
7 (CC_NO_CARRY): Likewise.
8 (NOTICE_UPDATE_CC): Define to nothing.
9 * config/v850/v850.md: Remove block comment on cc0 handling
10 Remove "cc" attribute from all patterns. Remove cc_status handling
11 from all patterns. Minor formatting fixes.
13 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
16 (cortex-a76.cortex-a55): Likewise.
17 * config/aarch64/aarch64-tune.md: Regenerate.
18 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
19 cortex-a76.cortex-a55.
21 2018-06-27 Jeff Law <law@redhat.com>
23 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
24 (MULTILIB_DIRNAMES): Similarly.
26 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
28 * gimple.h (gimple_return_retbnd): Delete.
29 (gimple_return_set_retbnd): Likewise.
30 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
31 gimple_return_set_retbnd.
32 * gimple-pretty-print.c (dump_gimple_return): Remove call to
33 gimple_return_retbnd and adjust.
34 * tree-inline.h (struct copy_body_data): Remove retbnd field.
35 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
36 Explicitly return NULL in a couple more cases. Move assertion
37 on debug statements and remove unreachable code.
38 (reset_debug_binding): Do not test id->retbnd.
39 (expand_call_inline): Do not set it.
41 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
43 * configure.ac: Add --disable-gcov option.
44 * configure: Regenerate.
45 * Makefile.in: Honour @enable_gcov@.
46 * doc/install.texi: Document --disable-gcov.
48 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
50 * config/arm/arm-cpus.in (cortex-a76): New entry.
51 (cortex-a76.cortex-a55): Likewise.
52 * config/arm/arm-tables.opt: Regenerate.
53 * config/arm/arm-tune.md: Likewise.
54 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
55 * doc/invoke.texi (ARM Options): Document cortex-a76 and
56 cortex-a76.cortex-a55.
58 2018-06-27 Tamar Christina <tamar.christina@arm.com>
61 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
63 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
65 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
67 (EPILOGUE_USES): Likewise.
69 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
71 * tree-inline.c (remap_location): New function extracted from...
72 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
73 (copy_phis_for_bb): ...here. Call remap_location.
74 (copy_cfg_body): Adjust call to copy_edges_for_bb.
76 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
78 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
79 unaligned vsx for 16B memset.
81 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
84 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
85 ieee128_float_type_node to long_double_type_node unless
86 TARGET_LONG_DOUBLE_128 is set.
88 2018-06-26 David Malcolm <dmalcolm@redhat.com>
90 * cfgloop.c (get_loop_location): Convert return type from
91 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
92 by implicit construction from rtx_insn *, and using
93 dump_user_location_t::from_function_decl for the fallback case.
94 * cfgloop.h (get_loop_location): Convert return type from
95 location_t to dump_user_location_t.
96 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
97 dump_printf_loc to pass in a dump_location_t rather than a
98 location_t, via the gimple stmt.
99 * coverage.c (get_coverage_counts): Update calls to
100 dump_printf_loc to pass in dump_location_t rather than a
102 * doc/optinfo.texi (Dump types): Convert example of
103 dump_printf_loc from taking "locus" to taking "insn". Update
104 description of the "_loc" calls to cover dump_location_t.
105 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
107 (dump_user_location_t::dump_user_location_t): New constructors,
108 from gimple *stmt and rtx_insn *.
109 (dump_user_location_t::from_function_decl): New function.
110 (dump_loc): Make static.
111 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
112 const dump_location_t &.
113 (dump_generic_expr_loc): Delete.
114 (dump_printf_loc): Convert param "loc" from location_t to
115 const dump_location_t &.
116 (selftest::test_impl_location): New function.
117 (selftest::dumpfile_c_tests): New function.
118 * dumpfile.h: Include "profile-count.h".
119 (class dump_user_location_t): New class.
120 (struct dump_impl_location_t): New struct.
121 (class dump_location_t): New class.
122 (dump_printf_loc): Convert 2nd param from source_location to
123 const dump_location_t &.
124 (dump_generic_expr_loc): Delete.
125 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
126 const dump_location_t &.
127 * gimple-fold.c (fold_gimple_assign): Update call to
128 dump_printf_loc to pass in a dump_location_t rather than a
129 location_t, via the gimple stmt.
130 (gimple_fold_call): Likewise.
131 * gimple-loop-interchange.cc
132 (loop_cand::analyze_iloop_reduction_var): Update for change to
133 check_reduction_path.
134 (tree_loop_interchange::interchange): Update for change to
136 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
137 change in return-type of find_loop_location.
138 (graphite_regenerate_ast_isl): Likewise.
139 * graphite-optimize-isl.c (optimize_isl): Likewise.
140 * graphite.c (graphite_transform_loops): Likewise.
141 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
142 pass in a dump_location_t rather than a location_t, via the
144 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
145 * ipa.c (walk_polymorphic_call_targets): Likewise.
146 * loop-unroll.c (report_unroll): Convert "locus" param from
147 location_t to dump_location_t.
148 (decide_unrolling): Update for change to get_loop_location's
150 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
151 location_t to dump_user_location_t.
152 (grid_find_single_omp_among_assignments_1): Updates calls to
153 dump_printf_loc to pass in a dump_location_t rather than a
154 location_t, via the gimple stmt.
155 (grid_parallel_clauses_gridifiable): Convert "tloc" from
156 location_t to dump_location_t. Updates calls to dump_printf_loc
157 to pass in a dump_location_t rather than a location_t, via the
159 (grid_inner_loop_gridifiable_p): Likewise.
160 (grid_dist_follows_simple_pattern): Likewise.
161 (grid_gfor_follows_tiling_pattern): Likewise.
162 (grid_target_follows_gridifiable_pattern): Likewise.
163 (grid_attempt_target_gridification): Convert initialization
164 of local "grid" from memset to zero-initialization; FIXME: does
165 this require C++11? Update call to dump_printf_loc to pass in a
166 optinfo_location rather than a location_t, via the gimple stmt.
167 * profile.c (read_profile_edge_counts): Updates call to
168 dump_printf_loc to pass in a dump_location_t rather than a
170 (compute_branch_probabilities): Likewise.
171 * selftest-run-tests.c (selftest::run_tests): Call
173 * selftest.h (dumpfile_c_tests): New decl.
174 * tree-loop-distribution.c (pass_loop_distribution::execute):
175 Update for change in return type of find_loop_location.
176 * tree-parloops.c (parallelize_loops): Likewise.
177 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
178 "locus" from location_t to dump_user_location_t.
179 (canonicalize_loop_induction_variables): Likewise.
180 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
181 for change in return type of find_loop_location.
182 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
183 to dump_printf_loc to pass in a dump_location_t rather than a
184 location_t, via the stmt.
185 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
187 * tree-vect-loop-manip.c (find_loop_location): Convert return
188 type from source_location to dump_user_location_t.
189 (vect_do_peeling): Update for above change.
190 (vect_loop_versioning): Update for change in type of
192 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
193 from location_t to dump_user_location_t.
194 (vect_estimate_min_profitable_iters): Update for change in type
196 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
197 location_t to dump_location_t.
198 (vect_slp_bb): Update for change in type of vect_location.
199 * tree-vectorizer.c (vect_location): Convert from source_location
200 to dump_user_location_t.
201 (try_vectorize_loop_1): Update for change in vect_location's type.
202 (vectorize_loops): Likewise.
203 (increase_alignment): Likewise.
204 * tree-vectorizer.h (vect_location): Convert from source_location
205 to dump_user_location_t.
206 (find_loop_location): Convert return type from source_location to
207 dump_user_location_t.
208 (check_reduction_path): Convert 1st param from location_t to
209 dump_user_location_t.
210 * value-prof.c (check_counter): Update call to dump_printf_loc to
211 pass in a dump_user_location_t rather than a location_t; update
212 call to error_at for change in type of "locus".
213 (check_ic_target): Update call to dump_printf_loc to
214 pass in a dump_user_location_t rather than a location_t, via the
217 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
219 * config/s390/s390.h (enum processor_flags): Do not use
220 default tune parameter when -march was specified.
222 2018-06-26 Jakub Jelinek <jakub@redhat.com>
225 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
226 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
229 2018-06-26 Richard Biener <rguenther@suse.de>
231 PR tree-optimization/86287
233 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
234 (vect_analyze_loop): Initialize n_stmts.
236 2018-06-26 Richard Biener <rguenther@suse.de>
239 * fold-const.c (fold_convertible_p): Pointer extension
242 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
245 * dwarf2out.c (loc_list_has_views): Adjust comments.
246 (dw_loc_list): Split single cross-partition range with
249 2018-06-25 Jeff Law <law@redhat.com>
251 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
252 on -mbig-switch by default.
254 * config/v850/predicates.md (const_float_1_operand): Fix match_code
256 (const_float_0_operand): Remove unused predicate.
257 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
258 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
259 (recipsf2): New expander. Original pattern now called
261 (recipdf2, recipdf2_insn): Similarly.
262 (rsqrtsf2, rsqrtsf2_insn): Similarly
263 (rsqrtdf2, rsqrtdf2_insn): Similarly
265 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
267 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
268 Simplify logic for FreeBSD (twice).
270 2018-06-25 Martin Sebor <msebor@redhat.com>
272 PR tree-optimization/86204
273 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
274 a strnlen result if it's less than the length of the string.
276 2018-06-25 Martin Sebor <msebor@redhat.com>
278 PR tree-optimization/85700
279 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
280 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
281 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
283 2018-06-25 Martin Sebor <msebor@redhat.com>
285 * doc/extend.texi (Zero-length arrays): Update and clarify.
287 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
289 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
290 added IEEE/IBM long double multilib support on PowerPC little
291 endian Linux systems.
292 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
293 (MULTILIB_DEFAULTS): Likewise.
294 * config/rs6000/rs6000.c (rs6000_option_override_internal):
296 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
297 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
298 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
300 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
303 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
304 (REORDER_45): Likewise.
306 2018-06-25 Jeff Law <law@redhat.com>
308 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
309 dividend to 32 bits. Adjust length.
310 (udivmodhi4): Cleanup output template. Fix length.
312 2018-06-25 Carl Love <cel@us.ibm.com>
314 * config/rs6000/vsx.md: Change word selector to prefered location.
316 2018-06-25 Richard Biener <rguenther@suse.de>
318 PR tree-optimization/86304
319 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
320 epilogue-if-converted loops as well.
322 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
324 * lto-section-out.c (lto_begin_section): Do not print section
325 name for noaddr and unnumbered dumps.
327 2018-06-25 Richard Biener <rguenther@suse.de>
329 * tree-vectorizer.h (struct vec_info_shared): New structure
330 with parts split out from struct vec_info and loop_nest from
331 struct _loop_vec_info.
332 (struct vec_info): Adjust accordingly.
333 (struct _loop_vec_info): Likewise.
334 (LOOP_VINFO_LOOP_NEST): Adjust.
335 (LOOP_VINFO_DATAREFS): Likewise.
336 (LOOP_VINFO_DDRS): Likewise.
337 (struct _bb_vec_info): Likewise.
338 (BB_VINFO_DATAREFS): Likewise.
339 (BB_VINFO_DDRS): Likewise.
340 (struct _stmt_vec_info): Add dr_aux member.
341 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
342 (DR_MISALIGNMENT_UNINITIALIZED): New.
343 (set_dr_misalignment): Adjust.
344 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
345 (vect_analyze_loop): Adjust prototype.
346 (vect_analyze_loop_form): Likewise.
347 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
348 Compute dependences lazily.
349 (vect_record_base_alignments): Use shared datarefs/ddrs.
350 (vect_verify_datarefs_alignment): Likewise.
351 (vect_analyze_data_refs_alignment): Likewise.
352 (vect_analyze_data_ref_accesses): Likewise.
353 (vect_analyze_data_refs): Likewise.
354 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
355 constructor parameter for shared part.
356 (vect_analyze_loop_form): Pass in shared part and adjust.
357 (vect_analyze_loop_2): Pass in storage for the number of
358 stmts. Move loop nest finding to the caller. Compute
360 (vect_analyze_loop): Pass in shared part.
361 (vect_transform_loop): Verify shared datarefs are unchanged.
362 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
363 constructor parameter for shared part.
364 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
365 (vect_slp_bb): Verify shared datarefs are unchanged before
367 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
369 (new_stmt_vec_info): Initialize DR_AUX misalignment to
370 DR_MISALIGNMENT_UNINITIALIZED.
371 * tree-vectorizer.c (vec_info::vec_info): Add constructor
372 parameter for shared part.
373 (vec_info::~vec_info): Adjust.
374 (vec_info_shared::vec_info_shared): New.
375 (vec_info_shared::~vec_info_shared): Likewise.
376 (vec_info_shared::save_datarefs): Likewise.
377 (vec_info_shared::check_datarefs): Likewise.
378 (try_vectorize_loop_1): Construct shared part live for analyses
379 of a single loop for multiple vector sizes.
380 * tree-parloops.c (gather_scalar_reductions): Adjust.
382 2018-06-25 Richard Biener <rguenther@suse.de>
384 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
385 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
386 (vect_analyze_data_refs): Remove similar code from here and
387 simplify accordingly.
389 2018-06-25 Richard Biener <rguenther@suse.de>
391 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
392 for reverse storage order accesses rather than asserting
393 they cannot happen here.
395 2018-06-25 Tom de Vries <tdevries@suse.de>
398 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
399 Use data16 instead of .byte for insn prefix.
401 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
404 * parser.c (make_char_string_pack): Pass this literal chars
405 through cpp_interpret_string.
406 (cp_parser_userdef_numeric_literal): Check the result of
407 make_char_string_pack.
409 2018-06-24 Maya Rashish <coypu@sdf.org>
411 * ginclude/stddef.h: Simplify conditions around avoiding
412 re-definition of __size_t.
414 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
416 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
417 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
419 2018-06-22 Maya Rashish <coypu@sdf.org>
421 * doc/invoke.texi (mno-fancy-math-387): Update for changes
422 made to OpenBSD and NetBSD through the years.
424 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
426 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
427 behavior of vec_pack (vector double, vector double) to match
428 behavior of vec_float2 (vector double, vector double).
430 2018-06-22 Olivier Hainque <hainque@adacore.com>
432 * gimplify.c (gimplify_function_tree): Prevent creation
433 of a trampoline for the address of the current function
434 passed to entry/exit instrumentation hooks.
436 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
439 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
442 2018-06-22 Martin Liska <mliska@suse.cz>
444 PR tree-optimization/86263
445 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
446 Bail out if is_enabled is false.
447 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
449 (jump_table_cluster::is_enabled): New function.
451 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
453 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
454 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
455 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
456 (lto_input_ts_binfo_tree_pointers): Likewise.
457 * tree-streamer-out.c (streamer_write_tree_bitfields,
458 write_ts_binfo_tree_pointers): Likewise.
459 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
461 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
463 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
465 2018-06-22 Martin Liska <mliska@suse.cz>
467 * symbol-summary.h (get): Make it pure and inline move
468 functionality from ::get function.
469 (get): Remove and inline into ::get and ::get_create.
470 (get_create): Move code from ::get function.
472 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
475 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
476 -x assembler-with-cpp.
478 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
480 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
481 _FILE_OFFSET_BITS=64 for C++.
483 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
485 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
486 conversion insn that shows up when pr85657-3.c is compiled using
487 IEEE 128-bit long double.
488 (neg<mode>2_internal): Use the correct mode to check whether the
489 mode is IBM extended.
490 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
491 multiply and divide external functions from being created more
494 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
496 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
498 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
499 the edge can be forwarded.
500 (cfg_layout_merge_blocks): Likewise.
502 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
504 * except.c (finish_eh_generation): Commit edge insertions only after
505 the EH edges have been redirected from post-landing to landing pads.
507 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
509 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
510 create_tmp_var_for to create the FRAME decl.
511 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
513 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
515 * tree-inline.c (copy_edges_for_bb): Minor tweak.
516 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
517 debug statement when resetting its value.
518 (expand_call_inline): Copy the locus of the call onto the assignment
519 of the return value, if any. Use local variable in more cases.
521 2018-06-21 Martin Liska <mliska@suse.cz>
523 * ipa-pure-const.c (propagate_nothrow): Use
524 funct_state_summaries->get.
525 (dump_malloc_lattice): Likewise.
526 (propagate_malloc): Likewise.
528 2018-06-21 Richard Biener <rguenther@suse.de>
530 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
531 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
532 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
534 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
535 BLOCK_ABSTRACT_ORIGIN unconditionally.
537 2018-06-21 David Malcolm <dmalcolm@redhat.com>
539 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
541 * ipa-reference.c (ipa_reference_c_finalize): Delete
542 ipa_ref_opt_sum_summaries and set it to NULL.
544 2018-06-21 Tom de Vries <tdevries@suse.de>
546 PR tree-optimization/85859
547 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
548 test with comment from bb_no_side_effects_p.
550 2018-06-21 Richard Biener <rguenther@suse.de>
552 PR tree-optimization/86232
553 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
554 max for constant niter.
556 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
558 * config/aarch64/aarch64-simd.md
559 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
561 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
563 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
564 Make opernads of the unspec commutative.
566 2018-06-21 Richard Biener <rguenther@suse.de>
568 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
569 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
570 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
571 (vect_analyze_data_ref_dependence): Re-order checks to deal with
573 (vect_record_base_alignments): Do not record base alignment
574 for gathers or scatters.
575 (vect_compute_data_ref_alignment): Drop return value that is always
576 true. Bail out early for gathers or scatters.
577 (vect_enhance_data_refs_alignment): Bail out early for gathers
579 (vect_find_same_alignment_drs): Likewise.
580 (vect_analyze_data_refs_alignment): Remove dead code.
581 (vect_slp_analyze_and_verify_node_alignment): Likewise.
582 (vect_analyze_data_refs): For possible gathers or scatters do
583 not create an alternate DR, just check their possible validity
584 and mark them. Adjust DECL_NONALIASED handling to not rely
586 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
587 update inits of gathers or scatters.
588 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
589 Also copy gather/scatter flag to pattern vinfo.
591 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
593 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
594 behavior of vec_packsu (vector unsigned long long, vector unsigned
595 long long) to match behavior of vec_packs with same signature.
597 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
598 Thomas Schwinge <thomas@codesourcery.com>
599 Cesar Philippidis <cesar@codesourcery.com>
601 * gimplify.c (gimplify_scan_omp_clauses): Add support for
602 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
603 (gimplify_adjust_omp_clauses): Likewise.
604 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
605 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
606 (gimplify_omp_target_update): Update handling of acc update and
608 * omp-low.c (install_var_field): Remove unused parameter
609 base_pointers_restrict.
610 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
611 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
613 (omp_target_base_pointers_restrict_p): Delete.
614 (scan_omp_target): Update call to scan_sharing_clauses.
615 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
617 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
618 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
619 (convert_local_omp_clauses): Likewise.
620 * tree-pretty-print.c (dump_omp_clause): Likewise.
621 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
623 (omp_clause_code_name): Likewise.
625 2018-06-20 Jakub Jelinek <jakub@redhat.com>
628 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
631 PR tree-optimization/86231
632 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
633 anti-range don't overwrite *vr0min before using it to compute *vr0max.
635 2018-06-20 Tom de Vries <tdevries@suse.de>
637 PR tree-optimization/86097
638 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
639 iv type if signedness of iv type is not the same as that of *nit.
641 2018-06-20 Jakub Jelinek <jakub@redhat.com>
643 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
644 EDGE_EH edges, verify they are all EDGE_EH.
646 2018-06-20 Maya Rashish <coypu@sdf.org>
648 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
650 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
652 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
653 * config/aarch64/aarch64.c (xgene1_tunings): Add
654 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
655 (aarch64_mode_valid_for_sched_fusion_p):
657 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
658 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
660 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
662 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
663 * config/aarch64/iterators.md (VQ2): New mode iterator.
665 2018-06-20 Martin Liska <mliska@suse.cz>
667 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
668 Change default ratio from 10 to 8.
670 2018-06-20 Martin Liska <mliska@suse.cz>
672 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
674 (bit_test_cluster::find_bit_tests): Likewise.
675 (switch_decision_tree::analyze_switch_statement): Find clusters.
676 * tree-switch-conversion.h (struct jump_table_cluster): Document
679 2018-06-20 Martin Liska <mliska@suse.cz>
681 * tree-switch-conversion.c (switch_conversion::collect):
682 Record m_uniq property.
683 (switch_conversion::expand): Bail out for special conditions.
684 (group_cluster::~group_cluster): New.
685 (group_cluster::group_cluster): Likewise.
686 (group_cluster::dump): Likewise.
687 (jump_table_cluster::emit): New.
688 (switch_decision_tree::fix_phi_operands_for_edges): New.
689 (struct case_node): Remove struct.
690 (jump_table_cluster::can_be_handled): New.
691 (case_values_threshold): Moved to header.
692 (reset_out_edges_aux): Likewise.
693 (jump_table_cluster::is_beneficial): New.
694 (bit_test_cluster::can_be_handled): Likewise.
695 (add_case_node): Remove.
696 (bit_test_cluster::is_beneficial): New.
697 (case_bit_test::cmp): New.
698 (bit_test_cluster::emit): New.
699 (expand_switch_as_decision_tree_p): Remove.
700 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
701 (fix_phi_operands_for_edge): Likewise.
702 (switch_decision_tree::analyze_switch_statement): New.
703 (compute_cases_per_edge): Move ...
704 (switch_decision_tree::compute_cases_per_edge): ... here.
705 (try_switch_expansion): Likewise.
706 (switch_decision_tree::try_switch_expansion): Likewise.
707 (record_phi_operand_mapping): Likewise.
708 (switch_decision_tree::record_phi_operand_mapping): Likewise.
709 (emit_case_decision_tree): Likewise.
710 (switch_decision_tree::emit): Likewise.
711 (balance_case_nodes): Likewise.
712 (switch_decision_tree::balance_case_nodes): Likewise.
713 (dump_case_nodes): Likewise.
714 (switch_decision_tree::dump_case_nodes): Likewise.
715 (emit_jump): Likewise.
716 (switch_decision_tree::emit_jump): Likewise.
717 (emit_cmp_and_jump_insns): Likewise.
718 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
719 (emit_case_nodes): Likewise.
720 (switch_decision_tree::emit_case_nodes): Likewise.
721 (conditional_probability): Remove.
722 * tree-switch-conversion.h (enum cluster_type): New.
724 (struct cluster): Likewise.
725 (cluster::cluster): Likewise.
726 (struct simple_cluster): Likewise.
727 (simple_cluster::simple_cluster): Likewise.
728 (struct group_cluster): Likewise.
729 (struct jump_table_cluster): Likewise.
730 (struct bit_test_cluster): Likewise.
731 (struct min_cluster_item): Likewise.
732 (struct case_tree_node): Likewise.
733 (case_tree_node::case_tree_node): Likewise.
734 (jump_table_cluster::case_values_threshold): Likewise.
735 (struct case_bit_test): Likewise.
736 (struct switch_decision_tree): Likewise.
737 (struct switch_conversion): Likewise.
738 (switch_decision_tree::reset_out_edges_aux): Likewise.
740 2018-06-20 Martin Liska <mliska@suse.cz>
742 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
743 (hoist_edge_and_branch_if_true): Likewise.
744 (expand_switch_using_bit_tests_p): Likewise.
745 (struct case_bit_test): Likewise.
746 (case_bit_test_cmp): Likewise.
747 (emit_case_bit_tests): Likewise.
748 (switch_conversion::switch_conversion): New class.
749 (struct switch_conv_info): Remove old struct.
750 (collect_switch_conv_info): More to ...
751 (switch_conversion::collect): ... this.
752 (check_range): Likewise.
753 (switch_conversion::check_range): Likewise.
754 (check_all_empty_except_final): Likewise.
755 (switch_conversion::check_all_empty_except_final): Likewise.
756 (check_final_bb): Likewise.
757 (switch_conversion::check_final_bb): Likewise.
758 (create_temp_arrays): Likewise.
759 (switch_conversion::create_temp_arrays): Likewise.
760 (free_temp_arrays): Likewise.
761 (gather_default_values): Likewise.
762 (switch_conversion::gather_default_values): Likewise.
763 (build_constructors): Likewise.
764 (switch_conversion::build_constructors): Likewise.
765 (constructor_contains_same_values_p): Likewise.
766 (switch_conversion::contains_same_values_p): Likewise.
767 (array_value_type): Likewise.
768 (switch_conversion::array_value_type): Likewise.
769 (build_one_array): Likewise.
770 (switch_conversion::build_one_array): Likewise.
771 (build_arrays): Likewise.
772 (switch_conversion::build_arrays): Likewise.
773 (gen_def_assigns): Likewise.
774 (switch_conversion::gen_def_assigns): Likewise.
775 (prune_bbs): Likewise.
776 (switch_conversion::prune_bbs): Likewise.
777 (fix_phi_nodes): Likewise.
778 (switch_conversion::fix_phi_nodes): Likewise.
779 (gen_inbound_check): Likewise.
780 (switch_conversion::gen_inbound_check): Likewise.
781 (process_switch): Use the newly created class.
782 (switch_conversion::expand): New.
783 (switch_conversion::~switch_conversion): New.
784 * tree-switch-conversion.h: New file.
786 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
788 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
789 tree-vect-patterns.c.
790 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
791 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
792 (vect_recog_sad_pattern): Likewise.
793 (vect_recog_widen_sum_pattern): Likewise.
794 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
795 (vect_recog_widen_shift_pattern): Remove the type_in argument.
796 (vect_recog_rotate_pattern): Likewise.
797 (vect_recog_mult_pattern): Likewise.
798 (vect_recog_vector_vector_shift_pattern): Likewise.
799 (vect_recog_divmod_pattern): Likewise.
800 (vect_recog_mixed_size_cond_pattern): Likewise.
801 (vect_recog_bool_pattern): Likewise.
802 (vect_recog_mask_conversion_pattern): Likewise.
803 (vect_try_gather_scatter_pattern): Likewise.
804 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
805 (vect_recog_over_widening_pattern): Likewise.
806 (vect_recog_gather_scatter_pattern): Likewise.
807 (vect_recog_func_ptr): Move from tree-vectorizer.h
808 (vect_vect_recog_func_ptrs): Move further down the file.
809 (vect_recog_func): Likewise. Remove the third argument.
810 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
811 (vect_pattern_recog_1): Expect the pattern function to do any
812 necessary target tests. Also expect it to provide a vector type.
813 Remove the type_in handling.
815 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
817 * tree-vect-patterns.c (vect_pattern_detected): New function.
818 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
819 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
820 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
821 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
822 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
823 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
824 (vect_recog_mask_conversion_pattern)
825 (vect_try_gather_scatter_pattern): Likewise.
827 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
829 * tree-vect-patterns.c (vect_get_internal_def): New function.
830 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
831 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
832 (search_type_for_mask_1): Use it.
834 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
836 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
837 redundant WIDEN_SUM_EXPR handling.
838 (vect_recog_sad_pattern): Likewise.
840 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
842 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
843 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
844 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
845 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
846 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
848 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
850 * tree-vect-stmts.c (vectorizable_call): Make sure that we
851 use the stmt_vec_info of the original bb statement for the
852 new zero assignment, even if the call is part of a pattern.
854 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
856 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
857 that the sequence is attached to the original statement rather
858 than the pattern statement.
859 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
860 PATTERN_DEF_SEQ from the original statement rather than
861 the main pattern statement.
862 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
863 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
864 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
866 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
868 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
869 definition statements before the early exit for statements that aren't
871 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
873 (vect_transform_loop): ...here. Process pattern definition
874 statements without first checking whether the main pattern
875 statement is live or relevant.
877 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
879 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
880 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
882 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
884 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
885 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
886 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
887 (expand_block_compare): Change select_block_compare_mode call.
888 (expand_strncmp_align_check): Use new functions, fix comment.
889 (emit_final_str_compare_gpr): New function.
890 (expand_strn_compare): Refactor and clean up code.
891 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
893 2018-06-19 Tony Reix <tony.reix@atos.com>
894 Damien Bergamini <damien.bergamini@atos.com>
895 David Edelsohn <dje.gcc@gmail.com>
897 * collect2.c (static_obj): New variable.
898 (static_libs): New variable.
899 (is_in_list): Uncomment declaration.
900 (main): Track AIX libraries linked statically.
901 (is_in_list): Uncomment definition.
902 (scan_prog_file): Don't add AIX shared libraries initializer
903 to constructor list if linking statically.
905 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
907 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
909 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
911 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
913 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
916 2018-06-19 Martin Liska <mliska@suse.cz>
918 * config/i386/i386.c (ix86_can_inline_p): Do not use
919 ipa_fn_summaries::get_create.
920 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
922 (devirtualization_time_bonus): Likewise.
923 (ipcp_propagate_stage): Likewise.
924 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
925 (edge_set_predicate): Likewise.
926 (evaluate_conditions_for_known_args): Likewise.
927 (evaluate_properties_for_edge): Likewise.
928 (ipa_call_summary::reset): Tranform to ...
929 (ipa_call_summary::~ipa_call_summary): ... this.
930 (ipa_fn_summary::reset): Transform to ...
931 (ipa_fn_summary::~ipa_fn_summary): ... this.
932 (ipa_fn_summary_t::remove): Rename to ...
933 (ipa_fn_summary_t::remove_callees): ... this.
934 (ipa_fn_summary_t::duplicate): Use placement new
935 instead of memory copy.
936 (ipa_call_summary_t::duplicate): Likewise.
937 (ipa_call_summary_t::remove): Remove.
938 (dump_ipa_call_summary): Change get_create to get.
939 (ipa_dump_fn_summary): Dump only when summary exists.
940 (analyze_function_body): Use symbol_summary::get instead
942 (compute_fn_summary): Likewise.
943 (estimate_edge_devirt_benefit): Likewise.
944 (estimate_edge_size_and_time): Likewise.
945 (inline_update_callee_summaries): Likewise.
946 (remap_edge_change_prob): Likewise.
947 (remap_edge_summaries): Likewise.
948 (ipa_merge_fn_summary_after_inlining): Likewise.
949 (write_ipa_call_summary): Likewise.
950 (ipa_fn_summary_write): Likewise.
951 (ipa_free_fn_summary): Likewise.
952 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
953 (struct ipa_call_summary): Likewise.
954 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
956 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
957 (estimate_size_after_inlining): Likewise.
958 (estimate_growth): Likewise.
959 (growth_likely_positive): Likewise.
960 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
961 (inline_call): Likewise.
962 * ipa-inline.c (caller_growth_limits): Likewise.
963 (can_inline_edge_p): Likewise.
964 (can_inline_edge_by_limits_p): Likewise.
965 (compute_uninlined_call_time): Likewise.
966 (compute_inlined_call_time): Likewise.
967 (want_inline_small_function_p): Likewise.
968 (edge_badness): Likewise.
969 (update_caller_keys): Likewise.
970 (update_callee_keys): Likewise.
971 (inline_small_functions): Likewise.
972 (inline_to_all_callers_1): Likewise.
973 (dump_overall_stats): Likewise.
974 (early_inline_small_functions): Likewise.
975 (early_inliner): Likewise.
976 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
977 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
978 * ipa-pure-const.c (malloc_candidate_p): Likewise.
979 * ipa-split.c (execute_split_functions): Likewise.
980 * symbol-summary.h: Likewise.
981 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
983 2018-06-19 Richard Biener <rguenther@suse.de>
985 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
986 (vectorize_loops): ... here. Fix dbgcnt handling.
987 (try_vectorize_loop): Wrap try_vectorize_loop_1.
989 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
992 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
993 ieee128 argument takes up only one (vector) register, not two (floating
996 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
998 * gimplify.c (gimplify_init_constructor): Really never clear for an
999 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
1001 2018-06-19 Richard Biener <rguenther@suse.de>
1003 PR tree-optimization/86179
1004 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
1005 after failed recognition.
1007 2018-06-18 Martin Sebor <msebor@redhat.com>
1010 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
1011 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
1012 Handle integer subtraction.
1013 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
1014 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
1016 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1018 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
1019 param from rtx to rtx_insn *.
1020 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
1022 (frv_ifcvt_modify_insn): Likwise.
1023 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
1024 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
1025 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
1026 as_a <rtx_insn *> cast to local "unprotected_region" once
1027 it's been established that it's not NULL or pc_rtx.
1028 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
1029 param "sethi" from rtx to rtx_insn *.
1030 (nds32_group_float_insns): Likewise for param "insn".
1031 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
1033 (vax_output_int_subtract): Likewise.
1034 * config/vax/vax.c (vax_output_int_add): Likewise for param
1036 (vax_output_int_subtract): Likewise.
1037 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
1038 (emit_pattern_after): Likewise for param "after".
1039 (emit_insn_after): Likewise.
1040 (emit_jump_insn_after): Likewise.
1041 (emit_call_insn_after): Likewise.
1042 (emit_debug_insn_after): Likewise.
1043 (emit_pattern_before): Likewise for param "before".
1044 (emit_insn_before): Likewise.
1045 (emit_jump_insn_before): Likewise.
1046 * final.c (get_insn_template): Likewise for param "insn", removing
1048 * output.h (get_insn_template): Likewise for 2nd param.
1049 * rtl.h (emit_insn_before): Likewise.
1050 (emit_jump_insn_before): Likewise.
1051 (emit_debug_insn_before_noloc): Likewise.
1052 (emit_insn_after): Likewise.
1053 (emit_jump_insn_after): Likewise.
1054 (emit_call_insn_after): Likewise.
1055 (emit_debug_insn_after): Likewise.
1056 (set_insn_deleted): Likewise for param.
1058 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
1061 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
1062 floating point modes, so that IFmode is numerically greater than
1063 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
1064 to declare the ordering. This prevents IFmode from being
1065 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
1066 machine. Include rs6000-modes.h to share the fractional values
1067 between genmodes* and the rest of the compiler.
1071 * config/rs6000/rs6000-modes.h: New file.
1072 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
1073 meaning of rs6000_long_double_size so that 126..128 selects an
1074 appropriate 128-bit floating point type.
1075 (rs6000_option_override_internal): Likewise.
1076 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
1077 (TARGET_LONG_DOUBLE_128): Change the meaning of
1078 rs6000_long_double_size so that 126..128 selects an appropriate
1079 128-bit floating point type.
1080 (LONG_DOUBLE_TYPE_SIZE): Update comment.
1081 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
1082 source and destination to match the standard usage.
1083 (truncifkf2): Likewise.
1084 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
1085 ISA 2.07 to use an explicit clobber, instead of passing in a
1087 (copysign<mode>3_soft): Likewise.
1089 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1091 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1092 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
1093 (vect_slp_analyze_instance_dependence): Likewise.
1094 (vect_enhance_data_refs_alignment): Likewise.
1095 (vect_analyze_data_refs_alignment): Likewise.
1096 (vect_slp_analyze_and_verify_instance_alignment
1097 (vect_analyze_data_ref_accesses): Likewise.
1098 (vect_prune_runtime_alias_test_list): Likewise.
1099 (vect_analyze_data_refs): Likewise.
1100 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
1101 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
1102 (vect_analyze_scalar_cycles_1): Likewise.
1103 (vect_get_loop_niters): Likewise.
1104 (vect_analyze_loop_form_1): Likewise.
1105 (vect_update_vf_for_slp): Likewise.
1106 (vect_analyze_loop_operations): Likewise.
1107 (vect_analyze_loop): Likewise.
1108 (vectorizable_induction): Likewise.
1109 (vect_transform_loop): Likewise.
1110 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
1111 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1112 (vect_make_slp_decision): Likewise.
1113 (vect_detect_hybrid_slp): Likewise.
1114 (vect_slp_analyze_operations): Likewise.
1115 (vect_slp_bb): Likewise.
1116 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
1117 (vectorizable_bswap): Likewise.
1118 (vectorizable_call): Likewise.
1119 (vectorizable_simd_clone_call): Likewise.
1120 (vectorizable_conversion): Likewise.
1121 (vectorizable_assignment): Likewise.
1122 (vectorizable_shift): Likewise.
1123 (vectorizable_operation): Likewise.
1124 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
1126 2018-06-18 Martin Sebor <msebor@redhat.com>
1128 PR tree-optimization/81384
1129 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
1130 * builtins.c (expand_builtin_strnlen): New function.
1131 (expand_builtin): Call it.
1132 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
1133 * builtins.def (BUILT_IN_STRNLEN): New.
1134 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
1135 Warn for bounds in excess of maximum object size.
1136 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
1137 single-value ranges. Handle strnlen.
1138 (handle_builtin_strlen): Handle strnlen.
1139 (strlen_check_and_optimize_stmt): Same.
1140 * doc/extend.texi (Other Builtins): Document strnlen.
1142 2018-06-18 Maya Rashish <coypu@sdf.org>
1144 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
1145 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
1146 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
1148 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
1150 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
1152 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1154 * tree.c (escaped_string::escape): Replace cast to char * by
1155 const_cast<char *> (unescaped).
1157 2018-06-18 Nick Clifton <nickc@redhat.com>
1160 * tree.c (escaped_string): New class. Converts an unescaped
1161 string into its escaped equivalent.
1162 (warn_deprecated_use): Use the new class to convert the
1163 deprecation message, if present.
1164 (test_escaped_strings): New self test.
1165 (test_c_tests): Add test_escaped_strings.
1166 * doc/extend.texi (deprecated): Add a note that the
1167 deprecation message is affected by the -fmessage-length
1168 option, and that control characters will be escaped.
1169 (#pragma GCC error): Document this pragma.
1170 (#pragma GCC warning): Likewise.
1171 * doc/invoke.texi (-fmessage-length): Document this option's
1172 effect on the #warning and #error preprocessor directives and
1173 the deprecated attribute.
1175 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
1177 * tree.c (decl_value_expr_lookup): Revert latest change.
1178 (decl_value_expr_insert): Likewise.
1180 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
1182 * gimplify.c (nonlocal_vlas): Delete.
1183 (nonlocal_vla_vars): Likewise.
1184 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
1186 (gimplify_body): Do not create and destroy nonlocal_vlas.
1187 * tree-nested.c: Include diagnostic.h.
1188 (use_pointer_in_frame): Tweak.
1189 (lookup_field_for_decl): Add assertion and declare the transformation.
1190 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
1191 internal error when the reference is in a wrong context. Do not
1192 create a debug decl by default.
1193 (note_nonlocal_block_vlas): Delete.
1194 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
1195 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
1196 create a debug decl by default.
1197 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
1198 call is in a wrong context.
1199 (fixup_vla_decls): New function.
1200 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
1201 debug variables were created.
1202 * tree.c (decl_value_expr_lookup): Add checking assertion.
1203 (decl_value_expr_insert): Likewise.
1205 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1208 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
1209 * tree-scalar-evolution.c (interpret_expr): Likewise.
1210 (expression_expensive_p): Likewise.
1211 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
1212 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
1213 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
1214 (ssa_defined_by_minus_one_stmt_p): New.
1216 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1219 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
1220 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
1221 * dojump.c (do_jump): Likewise.
1222 * expr.c (expand_expr_real_2): Check operand type's sign.
1223 * fold-const.c (const_unop): Handle ABSU_EXPR.
1224 (fold_abs_const): Likewise.
1225 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
1226 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
1227 (strip_sign_op_1): Likesise.
1228 * match.pd: Add new pattern to generate ABSU_EXPR.
1229 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
1230 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1231 * tree-eh.c (operation_could_trap_helper_p): Likewise.
1232 * tree-inline.c (estimate_operator_cost): Likewise.
1233 * tree-pretty-print.c (dump_generic_node): Likewise.
1234 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
1235 * tree.def (ABSU_EXPR): New.
1237 2018-06-16 Jakub Jelinek <jakub@redhat.com>
1240 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
1241 documented as preserved for backward compatibility only.
1242 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
1244 PR rtl-optimization/86108
1245 * bb-reorder.c (create_forwarder_block): Renamed to ...
1246 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
1247 jump from new landing pad to the second part.
1248 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
1251 2018-06-15 Jakub Jelinek <jakub@redhat.com>
1254 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
1255 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
1256 Only call store_expr for halves if the mode is the same.
1259 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
1260 Fix up comment formatting.
1262 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
1264 * typed-splay-tree.h (typed_splay_tree::remove): New function.
1265 (typed_splay_tree::closure,
1266 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
1267 (typed_splay_tree::typed_splay_tree,
1268 typed_splay_tree::operator =): Declared private.
1269 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
1270 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
1271 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
1272 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
1273 typed_splay_tree::splay_tree_splay,
1274 typed_splay_tree::splay_tree_foreach_helper,
1275 typed_splay_tree::splay_tree_insert,
1276 typed_splay_tree::splay_tree_remove,
1277 typed_splay_tree::splay_tree_lookup,
1278 typed_splay_tree::splay_tree_predecessor,
1279 typed_splay_tree::splay_tree_successor,
1280 typed_splay_tree::splay_tree_min,
1281 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
1282 (typed_splay_tree::root, typed_splay_tree::comp,
1283 typed_splay_tree::delete_key,
1284 typed_splay_tree::delete_value): New data members.
1285 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
1286 typed_splay_tree::remove.
1288 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
1290 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
1291 -mginv and -mno-ginv to the assembler.
1292 * config/mips/mips.opt (-mcrc): New option.
1294 * doc/invoke.text (-mcrc): Document.
1297 2018-06-15 Nick Clifton <nickc@redhat.com>
1300 * tree.c (escaped_string): New class. Converts an unescaped
1301 string into its escaped equivalent.
1302 (warn_deprecated_use): Use the new class to convert the
1303 deprecation message, if present.
1304 (test_escaped_strings): New self test.
1305 (test_c_tests): Add test_escaped_strings.
1306 * doc/extend.texi (deprecated): Add a note that the
1307 deprecation message is affected by the -fmessage-length
1308 option, and that control characters will be escaped.
1309 (#pragma GCC error): Document this pragma.
1310 (#pragma GCC warning): Likewise.
1311 * doc/invoke.texi (-fmessage-length): Document this option's
1312 effect on the #warning and #error preprocessor directives and
1313 the deprecated attribute.
1315 2018-06-15 Richard Biener <rguenther@suse.de>
1317 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
1318 here, also noting vector size used.
1319 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
1320 size used in MSG_OPTIMIZED_LOCATIONS dump.
1321 (pass_slp_vectorize::execute): Adjust.
1323 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
1326 * config/arc/arc.c (arc_return_address_register): Fix
1329 2018-06-15 Richard Biener <rguenther@suse.de>
1332 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
1333 leave useless conversion stripping to force_gimple_operand_gsi.
1334 (gimplify_build2): Likewise.
1335 (gimplify_build1): Likewise.
1337 2018-06-15 Richard Biener <rguenther@suse.de>
1340 * tree-cfg.c (move_stmt_op): unshare invariant addresses
1341 before adjusting their block.
1343 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
1345 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
1346 multilibs for *-*-rtems*.
1347 * config/riscv/t-rtems: New file.
1349 2018-06-14 Jakub Jelinek <jakub@redhat.com>
1352 * match.pd ((A +- CST1) +- CST2): Punt if last resort
1353 unsigned_type_for returns NULL.
1356 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
1357 subregs of multi-word pseudos unless the float mode has word size.
1359 2018-06-14 Richard Biener <rguenther@suse.de>
1362 * tree-vect-generic.c (build_word_mode_vector_type): Remove
1363 duplicate and harmful type_hash_canon.
1364 * tree.c (type_hash_canon): Assert we didn't find ourselves.
1366 2018-06-14 Richard Biener <rguenther@suse.de>
1369 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
1372 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
1374 * config/rtems.h (STDINT_LONG32): Define.
1376 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
1377 Prachi Godbole <prachi.godbole@imgtec.com>
1379 * config/mips/mips-cpus.def: Define P6600.
1380 * config/mips/mips-tables.opt: Regenerate.
1381 * config/mips/mips.c (mips_ucbranch_type): New enum.
1382 (mips_rtx_cost_data): Add support for P6600.
1383 (mips_issue_rate): Likewise.
1384 (mips_multipass_dfa_lookahead): Likewise.
1385 (mips_avoid_hazard): Likewise.
1386 (mips_reorg_process_insns): Likewise.
1387 (mips_classify_branch_p6600): New function.
1388 * config/mips/mips.h (TUNE_P6600): New define.
1389 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
1390 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
1391 * config/mips/mips.md: Include p6600.md.
1392 (processor): Add p6600.
1393 * config/mips/p6600.md: New file.
1394 * doc/invoke.texi: Add p6600 to supported architectures.
1396 2018-06-13 Martin Sebor <msebor@redhat.com>
1398 PR tree-optimization/86114
1399 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
1401 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
1403 2018-06-13 Richard Biener <rguenther@suse.de>
1405 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
1406 Properly set vector type of the intermediate stmt.
1407 * tree-vect-stmts.c (vectorizable_operation): The destination
1408 var always has vectype_out type.
1410 2018-06-13 Jeff Law <law@redhat.com>
1412 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
1413 integer 0 for argument to print_rtl_with_bb.
1414 (rl78_reorg): Likewise.
1416 2018-06-13 David Malcolm <dmalcolm@redhat.com>
1418 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
1419 from rtx to rtx_insn *.
1420 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
1422 (add_sched_insns_for_speculation): Likewise for local "target",
1423 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1424 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
1425 from rtx_insn ** to rtx_code_label **.
1426 (reorg_emit_nops): Likewise.
1427 (c6x_reorg): Likewise for local "call_labels".
1428 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
1430 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
1431 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
1432 the loops over LABEL_REFS.
1433 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
1435 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1436 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
1437 (split_branches): Strengthen local "olabel" from rtx to
1438 rtx_insn *, adding a safe_as_a cast.
1439 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
1441 (add_insn_after): Likewise for first two params.
1442 (add_insn_before): Likewise.
1443 (remove_insn): Likewise for param.
1444 (emit_pattern_before_noloc): Likewise for second and third params.
1445 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
1446 (emit_call_insn_before_noloc): Likewise.
1447 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
1449 (emit_barrier_before): Likewise.
1450 (emit_label_before): Strengthen "label" param from "rtx" to
1451 "rtx_code_label *". Strengthen "before" param from "rtx" to
1453 (emit_insn_after_1): Strengthen "after" param from "rtx" to
1455 (emit_pattern_after_noloc): Likewise.
1456 (emit_insn_after_noloc): Likewise.
1457 (emit_jump_insn_after_noloc): Likewise.
1458 (emit_call_insn_after_noloc): Likewise.
1459 (emit_debug_insn_after_noloc): Likewise.
1460 (emit_barrier_after): Likewise.
1461 (emit_label_after): Likewise for both params.
1462 (emit_pattern_after_setloc): Likewise for "after" param. Convert
1463 "loc" param from "int" to "location_t".
1464 (emit_insn_after_setloc): Likewise.
1465 (emit_jump_insn_after_setloc): Likewise.
1466 (emit_call_insn_after_setloc): Likewise.
1467 (emit_debug_insn_after_setloc): Likewise.
1468 (emit_pattern_before_setloc): Likewise for "before" param. Convert
1469 "loc" param from "int" to "location_t".
1470 (emit_pattern_before): Convert NULL_RTX to NULL.
1471 (emit_insn_before_setloc): Convert "loc" param from "int" to
1473 (emit_jump_insn_before_setloc): Likewise.
1474 (emit_call_insn_before_setloc): Likewise.
1475 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
1476 rtx_insn *. Convert "loc" param from "int" to "location_t".
1477 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
1478 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
1479 Convert 3rd param from "int" to "location_t".
1480 (emit_barrier_before, emit_barrier_after, next_real_insn):
1481 Strengthen param from rtx to rtx_insn *.
1482 (emit_label_before): Strengthen 1st param from "rtx" to
1483 "rtx_code_label *". Strengthen 2nd param from "rtx" to
1485 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
1486 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
1487 Strengthen 2nd param from "rtx" to "rtx_insn *".
1488 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
1489 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
1490 Likewise. Convert 3rd param from "int" to "location_t".
1491 (emit_label_after): Strengthen 1st param from "rtx" to
1493 (next_real_insn, remove_insn): Strengthen param from "rtx" to
1495 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
1496 from "rtx" to "rtx_insn *".
1498 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
1500 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
1501 bodies streamed in with -Q.
1502 * dumpfile.c (dump_files): Add lto-stream-out dump file.
1503 * dumpfile.h (tree_dump_index): Add lto_stream_out.
1504 * gimple-streamer-out.c: Include gimple-pretty-print.h
1505 (output_bb): Dump stmts streamed.
1506 * lto-section-out.c: Include print-tree.h
1507 (lto_begin_section): Dump sections created.
1508 (lto_output_decl_index): Dump decl encoded.
1509 * lto-streamer-out.c: Include print-tree.h
1510 (create_output_block): Dump output block created.
1511 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
1512 (output_function): Dump function output.
1513 (output_constructor): Dump constructor streamed.
1514 (write_global_stream): Output indexes encoded.
1515 (produce_asm_for_decls): Dump streams encoded.
1516 * lto-streamer.c (streamer_dump_file): New global var.
1517 * lto-streamer.h (streamer_dump_file): Declare.
1518 * passes.c (ipa_write_summaries): Initialize streamer dump.
1519 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1522 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1525 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1526 offsets for register save directives. Emit a second batch of save
1527 directives, if need be, when the function accesses prior frames.
1529 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1531 * config/arc/fpu.md (fmasf4): Force operand to register.
1532 (fnmasf4): Likewise.
1534 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1536 * config/arc/arc-protos.h (arc_pad_return): Remove.
1537 * config/arc/arc.c (machine_function): Remove force_short_suffix
1539 (arc_print_operand): Adjust printing of '&'.
1540 (arc_verify_short): Remove conditional printing of short suffix.
1541 (arc_final_prescan_insn): Remove reference to size_reason.
1542 (pad_return): New function.
1543 (arc_reorg): Call pad_return.
1544 (arc_pad_return): Remove.
1545 (arc_init_machine_status): Remove reference to force_short_suffix.
1546 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
1547 (attr length): When attribute iscompact is true force to 2
1548 regardless; in the case of maybe check if we want to force the
1549 instruction to have 4 bytes length.
1550 (nopv): Change it to generate 4 byte long nop as well.
1551 (blockage): New pattern.
1552 (simple_return): Remove call to arc_pad_return.
1553 (p_return_i): Likewise.
1555 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1557 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1559 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1561 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1564 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1566 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1567 for ARC700 and ARCv2.
1569 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1572 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1573 operands[2] instead of operands[1].
1576 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1578 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1579 case, check whether the outer register overlaps an unallocatable
1580 register, not just whether it fits the required class.
1582 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1584 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1586 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1587 (memrefs_conflict_p): Likewise.
1588 (init_alias_analysis): Likewise.
1589 * cfgexpand.c (expand_debug_expr): Likewise.
1590 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1591 * cse.c (fold_rtx): Likewise.
1592 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1593 * expr.c (emit_block_move_hints): Likewise.
1594 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1595 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1596 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1597 (emit_group_store): Likewise.
1598 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1599 to read the PRE/POST_MODIFY increment.
1600 * calls.c (store_one_arg): Use strip_offset.
1601 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1603 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1605 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1606 (simplify_binary_operation_1): Extend CONST_INT handling to
1608 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1609 than a HOST_WIDE_INT.
1610 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1612 (adjust_mems, add_stores): Update accodingly.
1613 (vt_canonicalize_addr): Track polynomial offsets.
1614 (emit_note_insn_var_location): Likewise.
1615 (vt_add_function_parameter): Likewise.
1616 (vt_initialize): Likewise.
1618 2018-06-12 Jeff Law <law@redhat.com>
1620 * config.gcc (alpha*-*-freebsd*): Remove.
1621 * config/alpha/freebsd.h: Remove.
1623 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1626 * spellcheck-tree.c (levenshtein_distance): Rename to...
1627 (get_edit_distance): ...this, and update for underlying renaming.
1628 * spellcheck-tree.h (levenshtein_distance): Rename to...
1629 (get_edit_distance): ...this.
1630 * spellcheck.c (levenshtein_distance): Rename to...
1631 (get_edit_distance): ...this. Convert from Levenshtein distance
1632 to Damerau-Levenshtein distance by supporting transpositions of
1633 adjacent characters. Rename "v1" to "v_next" and "v0" to
1635 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1636 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1637 update for underlying renaming.
1638 (selftest::levenshtein_distance_unit_test): Rename to...
1639 (selftest::test_get_edit_distance_unit): ...this, and update for
1640 underlying renaming.
1641 (selftest::test_find_closest_string): Add example from PR 69968
1642 where transposition helps
1643 (selftest::test_metric_conditions): Update for renaming.
1644 (selftest::test_metric_conditions): Likewise.
1645 (selftest::spellcheck_c_tests): Likewise.
1646 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1647 (get_edit_distance): ...this.
1648 (best_match::consider): Update for renaming.
1650 2018-06-12 Martin Sebor <msebor@redhat.com>
1652 PR tree-optimization/85259
1653 * builtins.c (compute_objsize): Handle constant offsets.
1654 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1655 true iff a warning has been issued.
1656 * gimple.h (gimple_nonartificial_location): New function.
1657 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1658 gimple_nonartificial_location and handle -Wno-system-headers.
1659 (handle_builtin_stxncpy): Same.
1661 2018-06-12 Martin Sebor <msebor@redhat.com>
1664 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1666 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1668 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1669 BUILTIN_VEC_XST entries for pointer to double and long long.
1671 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1674 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1676 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1679 2018-06-12 Martin Liska <mliska@suse.cz>
1681 * doc/options.texi: Document IntegerRange.
1683 2018-06-12 Martin Liska <mliska@suse.cz>
1685 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1686 * opt-functions.awk: Handle Deprecated flag.
1687 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1689 (read_cmdline_option): Report warning for a deprecated option.
1690 * opts.h (struct cl_option): Add new field cl_deprecated.
1691 (CL_ERR_DEPRECATED): New.
1693 2018-06-12 Martin Liska <mliska@suse.cz>
1695 * doc/options.texi: Document Deprecated option flag.
1697 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1699 * config/arc/arc-arch.h (arc_extras): New enum.
1700 (arc_cpu_t):Add field extra.
1701 (arc_cpu_types): Consider the extras.
1702 * config/arc/arc-cpus.def: Add extras info.
1703 * config/arc/arc-opts.h (processor_type): Consider extra field.
1704 * config/arc/arc.c (arc_override_options): Handle extra field.
1706 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1708 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1709 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1711 * config/arc/arc.c (arc_init): Update pic warning.
1712 (irq_range): Update irq range parsing warnings.
1713 (arc_override_options): Update various warning messages.
1714 (arc_handle_aux_attribute): Likewise.
1716 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1718 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1720 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1722 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1723 compiler messages directives.
1725 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1727 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1728 * config/mips/mips-tables.opt: Regenerate.
1729 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1731 * doc/invoke.texi: Document -march=i6500.
1733 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1735 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1736 (i6400_gpmul): Add cpu_unit.
1737 (i6400_gpdiv): Likewise.
1738 (i6400_msa_add_d): Update reservations.
1739 (i6400_msa_int_add) Likewise.
1740 (i6400_msa_short_logic3) Likewise.
1741 (i6400_msa_short_logic2) Likewise.
1742 (i6400_msa_short_logic) Likewise.
1743 (i6400_msa_move) Likewise.
1744 (i6400_msa_cmp) Likewise.
1745 (i6400_msa_short_float2) Likewise.
1746 (i6400_msa_div_d) Likewise.
1747 (i6400_msa_long_logic1) Likewise.
1748 (i6400_msa_long_logic2) Likewise.
1749 (i6400_msa_mult) Likewise.
1750 (i6400_msa_long_float2) Likewise.
1751 (i6400_msa_long_float4) Likewise.
1752 (i6400_msa_long_float5) Likewise.
1753 (i6400_msa_long_float8) Likewise.
1754 (i6400_fpu_fadd): Include frint type.
1755 (i6400_fpu_store): New define_insn_reservation.
1756 (i6400_fpu_load): Likewise.
1757 (i6400_fpu_move): Likewise.
1758 (i6400_fpu_fcmp): Likewise.
1759 (i6400_fpu_fmadd): Likewise.
1760 (i6400_int_mult): Include imul3nc type and update reservation.
1761 (i6400_int_div): Include idiv3 type and update reservation.
1762 (i6400_int_load): Update to check type not move_type.
1763 (i6400_int_store): Likewise.
1764 (i6400_int_prefetch): Set zero latency.
1766 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1768 * gcc.c: Document new %@{...} sequence.
1769 (LINK_COMMAND_SPEC): Use it for the -L switches.
1770 (cpp_unique_options): Use it for the -I switches.
1771 (at_file_argbuf): New global variable.
1772 (in_at_file): Likewise.
1773 (alloc_args): Create at_file_argbuf.
1774 (clear_args): Truncate at_file_argbuf.
1775 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1776 (open_at_file): New function.
1777 (close_at_file): Likewise.
1778 (create_at_file): Delete.
1779 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1782 (validate_switches_from_spec): Deal with %@{...} sequence.
1783 (validate_switches): Likewise.
1784 (driver::finalize): Call clear_args.
1786 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1788 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1790 2018-06-11 Martin Sebor <msebor@redhat.com>
1792 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1793 (Wno-class-memaccess): Add @opindex.
1794 (Wno-templates, Wno-multiple-inheritance): Same.
1795 (Wno-virtual-inheritance, Wno-namespaces): Same.
1796 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1797 (Wno-format-overflow, Wno-format-truncation): Same.
1798 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1799 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1800 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1801 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1802 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1803 (Wno-misspelled-isr): Same.
1805 2018-06-11 Martin Sebor <msebor@redhat.com>
1807 * PR tree-optimization/86083
1808 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1810 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1812 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1814 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1817 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1818 on the correct operand.
1819 (*movdi_internal64): Ditto.
1821 2018-06-11 Martin Liska <mliska@suse.cz>
1823 PR tree-optimization/86089
1824 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1826 2018-06-11 Julia Koval <julia.koval@intel.com>
1828 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1829 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1830 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1832 2018-06-11 Olivier Hainque <hainque@adacore.com>
1834 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1835 for Ada with strict dwarf2.
1837 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1840 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1843 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1845 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1847 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1849 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1852 2018-06-08 Martin Liska <mliska@suse.cz>
1854 * tree-cfg.h (debug_function): Fix argument type to match
1857 2018-06-08 Martin Liska <mliska@suse.cz>
1859 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1860 Remove usage of MPX-related (and removed) fields.
1861 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1863 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1865 * cfg.c (debug): Use TDF_NONE rather than 0.
1866 * cfghooks.c (debug): Likewise.
1867 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1868 (struct dump_option_value_info): Convert to...
1869 (struct kv_pair): ...this template type.
1870 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1872 (optinfo_verbosity_options): Likewise.
1873 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1875 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1876 than int for "optgroup_flags" param.
1877 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1879 (dump_dec): Likewise.
1880 (dump_finish): Use TDF_NONE rather than 0.
1881 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1882 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1883 than 0. Update for change to option_ptr.
1884 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1885 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1886 0. Update for changes to optinfo_verbosity_options and
1888 (opt_info_switch_p): Convert optgroup_flags from int to
1890 (dump_basic_block): Use dump_flags_t rather than int
1891 for "dump_kind" param.
1892 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1893 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1894 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1895 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1896 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1897 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1898 TDF_NONE): Convert from macros to...
1899 (enum dump_flag): ...this new enum.
1900 (dump_flags_t): Update to use enum.
1901 (operator|, operator&, operator~, operator|=, operator&=):
1902 Implement for dump_flags_t.
1903 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1904 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1905 Convert from macros to...
1906 (enum optgroup_flag): ...this new enum.
1907 (optgroup_flags_t): New typedef.
1908 (operator|, operator|=): Implement for optgroup_flags_t.
1909 (struct dump_file_info): Convert field "alt_flags" to
1910 dump_flags_t. Convert field "optgroup_flags" to
1912 (dump_basic_block): Use dump_flags_t rather than int for param.
1913 (dump_generic_expr_loc): Likewise.
1914 (dump_dec): Likewise.
1915 (dump_register): Convert param "optgroup_flags" to
1917 (opt_info_enable_passes): Likewise.
1918 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1920 * gimple-pretty-print.c (debug): Likewise.
1921 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1922 (merged_store_group::apply_stores): Likewise.
1923 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1924 * gimple.c (verify_gimple_pp): Likewise.
1925 * graphite-poly.c (print_pbb_body): Likewise.
1926 * passes.c (pass_manager::register_one_dump_file): Convert
1927 local "optgroup_flags" to optgroup_flags_t.
1928 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1930 (debug_body): Likewise.
1931 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1932 to optgroup_flags_t.
1933 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1935 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1936 (convert_mult_to_fma): Likewise.
1937 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1938 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1939 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1940 "dump_kind" to dump_flags_t.
1942 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1944 * config/rs6000/rs6000.c (min, max): Delete.
1946 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1948 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1949 -mabi=spe and -mabi=no-spe.
1951 2018-06-08 Martin Liska <mliska@suse.cz>
1953 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1954 where we expect an existing summary.
1956 2018-06-08 Martin Liska <mliska@suse.cz>
1958 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1959 * ipa-inline.h (estimate_edge_growth): Likewise.
1961 2018-06-08 Martin Liska <mliska@suse.cz>
1963 * cgraph.c (function_version_hasher::hash): Use
1964 cgraph_node::get_uid ().
1965 (function_version_hasher::equal):
1966 * cgraph.h (cgraph_node::get_uid): New method.
1967 * ipa-inline.c (update_caller_keys): Use
1968 cgraph_node::get_uid ().
1969 (update_callee_keys): Likewise.
1970 * ipa-utils.c (searchc): Likewise.
1971 (ipa_reduced_postorder): Likewise.
1972 * lto-cgraph.c (input_node): Likewise.
1973 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1974 * symbol-summary.h (symtab_insertion): Likewise.
1975 (symtab_removal): Likewise.
1976 (symtab_duplication): Likewise.
1977 * tree-pretty-print.c (dump_function_header): Likewise.
1978 * tree-sra.c (convert_callers_for_node): Likewise.
1980 2018-06-08 Martin Liska <mliska@suse.cz>
1982 * cgraph.c (symbol_table::create_edge): Always assign a new
1984 (symbol_table::free_edge): Do not recycle numbers.
1985 * cgraph.h (cgraph_edge::get): New method.
1986 * symbol-summary.h (symtab_removal): Use it.
1987 (symtab_duplication): Likewise.
1988 (call_summary::hashable_uid): Remove.
1990 2018-06-08 Martin Liska <mliska@suse.cz>
1992 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1993 (initialize_growth_caches): Remove.
1994 (free_growth_caches): Likewise.
1995 (do_estimate_edge_time): Use edge_growth_cache.
1996 (do_estimate_edge_size): Likewise.
1997 (do_estimate_edge_hints): Likewise.
1998 * ipa-inline.c (reset_edge_caches): Likewise.
1999 (recursive_inlining): Likewise.
2000 (inline_small_functions): Likewise.
2001 * ipa-inline.h (initialize_growth_caches): Remove.
2002 (estimate_edge_size): Likewise.
2003 (estimate_edge_time): Likewise.
2004 (estimate_edge_hints): Likewise.
2005 (reset_edge_growth_cache): Likewise.
2006 * symbol-summary.h (call_summary::remove): New method.
2008 2018-06-08 Martin Liska <mliska@suse.cz>
2010 * ipa-cp.c (class edge_clone_summary): New summary.
2011 (grow_edge_clone_vectors): Remove.
2012 (ipcp_edge_duplication_hook): Remove.
2013 (class edge_clone_summary_t): New call_summary class.
2014 (ipcp_edge_removal_hook): Remove.
2015 (edge_clone_summary_t::duplicate): New function.
2016 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
2017 (create_specialized_node): Likewise.
2018 (ipcp_driver): Initialize edge_clone_summaries and do not
2021 2018-06-08 Martin Liska <mliska@suse.cz>
2023 * symbol-summary.h (get): New function.
2024 (call_summary::m_initialize_when_cloning): New class member.
2026 2018-06-08 Martin Liska <mliska@suse.cz>
2028 * cgraph.c (cgraph_node::remove): Do not recycle uid.
2029 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
2030 (symbol_table::allocate_cgraph_symbol): Do not set uid.
2031 * passes.c (uid_hash_t): Record removed_nodes by their uids.
2032 (remove_cgraph_node_from_order): Use the removed_nodes set.
2033 (do_per_function_toporder): Likwise.
2034 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
2035 instead of summary_uid.
2036 (symtab_removal): Likewise.
2037 (symtab_duplication): Likewise.
2039 2018-06-08 Martin Liska <mliska@suse.cz>
2041 * ipa-cp.c (ipcp_store_bits_results): Use
2042 ipcp_transformation_sum.
2043 (ipcp_store_vr_results): Likewise.
2044 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
2046 (ipcp_transformation_initialize): ... this.
2047 (ipa_set_node_agg_value_chain):
2048 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
2049 (write_ipcp_transformation_info): Likewise.
2050 (read_ipcp_transformation_info): Likewise.
2051 (ipcp_update_bits): Likewise.
2052 (ipcp_update_vr): Likewise.
2053 (ipcp_transform_function): Likewise.
2054 * ipa-prop.h: Rename ipcp_transformation_summary to
2055 ipcp_transformation.
2056 (class ipcp_transformation_t): New function summary.
2057 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
2058 (ipa_get_agg_replacements_for_node): Likewise.
2060 2018-06-08 Martin Liska <mliska@suse.cz>
2062 * ipa-pure-const.c (struct funct_state_d): Do it class instead
2064 (class funct_state_summary_t): New function_summary class.
2065 (has_function_state): Remove.
2066 (get_function_state): Likewise.
2067 (set_function_state): Likewise.
2068 (add_new_function): Likewise.
2069 (funct_state_summary_t::insert): New function.
2070 (duplicate_node_data): Remove.
2071 (remove_node_data): Remove.
2072 (funct_state_summary_t::duplicate): New function.
2073 (register_hooks): Create new funct_state_summaries.
2074 (pure_const_generate_summary): Use it.
2075 (pure_const_write_summary): Likewise.
2076 (pure_const_read_summary): Likewise.
2077 (propagate_pure_const): Likewise.
2078 (propagate_nothrow): Likewise.
2079 (dump_malloc_lattice): Likewise.
2080 (propagate_malloc): Likewise.
2081 (execute): Do not register hooks, just remove summary
2083 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
2086 2018-06-08 Martin Liska <mliska@suse.cz>
2088 * ipa-reference.c (remove_node_data): Remove.
2089 (duplicate_node_data): Likewise.
2090 (class ipa_ref_var_info_summary_t): New class.
2091 (class ipa_ref_opt_summary_t): Likewise.
2092 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
2093 (get_reference_optimization_summary): Use
2094 ipa_ref_opt_sum_summaries.
2095 (set_reference_vars_info): Remove.
2096 (set_reference_optimization_summary): Likewise.
2097 (ipa_init): Create summaries.
2098 (init_function_info): Use function summary.
2099 (ipa_ref_opt_summary_t::duplicate): New function.
2100 (ipa_ref_opt_summary_t::remove): New function.
2101 (get_read_write_all_from_node): Fix GNU coding style.
2102 (propagate): Use function summary.
2103 (write_node_summary_p): Fix GNU coding style.
2104 (stream_out_bitmap): Likewise.
2105 (ipa_reference_read_optimization_summary): Use function summary.
2106 (ipa_reference_c_finalize): Do not release hooks.
2108 2018-06-08 Martin Liska <mliska@suse.cz>
2110 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
2111 (analyze_function_body): Extract multiple calls of get_create.
2112 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2113 * ipa-inline.c (recursive_inlining): Use ::get method.
2114 * ipa-inline.h (estimate_edge_growth): Likewise.
2116 2018-06-08 Martin Liska <mliska@suse.cz>
2118 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
2120 (hsa_function_summary::hsa_function_summary): Use the new enum
2122 (hsa_gpu_implementation_p): Use hsa_summaries::get.
2123 * hsa-gen.c (hsa_get_host_function): Likewise.
2124 (get_brig_function_name): Likewise.
2125 * ipa-hsa.c (process_hsa_functions): Likewise.
2126 (ipa_hsa_write_summary): Likewise.
2127 * symbol-summary.h (symtab_duplication): Use ::get function/
2128 (get): New function.
2130 2018-06-08 Martin Liska <mliska@suse.cz>
2132 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
2134 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
2135 (hsa_register_kernel): Likewise.
2136 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
2137 * hsa-gen.c (hsa_get_host_function): Likewise.
2138 (get_brig_function_name): Likewise.
2139 (generate_hsa): Likewise.
2140 (pass_gen_hsail::execute): Likewise.
2141 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
2142 (devirtualization_time_bonus): Likewise.
2143 (ipcp_propagate_stage): Likewise.
2144 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2145 (edge_set_predicate): Likewise.
2146 (evaluate_conditions_for_known_args): Likewise.
2147 (evaluate_properties_for_edge): Likewise.
2148 (ipa_fn_summary::reset): Likewise.
2149 (ipa_fn_summary_t::duplicate): Likewise.
2150 (dump_ipa_call_summary): Likewise.
2151 (ipa_dump_fn_summary): Likewise.
2152 (analyze_function_body): Likewise.
2153 (compute_fn_summary): Likewise.
2154 (estimate_edge_devirt_benefit): Likewise.
2155 (estimate_edge_size_and_time): Likewise.
2156 (estimate_calls_size_and_time): Likewise.
2157 (estimate_node_size_and_time): Likewise.
2158 (inline_update_callee_summaries): Likewise.
2159 (remap_edge_change_prob): Likewise.
2160 (remap_edge_summaries): Likewise.
2161 (ipa_merge_fn_summary_after_inlining): Likewise.
2162 (ipa_update_overall_fn_summary): Likewise.
2163 (read_ipa_call_summary): Likewise.
2164 (inline_read_section): Likewise.
2165 (write_ipa_call_summary): Likewise.
2166 (ipa_fn_summary_write): Likewise.
2167 (ipa_free_fn_summary): Likewise.
2168 * ipa-hsa.c (process_hsa_functions): Likewise.
2169 (ipa_hsa_write_summary): Likewise.
2170 (ipa_hsa_read_section): Likewise.
2171 * ipa-icf.c (sem_function::merge): Likewise.
2172 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2173 (do_estimate_edge_time): Likewise.
2174 (estimate_size_after_inlining): Likewise.
2175 (estimate_growth): Likewise.
2176 (growth_likely_positive): Likewise.
2177 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2178 (inline_call): Likewise.
2179 * ipa-inline.c (caller_growth_limits): Likewise.
2180 (can_inline_edge_p): Likewise.
2181 (can_inline_edge_by_limits_p): Likewise.
2182 (compute_uninlined_call_time): Likewise.
2183 (compute_inlined_call_time): Likewise.
2184 (want_inline_small_function_p): Likewise.
2185 (edge_badness): Likewise.
2186 (update_caller_keys): Likewise.
2187 (update_callee_keys): Likewise.
2188 (recursive_inlining): Likewise.
2189 (inline_small_functions): Likewise.
2190 (inline_to_all_callers_1): Likewise.
2191 (dump_overall_stats): Likewise.
2192 (early_inline_small_functions): Likewise.
2193 (early_inliner): Likewise.
2194 * ipa-inline.h (estimate_edge_growth): Likewise.
2195 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2196 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2197 * ipa-prop.h (IPA_NODE_REF): Likewise.
2198 (IPA_EDGE_REF): Likewise.
2199 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2200 (propagate_malloc): Likewise.
2201 * ipa-split.c (execute_split_functions): Likewise.
2202 * symbol-summary.h: Rename get to get_create.
2204 (get_create): Likewise.
2205 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2207 2018-06-08 Martin Liska <mliska@suse.cz>
2209 * symbol-summary.h (release): Move definition out of class
2211 (symtab_removal): Likewise.
2212 (symtab_duplication): Likewise.
2214 2018-06-08 Martin Liska <mliska@suse.cz>
2216 * symbol-summary.h (function_summary): Move constructor
2217 implementation out of class declaration.
2218 (release): Likewise.
2219 (symtab_insertion): Likewise.
2220 (symtab_removal): Likewise.
2221 (symtab_duplication): Likewise.
2224 2018-06-08 Martin Liska <mliska@suse.cz>
2226 * Makefile.in: Remove support for MPX (macros, related functions,
2227 fields in cgraph_node, ...).
2228 * builtin-types.def (BT_BND): Likewise.
2229 (BT_FN_BND_CONST_PTR): Likewise.
2230 (BT_FN_CONST_PTR_BND): Likewise.
2231 (BT_FN_VOID_PTR_BND): Likewise.
2232 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
2233 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
2234 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
2235 (expand_builtin_mempcpy_with_bounds): Likewise.
2236 (expand_builtin_memset_with_bounds): Likewise.
2237 (expand_builtin_memset_args): Likewise.
2238 (std_expand_builtin_va_start): Likewise.
2239 (expand_builtin): Likewise.
2240 (expand_builtin_with_bounds): Likewise.
2241 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
2242 (DEF_LIB_BUILTIN_CHKP): Likewise.
2243 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
2244 (DEF_CHKP_BUILTIN): Likewise.
2245 (BUILT_IN_MEMCPY): Likewise.
2246 (BUILT_IN_MEMMOVE): Likewise.
2247 (BUILT_IN_MEMPCPY): Likewise.
2248 (BUILT_IN_MEMSET): Likewise.
2249 (BUILT_IN_STPCPY): Likewise.
2250 (BUILT_IN_STRCAT): Likewise.
2251 (BUILT_IN_STRCHR): Likewise.
2252 (BUILT_IN_STRCPY): Likewise.
2253 (BUILT_IN_STRLEN): Likewise.
2254 (BUILT_IN_MEMCPY_CHK): Likewise.
2255 (BUILT_IN_MEMMOVE_CHK): Likewise.
2256 (BUILT_IN_MEMPCPY_CHK): Likewise.
2257 (BUILT_IN_MEMSET_CHK): Likewise.
2258 (BUILT_IN_STPCPY_CHK): Likewise.
2259 (BUILT_IN_STRCAT_CHK): Likewise.
2260 (BUILT_IN_STRCPY_CHK): Likewise.
2261 * calls.c (store_bounds): Likewise.
2262 (emit_call_1): Likewise.
2263 (special_function_p): Likewise.
2264 (maybe_warn_nonstring_arg): Likewise.
2265 (initialize_argument_information): Likewise.
2266 (finalize_must_preallocate): Likewise.
2267 (compute_argument_addresses): Likewise.
2268 (expand_call): Likewise.
2269 * cfgexpand.c (expand_call_stmt): Likewise.
2270 (expand_return): Likewise.
2271 (expand_gimple_stmt_1): Likewise.
2272 (pass_expand::execute): Likewise.
2273 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2274 (cgraph_node::remove): Likewise.
2275 (cgraph_node::dump): Likewise.
2276 (cgraph_node::verify_node): Likewise.
2277 * cgraph.h (chkp_function_instrumented_p): Likewise.
2278 (symtab_node::get_alias_target): Likewise.
2279 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
2280 (cgraph_local_p): Likewise.
2281 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
2282 (cgraph_edge::rebuild_references): Likewise.
2283 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
2284 (walk_polymorphic_call_targets): Likewise.
2285 (cgraph_node::expand_thunk): Likewise.
2286 (symbol_table::output_weakrefs): Likewise.
2287 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
2288 (ix86_handle_option): Likewise.
2289 * config/i386/constraints.md: Likewise.
2290 * config/i386/i386-builtin-types.def (BND): Likewise.
2294 * config/i386/i386-builtin.def (BDESC_END): Likewise.
2295 (BDESC_FIRST): Likewise.
2297 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
2298 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
2299 * config/i386/i386.c (enum reg_class): Likewise.
2300 (ix86_target_string): Likewise.
2301 (ix86_option_override_internal): Likewise.
2302 (ix86_conditional_register_usage): Likewise.
2303 (ix86_valid_target_attribute_inner_p): Likewise.
2304 (ix86_set_indirect_branch_type): Likewise.
2305 (ix86_set_current_function): Likewise.
2306 (ix86_function_arg_regno_p): Likewise.
2307 (init_cumulative_args): Likewise.
2308 (ix86_function_arg_advance): Likewise.
2309 (ix86_function_arg): Likewise.
2310 (ix86_pass_by_reference): Likewise.
2311 (ix86_function_value_regno_p): Likewise.
2312 (ix86_function_value_1): Likewise.
2313 (ix86_function_value_bounds): Likewise.
2314 (ix86_return_in_memory): Likewise.
2315 (ix86_setup_incoming_vararg_bounds): Likewise.
2316 (ix86_va_start): Likewise.
2317 (indirect_thunk_need_prefix): Likewise.
2318 (print_reg): Likewise.
2319 (ix86_print_operand): Likewise.
2320 (ix86_expand_call): Likewise.
2321 (ix86_output_function_return): Likewise.
2322 (reg_encoded_number): Likewise.
2323 (BDESC_VERIFYS): Likewise.
2324 (ix86_init_mpx_builtins): Likewise.
2325 (ix86_init_builtins): Likewise.
2326 (ix86_emit_cmove): Likewise.
2327 (ix86_emit_move_max): Likewise.
2328 (ix86_expand_builtin): Likewise.
2329 (ix86_builtin_mpx_function): Likewise.
2330 (ix86_get_arg_address_for_bt): Likewise.
2331 (ix86_load_bounds): Likewise.
2332 (ix86_store_bounds): Likewise.
2333 (ix86_load_returned_bounds): Likewise.
2334 (ix86_store_returned_bounds): Likewise.
2335 (ix86_class_likely_spilled_p): Likewise.
2336 (ix86_hard_regno_mode_ok): Likewise.
2337 (x86_order_regs_for_local_alloc): Likewise.
2338 (ix86_mitigate_rop): Likewise.
2339 (ix86_bnd_prefixed_insn_p): Likewise.
2340 (ix86_mpx_bound_mode): Likewise.
2341 (ix86_make_bounds_constant): Likewise.
2342 (ix86_initialize_bounds): Likewise.
2343 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
2344 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
2345 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
2346 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
2347 (TARGET_CHKP_BOUND_MODE): Likewise.
2348 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
2349 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
2350 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
2351 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
2352 * config/i386/i386.h (TARGET_MPX): Likewise.
2353 (TARGET_MPX_P): Likewise.
2354 (VALID_BND_REG_MODE): Likewise.
2355 (FIRST_BND_REG): Likewise.
2356 (LAST_BND_REG): Likewise.
2357 (enum reg_class): Likewise.
2358 (BND_REG_P): Likewise.
2359 (BND_REGNO_P): Likewise.
2360 (BNDmode): Likewise.
2361 (ADJUST_INSN_LENGTH): Likewise.
2362 * config/i386/i386.md: Likewise.
2363 * config/i386/i386.opt: Likewise.
2364 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
2365 (defined): Likewise.
2366 (LINK_MPX): Likewise.
2367 (MPX_SPEC): Likewise.
2368 (LIBMPX_SPEC): Likewise.
2369 (LIBMPXWRAPPERS_SPEC): Likewise.
2370 (CHKP_SPEC): Likewise.
2371 * config/i386/predicates.md: Likewise.
2372 * dbxout.c (dbxout_type): Likewise.
2373 * doc/extend.texi: Likewise.
2374 * doc/invoke.texi: Likewise.
2375 * doc/md.texi: Likewise.
2376 * doc/tm.texi: Likewise.
2377 * doc/tm.texi.in: Likewise.
2378 * dwarf2out.c (is_base_type): Likewise.
2379 (gen_formal_types_die): Likewise.
2380 (gen_subprogram_die): Likewise.
2381 (gen_type_die_with_usage): Likewise.
2382 (gen_decl_die): Likewise.
2383 (dwarf2out_late_global_decl): Likewise.
2384 * expr.c (expand_assignment): Likewise.
2385 (emit_storent_insn): Likewise.
2386 (store_expr_with_bounds): Likewise.
2387 (store_expr): Likewise.
2388 (expand_expr_real_1): Likewise.
2389 * expr.h (store_expr_with_bounds): Likewise.
2390 * function.c (use_register_for_decl): Likewise.
2391 (struct bounds_parm_data): Likewise.
2392 (assign_parms_augmented_arg_list): Likewise.
2393 (assign_parm_find_entry_rtl): Likewise.
2394 (assign_parm_is_stack_parm): Likewise.
2395 (assign_parm_load_bounds): Likewise.
2396 (assign_bounds): Likewise.
2397 (assign_parms): Likewise.
2398 (expand_function_start): Likewise.
2399 * gcc.c (CHKP_SPEC): Likewise.
2400 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
2401 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
2402 (wrestrict_dom_walker::check_call): Likewise.
2403 * gimple.c (gimple_build_call_from_tree): Likewise.
2404 * gimple.h (enum gf_mask): Likewise.
2405 (gimple_call_with_bounds_p): Likewise.
2406 (gimple_call_set_with_bounds): Likewise.
2407 * gimplify.c (gimplify_init_constructor): Likewise.
2408 * ipa-cp.c (initialize_node_lattices): Likewise.
2409 (propagate_constants_across_call): Likewise.
2410 (find_more_scalar_values_for_callers_subset): Likewise.
2411 * ipa-hsa.c (process_hsa_functions): Likewise.
2412 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
2413 * ipa-icf.c (sem_function::merge): Likewise.
2414 * ipa-inline.c (early_inliner): Likewise.
2415 * ipa-pure-const.c (warn_function_noreturn): Likewise.
2416 (warn_function_cold): Likewise.
2417 (propagate_pure_const): Likewise.
2418 * ipa-ref.h (enum GTY): Likewise.
2419 * ipa-split.c (find_retbnd): Likewise.
2420 (consider_split): Likewise.
2421 (split_function): Likewise.
2422 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2423 * ipa.c (walk_polymorphic_call_targets): Likewise.
2424 (symbol_table::remove_unreachable_nodes): Likewise.
2425 (process_references): Likewise.
2426 (cgraph_build_static_cdtor_1): Likewise.
2427 * lto-cgraph.c (lto_output_node): Likewise.
2428 (output_refs): Likewise.
2429 (compute_ltrans_boundary): Likewise.
2430 (input_overwrite_node): Likewise.
2431 (input_node): Likewise.
2432 (input_cgraph_1): Likewise.
2433 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
2434 * passes.c (pass_manager::execute_early_local_passes): Likewise.
2435 (class pass_chkp_instrumentation_passes): Likewise.
2436 (make_pass_chkp_instrumentation_passes): Likewise.
2437 * passes.def: Likewise.
2438 * rtl.h (struct GTY): Likewise.
2439 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
2440 * stor-layout.c (layout_type): Likewise.
2441 * symtab.c: Likewise.
2442 * target.def: Likewise.
2443 * targhooks.c (default_chkp_bound_type): Likewise.
2444 (default_chkp_bound_mode): Likewise.
2445 (default_builtin_chkp_function): Likewise.
2446 (default_chkp_function_value_bounds): Likewise.
2447 (default_chkp_make_bounds_constant): Likewise.
2448 (default_chkp_initialize_bounds): Likewise.
2449 * targhooks.h (default_chkp_bound_type): Likewise.
2450 (default_chkp_bound_mode): Likewise.
2451 (default_builtin_chkp_function): Likewise.
2452 (default_chkp_function_value_bounds): Likewise.
2453 (default_chkp_make_bounds_constant): Likewise.
2454 (default_chkp_initialize_bounds): Likewise.
2455 * toplev.c (compile_file): Likewise.
2456 (process_options): Likewise.
2457 * tree-core.h (DEF_BUILTIN): Likewise.
2458 (DEF_BUILTIN_CHKP): Likewise.
2459 * tree-inline.c (declare_return_variable): Likewise.
2460 (remap_gimple_stmt): Likewise.
2461 (copy_bb): Likewise.
2462 (initialize_inlined_parameters): Likewise.
2463 (expand_call_inline): Likewise.
2464 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
2465 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
2466 (make_pass_ipa_chkp_produce_thunks): Likewise.
2467 (make_pass_chkp): Likewise.
2468 (make_pass_chkp_opt): Likewise.
2469 (make_pass_chkp_instrumentation_passes): Likewise.
2470 * tree-pretty-print.c (dump_generic_node): Likewise.
2471 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
2472 * tree-ssa-dce.c (propagate_necessity): Likewise.
2473 (eliminate_unnecessary_stmts): Likewise.
2474 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
2475 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2476 * tree-ssa-sccvn.h: Likewise.
2477 * tree-ssa-strlen.c (get_string_length): Likewise.
2478 (valid_builtin_call): Likewise.
2479 (adjust_last_stmt): Likewise.
2480 (handle_builtin_strchr): Likewise.
2481 (handle_builtin_strcpy): Likewise.
2482 (handle_builtin_stxncpy): Likewise.
2483 (handle_builtin_memcpy): Likewise.
2484 (handle_builtin_strcat): Likewise.
2485 (strlen_check_and_optimize_stmt): Likewise.
2486 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
2487 * tree-streamer-in.c: Likewise.
2488 * tree-streamer.c (record_common_node): Likewise.
2489 * tree.c (tree_code_size): Likewise.
2490 (wide_int_to_tree_1): Likewise.
2491 (type_contains_placeholder_1): Likewise.
2492 (build_common_tree_nodes): Likewise.
2493 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
2494 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
2495 (POINTER_BOUNDS_P): Likewise.
2496 (BOUNDED_TYPE_P): Likewise.
2497 (BOUNDED_P): Likewise.
2498 (CALL_WITH_BOUNDS_P): Likewise.
2499 (pointer_bounds_type_node): Likewise.
2500 * value-prof.c (gimple_ic): Likewise.
2501 * var-tracking.c (vt_add_function_parameters): Likewise.
2502 * varasm.c (make_decl_rtl): Likewise.
2503 (assemble_start_function): Likewise.
2504 (output_constant): Likewise.
2505 (maybe_assemble_visibility): Likewise.
2506 * varpool.c (ctor_for_folding): Likewise.
2507 * chkp-builtins.def: Remove.
2508 * ipa-chkp.c: Remove.
2509 * ipa-chkp.h: Remove.
2510 * rtl-chkp.c: Remove.
2511 * rtl-chkp.h: Remove.
2512 * tree-chkp-opt.c: Remove.
2513 * tree-chkp.c: Remove.
2514 * tree-chkp.h: Remove.
2516 2018-06-07 Carl Love <cel@us.ibm.com>
2518 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
2519 vextract_fp_from_shortl): Add BE support.
2521 2018-06-07 Paul Koning <ni1d@arrl.net>
2523 * compare-elim.c (try_merge_compare): Don't merge compare if
2524 address contains a side effect.
2525 (try_eliminate_compare): Likewise.
2527 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2529 * config.gcc: Support "tremont".
2530 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2531 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2533 * config/i386/i386.c (m_TREMONT): Define.
2534 (processor_target_table): Add "tremont".
2535 (PTA_TREMONT): Define.
2536 (ix86_lea_outperforms): Add TARGET_TREMONT.
2537 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
2538 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
2539 and M_INTEL_GOLDMONT_PLUS.
2540 (fold_builtin_cpu): Add "tremont".
2541 (ix86_add_stmt_cost): Add TARGET_TREMONT.
2542 (ix86_option_override_internal): Add "tremont".
2543 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
2544 (processor_type): Add PROCESSOR_TREMONT.
2545 * config/i386/x86-tune.def: Add m_TREMONT.
2546 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
2548 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2550 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2551 symbol defined for msp430i* devices to be lower case.
2553 gcc/testsuite/gcc.target/msp430/
2554 * msp430i-device-symbol.c: New test.
2555 * msp430f-device-symbol.c: New test.
2556 * msp430.h: New test header file.
2558 2018-06-07 Richard Biener <rguenther@suse.de>
2560 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2561 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2562 Properly wrap signed arithmetic if overflow wraps.
2564 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2566 PR tree-optimization/69615
2567 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2568 of a cast from a same precision integral SSA_NAME in a bb dominated
2569 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2570 cast to utype if rhs2 has already a compatible type.
2572 2018-06-07 Richard Biener <rguenther@suse.de>
2574 PR tree-optimization/85935
2575 * graphite-scop-detection.c (find_params_in_bb): Analyze
2576 condition operands with respect to the correct loop. Assert
2577 the analysis doesn't fail.
2579 2018-06-04 Carl Love <cel@us.ibm.com>
2581 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2582 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2583 as it is slightly cheaper.
2584 (first_match_or_eos_index_<mode>):
2585 Calculate index using natural element order.
2586 (first_match_index_<mode>):
2587 Calculate index using natural element order.
2588 (first_match_or_eos_index_<mode>):
2589 Calculate index using natural order.
2590 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2591 for BE and LE modes.
2592 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2593 P9V_BUILTIN_VCLZLSBB_V16QI.
2594 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2597 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2599 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2600 indentation and line wrap for many prototypes. Add missing
2601 @smallexample directives around block of prototypes for vec_xl and
2604 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2606 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2607 track if we pass or return IEEE 128-bit floating point.
2608 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2609 C++ mangling that is compatible with GCC 8.1.
2610 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2611 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2612 floating point types.
2613 (rs6000_function_arg_advance_1): Likewise.
2614 (rs6000_mangle_type): Optionally generate mangled names that match
2615 what GCC 8.1 generated for IEEE 128-bit floating point types.
2616 (rs6000_globalize_decl_name): If we have an external function that
2617 passes or returns IEEE 128-bit types, generate a weak reference
2618 from the mangled name used in GCC 8.1 to the current mangled
2620 (rs6000_init_builtins): Make __ibm128 use the long double type if
2621 long double is IBM extended double. Make __float128 use the long
2622 double type if long double is IEEE 128-bit.
2625 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2626 macro for __ibm128 built-in functions.
2627 (PACK_IF): Add __ibm128 pack/unpack functions.
2628 (UNPACK_IF): Likewise.
2629 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2630 enable long double built-in functions if long double is IEEE
2631 128-bit floating point.
2632 (rs6000_invalid_builtin): Update long double built-in function
2634 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2635 functions, adjust the built-in function to use the long double
2636 built-in function if __ibm128 and long double are the same type.
2637 * doc/extend.texi (PowerPC builtins): Update documention for
2638 __builtin_{,un}pack_longdouble. Add documentation for
2639 __builtin_{,un}pack_ibm128.
2641 2018-06-06 Jim Wilson <jimw@sifive.com>
2643 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2644 (struct machine_function): New field interrupt_mode.
2645 (riscv_handle_type_attribute): New function. Add forward declaration.
2646 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2647 (riscv_expand_epilogue): Check interrupt_mode field.
2648 (riscv_set_current_function): Check interrupt attribute args and
2649 set interrupt_mode field.
2650 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2651 (riscv_sret, riscv_uret): New.
2652 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2653 new arguments to interrupt attribute.
2655 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2658 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2659 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2661 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2663 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2664 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2665 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2666 several redundant entries.
2668 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2670 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2671 type from "rtx" to "rtx_insn *".
2672 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2673 for local "call_insn", removing cast.
2674 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2676 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2678 PR tree-optimization/86066
2679 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2680 for BIT_INSERT_EXPR stores.
2682 2018-06-06 Richard Biener <rguenther@suse.de>
2684 PR tree-optimization/86062
2685 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2686 component refs ontop
2687 of to be offsetted base.
2689 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2691 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2692 to be static and remove check on interrupt attribute name.
2694 gcc/testsuite/gcc.target/msp430/
2695 * function-attributes-4.c: New test.
2696 * static-interrupts.c: New test.
2698 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2700 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2701 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2703 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2706 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2708 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2709 Remove second argument, change how error is called.
2710 (aarch64_layout_arg): Remove second argument from
2711 aarch64_err_no_fpadvsimd call.
2712 (aarch64_init_cumulative_args): Ditto.
2713 (aarch64_gimplify_va_arg_expr): Ditto.
2714 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2716 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2718 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2719 (*simple_return_indirect_internal<mode>): Rename from
2720 simple_return_indirect_internal. Use W mode iterator.
2721 (rstorssp): New expander.
2722 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2723 (clrssbsy): New expander.
2724 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2726 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2728 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2730 (cmse_check_pointed_object): Likewise.
2732 2018-06-05 Martin Liska <mliska@suse.cz>
2734 PR gcov-profile/47618
2735 * doc/invoke.texi: Document how -fprofile-dir format
2738 2018-06-05 Richard Biener <rguenther@suse.de>
2740 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2741 removal pretend DOM info isn't available so we do not update
2742 it and only remove edges, not dominated blocks. Actually free
2743 DOM info in case we removed something. Remove unreachable blocks.
2744 (mfb_keep_latches): Work with either DOM info or marked backedges.
2745 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2746 first. Mark backedges if DOM info isn't available.
2747 (Re-)compute DOM info after cleanup_control_flow_pre.
2749 2018-06-05 Richard Biener <rguenther@suse.de>
2751 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2752 (locus_discrim_hasher::hash): Adjust.
2753 (locus_discrim_hasher::equal): Likewise.
2754 (next_discriminator_for_locus): Work on line directly.
2755 (same_line_p): Pass in expanded locus1 as well.
2756 (assign_discriminators): Avoid redundant location expansions.
2758 2018-06-05 Richard Biener <rguenther@suse.de>
2760 PR tree-optimization/86046
2761 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2762 if required after clearing TREE_ADDRESSABLE.
2764 2018-06-05 Richard Biener <rguenther@suse.de>
2766 PR tree-optimization/86047
2767 * tree-ssa-loop.c (for_each_index): Glob handling of all
2768 decls and constants and really handle all of them.
2770 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2773 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2774 qualifier_void_pointer and qualifier_const_void_pointer.
2775 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2776 (arm_init_builtins): Handle the above.
2777 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2778 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2779 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2782 2018-06-05 Martin Liska <mliska@suse.cz>
2784 * auto-profile.c (read_autofdo_file): Do not use
2785 gcov_ctr_summary struct.
2786 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2787 * coverage.c (struct counts_entry): Likewise.
2788 (read_counts_file): Read just single summary entry.
2789 (get_coverage_counts): Use gcov_summary struct.
2790 * coverage.h (get_coverage_counts): Likewise.
2791 * gcov-dump.c (dump_working_sets): Likewise.
2792 (tag_summary): Dump just single summary.
2793 * gcov-io.c (gcov_write_summary): Write just histogram
2795 (gcov_read_summary): Read just single summary.
2796 (compute_working_sets): Use gcov_summary struct.
2797 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2798 of GCOV_COUNTERS_SUMMABLE.
2799 (GCOV_COUNTERS_SUMMABLE): Remove.
2800 (GCOV_FIRST_VALUE_COUNTER): Replace with
2801 GCOV_COUNTER_V_INTERVAL.
2802 (struct gcov_ctr_summary): Remove.
2803 (struct gcov_summary): Directly use fields of former
2805 (compute_working_sets): Use gcov_summary struct.
2806 * gcov.c (read_count_file): Do not use ctrs fields.
2807 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2809 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2811 * profile.c: Likewise.
2812 * profile.h: Likewise.
2814 2018-06-05 Martin Liska <mliska@suse.cz>
2816 PR gcov-profile/84846
2817 * gcov.c (output_lines): Print working directory only
2818 in intermediate format.
2820 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2822 * config/s390/s390-builtin-types.def: Add void function type.
2823 * config/s390/s390-builtins.def: Use the function type for the
2826 2018-06-04 Jim Wilson <jimw@sifive.com>
2828 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2830 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2831 and maybe_eh_return. Change regno to unsigned int. Use new args to
2832 handle EH_RETURN_DATA_REGNO registers properly.
2833 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2834 (riscv_expand_epilogue): Update comment. Change argument name and
2835 type. Update code to use new name and type. Pass new args to
2836 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2838 * config/riscv/riscv.md (NORMAL_RETURN): New.
2839 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2840 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2841 (eh_return): Call gen_eh_return_internal and emit barrier.
2842 (eh_return_internal): Call riscv_expand_epilogue.
2844 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2846 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2847 bit_insertion field and declare can_be_merged_into method.
2848 (merged_store_group::can_be_merged_into): New method.
2849 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2850 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2851 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2853 2018-06-04 Richard Biener <rguenther@suse.de>
2855 PR tree-optimization/85955
2856 * builtins.c (fold_builtin_sincos): Convert pointers to
2857 destination to appropriate type before dereferencing.
2859 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2861 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2863 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2865 * expr.c (expand_expr_real_1): Force the operand into memory if
2866 its TYPE_MODE is BLKmode and if there is no integer mode for
2867 the number of bits being extracted.
2869 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2873 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2874 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2876 2018-06-04 Richard Biener <rguenther@suse.de>
2878 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2879 (cleanup_tree_cfg_noloop): ... single caller. Do
2880 start_recording_case_labels later.
2882 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2884 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2885 to _IMMINTRIN_H_INCLUDED.
2886 * config/i386/pconfigintrin.h: Ditto.
2887 * config/i386/waitpkgintrin.h: Ditto.
2888 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2889 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2890 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2891 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2892 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2893 waitpkgintrin.h and cldemoteintrin.h.
2895 2018-06-04 Richard Biener <rguenther@suse.de>
2897 PR tree-optimization/86038
2898 * tracer.c (find_best_successor): Check probability for
2899 being initialized, bail out if not.
2901 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2904 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2905 of bits to ignore when comparing architectures.
2907 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2909 PR tree-optimization/69615
2910 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2911 maximum or minimum of the type, try to merge it also as if
2912 range1 is + [-, x - 1] or + [x + 1, -].
2915 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2917 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2919 PR tree-optimization/86034
2920 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2921 the unsigned bitfield type in a bit insertion sequence if it does not
2922 have a larger precision than the bitfield size.
2923 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2925 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2927 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2929 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2931 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2932 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2933 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2934 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2936 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2938 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2939 Disable -fdelete-null-pointer-checks for ELF toolchain.
2941 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2942 Kito Cheng <kito.cheng@gmail.com>
2944 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2945 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2946 (nds32*-*-*): Add glibc and uclibc conditions.
2947 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2948 (TARGET_EXCEPT_UNWIND_INFO): Define.
2949 * config/nds32/elf.h: New file.
2950 * config/nds32/linux.h: New file.
2951 * config/nds32/nds32-elf.opt: New file.
2952 * config/nds32/nds32-linux.opt: New file.
2953 * config/nds32/nds32-fp-as-gp.c
2954 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2955 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2957 (nds32_asm_file_end): Ditto.
2958 (nds32_print_operand): Ditto.
2959 (nds32_insert_attributes): Ditto.
2960 (nds32_init_libfuncs): New function.
2961 (TARGET_HAVE_TLS): Define.
2962 (TARGET_INIT_LIBFUNCS): Define.
2963 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2965 (TARGET_ELF): Apply different mcmodel setting.
2966 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2967 been migrated into elf.h and linux.h files.
2968 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2969 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2970 (mcmodel): The content has been migrated into nds32-elf.opt and
2971 nds32-linux.opt files.
2972 * config/nds32/t-elf: New file.
2973 * config/nds32/t-linux: New file.
2975 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2976 Shiva Chen <shiva0217@gmail.com>
2978 * config/nds32/constants.md (unspec_volatile_element): Add
2979 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2980 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2982 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2983 (make_pass_nds32_fp_as_gp): Declare.
2984 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2986 (nds32_asm_function_end_prologue): Remove unused asm output.
2987 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2988 (nds32_asm_file_start): Output necessary fp_as_gp information.
2989 (nds32_option_override): Adjust register usage.
2990 (nds32_expand_prologue): Consider fp_as_gp situation.
2991 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2992 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2995 (simple_return): Ditto.
2996 (omit_fp_begin): Output special directive for fp_as_gp.
2997 (omit_fp_end): Output special directive for fp_as_gp.
2998 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2999 mforbid-fp-as-gp): New options.
3001 2018-06-01 Mark Wielaard <mark@klomp.org>
3003 * dwarf2out.c (dwarf2out_finish): Remove generation of
3004 DW_AT_loclists_base.
3006 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
3008 * gimple-ssa-store-merging.c: Include gimple-fold.h.
3009 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
3010 (struct merged_store_group): Add bit_insertion field.
3011 (dump_char_array): Use standard hexadecimal format.
3012 (merged_store_group::merged_store_group): Set bit_insertion to false.
3013 (merged_store_group::apply_stores): Use optimal buffer size. Deal
3014 with BIT_INSERT_EXPR stores. Move up code updating the mask and
3015 also print the mask in the dump file.
3016 (pass_store_merging::gate): Minor tweak.
3017 (imm_store_chain_info::coalesce_immediate): Fix wrong association
3018 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
3019 stores with INTEGER_CST stores.
3020 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
3021 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
3022 and use it throughout. Generate bit insertion sequences if need be.
3023 (pass_store_merging::process_store): Remove redundant condition.
3024 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
3026 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3028 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3029 the 128-bit floating point types. Fix function comment.
3031 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3033 * config/aarch64/aarch64-simd.md
3034 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
3036 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
3039 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3041 PR tree-optimization/85989
3042 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3044 (backprop::intersect_uses): Check it when deciding whether this
3045 is a backedge reference.
3046 (backprop::process_block): Add each phi to m_visited_phis
3047 after visiting it, then clear it at the end.
3049 2018-06-01 Richard Biener <rguenther@suse.de>
3051 * tree-vectorizer.h (vect_dr_stmt): New function.
3052 (vect_get_load_cost): Adjust.
3053 (vect_get_store_cost): Likewise.
3054 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
3055 Use vect_dr_stmt instead of DR_SMTT.
3056 (vect_record_base_alignments): Likewise.
3057 (vect_calculate_target_alignment): Likewise.
3058 (vect_compute_data_ref_alignment): Likewise and make static.
3059 (vect_update_misalignment_for_peel): Likewise.
3060 (vect_verify_datarefs_alignment): Likewise.
3061 (vector_alignment_reachable_p): Likewise.
3062 (vect_get_data_access_cost): Likewise. Pass down
3063 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
3064 (vect_get_peeling_costs_all_drs): Likewise.
3065 (vect_peeling_hash_get_lowest_cost): Likewise.
3066 (vect_enhance_data_refs_alignment): Likewise.
3067 (vect_find_same_alignment_drs): Likewise.
3068 (vect_analyze_data_refs_alignment): Likewise.
3069 (vect_analyze_group_access_1): Likewise.
3070 (vect_analyze_group_access): Likewise.
3071 (vect_analyze_data_ref_access): Likewise.
3072 (vect_analyze_data_ref_accesses): Likewise.
3073 (vect_vfa_segment_size): Likewise.
3074 (vect_small_gap_p): Likewise.
3075 (vectorizable_with_step_bound_p): Likewise.
3076 (vect_prune_runtime_alias_test_list): Likewise.
3077 (vect_analyze_data_refs): Likewise.
3078 (vect_supportable_dr_alignment): Likewise.
3079 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3080 (vect_gen_prolog_loop_niters): Likewise.
3081 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3082 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
3084 (vect_recog_mask_conversion_pattern): Likewise.
3085 (vect_try_gather_scatter_pattern): Likewise.
3086 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
3087 to vect_get_store_cost.
3088 (vect_get_store_cost): Get stmt_info instead of DR.
3089 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
3090 (vect_get_load_cost): Get stmt_info instead of DR.
3092 2018-06-01 Richard Biener <rguenther@suse.de>
3095 * gimple-fold.c (var_decl_component_p): Also allow offsetted
3096 vars wrapped in MEM_REFs.
3098 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3100 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3101 Fix subreg tests so that we only return a choice between
3102 GENERAL_REGS and FP_REGS if the original classes included both.
3104 2018-06-01 Richard Biener <rguenther@suse.de>
3107 * tree-ssa-structalias.c (get_function_part_constraint):
3108 Handle NULL fi->decl.
3109 (find_func_aliases_for_call): Properly handle indirect
3110 fi from direct call.
3111 (find_func_clobbers): Likewise.
3112 (ipa_pta_execute): Likewise.
3113 (create_variable_info_for): For functions that are ifunc_resolver
3114 resolve to a varinfo that contains the result of the resolver call.
3115 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
3118 2018-05-31 Michael Collison <michael.collison@arm.com>
3120 * config/aarch64/aarch64.md:
3121 (*fix_to_zero_extenddfdi2): New pattern.
3122 * gcc.target/aarch64/fix_extend1.c: New testcase.
3124 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
3128 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
3129 and BUILT_IN_STRNCMP_EQ.
3130 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
3131 BUILT_IN_STRNCMP_EQ.
3132 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
3133 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3134 (gimple_fold_builtin): Likewise.
3135 * tree-ssa-strlen.c (compute_string_length): New function.
3136 (determine_min_obsize): New function.
3137 (handle_builtin_string_cmp): New function to handle calls to
3138 string compare functions.
3139 (strlen_optimize_stmt): Add handling to builtin string compare
3141 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3142 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3143 * tree.c (build_common_builtin_nodes): Add new defines of
3144 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
3146 2018-05-31 Jakub Jelinek <jakub@redhat.com>
3149 * bb-reorder.c (pass_partition_blocks::gate): Return false for
3150 functions with naked attribute.
3152 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
3154 * config/i386/sse.md (avx_vec_concat<mode>):
3155 Substitute concat_tg_mode mode attribute with xtg_mode.
3156 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
3157 (concat_tg_mode): Remove mode attribute.
3159 2018-05-31 Martin Sebor <msebor@redhat.com>
3162 * calls.c (alloc_max_size): Correct a logic error/typo.
3163 Treat excessive arguments as infinite. Warn for invalid arguments.
3164 * doc/invoke.texi (-Walloc-size-larger-than): Update.
3166 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
3169 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
3170 and movx for Haswell.
3172 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
3173 Cesar Philippidis <cesar@codesourcery.com>
3176 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
3177 when emitting error on private/firstprivate reductions.
3178 * omp-low.c (lower_omp_target): Avoid reference-type processing
3179 on pointers for firstprivate clause.
3181 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3183 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
3186 * config/aarch64/aarch64-simd.md
3187 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
3188 (aarch64_ld1_x3_<mode>): Likewise
3189 (aarch64_st1x2<VALLDIF:mode>): Likewise
3190 (aarch64_st1_x2_<mode>): Likewise
3191 (aarch64_st1x3<VALLDIF:mode>): Likewise
3192 (aarch64_st1_x3_<mode>): Likewise
3193 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
3194 (vld1_s8_x3): Likewise.
3195 (vld1_u16_x3): Likewise.
3196 (vld1_s16_x3): Likewise.
3197 (vld1_u32_x3): Likewise.
3198 (vld1_s32_x3): Likewise.
3199 (vld1_u64_x3): Likewise.
3200 (vld1_s64_x3): Likewise.
3201 (vld1_f16_x3): Likewise.
3202 (vld1_f32_x3): Likewise.
3203 (vld1_f64_x3): Likewise.
3204 (vld1_p8_x3): Likewise.
3205 (vld1_p16_x3): Likewise.
3206 (vld1_p64_x3): Likewise.
3207 (vld1q_u8_x3): Likewise.
3208 (vld1q_s8_x3): Likewise.
3209 (vld1q_u16_x3): Likewise.
3210 (vld1q_s16_x3): Likewise.
3211 (vld1q_u32_x3): Likewise.
3212 (vld1q_s32_x3): Likewise.
3213 (vld1q_u64_x3): Likewise.
3214 (vld1q_s64_x3): Likewise.
3215 (vld1q_f16_x3): Likewise.
3216 (vld1q_f32_x3): Likewise.
3217 (vld1q_f64_x3): Likewise.
3218 (vld1q_p8_x3): Likewise.
3219 (vld1q_p16_x3): Likewise.
3220 (vld1q_p64_x3): Likewise.
3221 (vst1_s64_x2): Likewise.
3222 (vst1_u64_x2): Likewise.
3223 (vst1_f64_x2): Likewise.
3224 (vst1_s8_x2): Likewise.
3225 (vst1_p8_x2): Likewise.
3226 (vst1_s16_x2): Likewise.
3227 (vst1_p16_x2): Likewise.
3228 (vst1_s32_x2): Likewise.
3229 (vst1_u8_x2): Likewise.
3230 (vst1_u16_x2): Likewise.
3231 (vst1_u32_x2): Likewise.
3232 (vst1_f16_x2): Likewise.
3233 (vst1_f32_x2): Likewise.
3234 (vst1_p64_x2): Likewise.
3235 (vst1q_s8_x2): Likewise.
3236 (vst1q_p8_x2): Likewise.
3237 (vst1q_s16_x2): Likewise.
3238 (vst1q_p16_x2): Likewise.
3239 (vst1q_s32_x2): Likewise.
3240 (vst1q_s64_x2): Likewise.
3241 (vst1q_u8_x2): Likewise.
3242 (vst1q_u16_x2): Likewise.
3243 (vst1q_u32_x2): Likewise.
3244 (vst1q_u64_x2): Likewise.
3245 (vst1q_f16_x2): Likewise.
3246 (vst1q_f32_x2): Likewise.
3247 (vst1q_f64_x2): Likewise.
3248 (vst1q_p64_x2): Likewise.
3249 (vst1_s64_x3): Likewise.
3250 (vst1_u64_x3): Likewise.
3251 (vst1_f64_x3): Likewise.
3252 (vst1_s8_x3): Likewise.
3253 (vst1_p8_x3): Likewise.
3254 (vst1_s16_x3): Likewise.
3255 (vst1_p16_x3): Likewise.
3256 (vst1_s32_x3): Likewise.
3257 (vst1_u8_x3): Likewise.
3258 (vst1_u16_x3): Likewise.
3259 (vst1_u32_x3): Likewise.
3260 (vst1_f16_x3): Likewise.
3261 (vst1_f32_x3): Likewise.
3262 (vst1_p64_x3): Likewise.
3263 (vst1q_s8_x3): Likewise.
3264 (vst1q_p8_x3): Likewise.
3265 (vst1q_s16_x3): Likewise.
3266 (vst1q_p16_x3): Likewise.
3267 (vst1q_s32_x3): Likewise.
3268 (vst1q_s64_x3): Likewise.
3269 (vst1q_u8_x3): Likewise.
3270 (vst1q_u16_x3): Likewise.
3271 (vst1q_u32_x3): Likewise.
3272 (vst1q_u64_x3): Likewise.
3273 (vst1q_f16_x3): Likewise.
3274 (vst1q_f32_x3): Likewise.
3275 (vst1q_f64_x3): Likewise.
3276 (vst1q_p64_x3): Likewise.
3278 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3280 * config/msp430/msp430.c (msp430_output_labelref): Prepend
3281 user_label_prefix to name.
3283 * tree-core.h: Update comment about the format of NAME string
3284 passed to handler in attribute_spec.
3286 * config/msp430/msp430.md: Remove erroneous subreg expression from
3287 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
3288 zero_extend{q,h}isi2.
3290 2018-05-30 Borislav Petkov <bp@suse.de>
3292 * doc/extend.texi: Document some architecture specific
3293 constraints and sort entries.
3295 2018-05-30 Martin Sebor <msebor@redhat.com>
3298 * builtins.c (expand_builtin_stpcpy_1): New function.
3299 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
3300 only if the former succeeds.
3302 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3304 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
3307 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3309 * doc/invoke.texi (-flinker-output): Document
3311 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3313 * passes.c (ipa_write_summaries): Only modify statements if body
3315 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
3316 incrementally linking.
3317 (ipa_passes): Likewise.
3318 * lto-cgraph.c (lto_output_node): When incrementally linking do not
3319 pass down resolution info.
3320 * common.opt (flag_incremental_link): Update info.
3321 * gcc.c (plugin specs): Turn flinker-output=* to
3322 -plugin-opt=-linker-output-known
3323 * toplev.c (compile_file): Also cut compilation when doing incremental
3325 * flag-types. (enum lto_partition_model): Add
3326 LTO_LINKER_OUTPUT_NOLTOREL.
3327 (invoke.texi): Add -flinker-output docs.
3328 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
3329 link same way as WPA; do not stream in dead initializers.
3331 * dwarf2out.c (dwarf2out_die_ref_for_decl,
3332 darf2out_register_external_decl): Support incremental link.
3334 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3336 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
3338 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3340 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
3341 it down to simple_object_copy_lto_debug_sections.
3342 (run_gcc): Determine incremental LTO link time and configure
3343 lto1 into non-wpa mode, disable renaming of debug sections.
3345 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
3347 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3348 descriptions of various incorrectly documented functions.
3350 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
3353 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3354 address check not strict.
3356 2018-05-30 Richard Biener <rguenther@suse.de>
3358 PR tree-optimization/85964
3359 * tracer.c (better_p): Drop initialized count check, we only
3360 call the function with initialized counts now.
3361 (find_best_successor): Do find a best edge if one
3362 has uninitialized count.
3363 (find_best_predecessor): Likewise. Do BB frequency check only
3364 if count is initialized.
3366 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
3368 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
3369 (aarch64_ldrstr_offset_compare): New.
3370 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
3371 load/store orderings.
3372 (aarch64_gen_adjusted_ldpstp): Likewise.
3374 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
3376 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3377 Check for subset of GENERAL_REGS and FP_REGS.
3378 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
3381 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
3383 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
3384 and wi::to_poly_offset. Add the current offset and then check
3385 whether the sum fits, rather than using an unchecked addition of
3386 a checked term. Check for a shwi rather than a uhwi.
3387 * expr.c (get_bit_range): Use tree_to_poly_uint64.
3388 (store_constructor): Use poly_int_tree_p.
3389 (expand_expr_real_1): Likewise.
3390 * function.c (assign_temp): Likewise.
3391 * fold-const.c (const_binop): Use poly_int_tree_p and
3393 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
3395 * ipa-icf-gimple.c (func_checker::compare_operand): Use
3396 to_poly_offset for MEM offsets.
3397 * ipa-icf.c (sem_variable::equals): Likewise.
3398 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
3399 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
3400 wi::to_poly_offset for BIT_FIELD_REF offsets.
3401 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
3403 * var-tracking.c (emit_note_insn_var_location): Use
3404 tree_to_poly_uint64.
3406 2018-05-29 Jim Wilson <jimw@sifive.com>
3408 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
3410 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
3413 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
3414 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
3416 (sse4_1_round<mode>2): Use nonimmediate_operand
3417 for operand 1 predicate.
3419 2018-05-29 Martin Sebor <msebor@redhat.com>
3420 Richard Biener <rguenther@suse.de>
3423 * calls.c (get_size_range): Call determine_value_range instead
3424 of get_value_range..
3425 * tree-vrp.h (determine_value_range): Declared new function.
3426 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
3428 2018-05-29 Richard Biener <rguenther@suse.de>
3430 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
3431 sure to use non-pattern stmts for get_earlier_stmt arguments.
3432 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
3433 called on pattern stmts.
3434 (get_later_stmt): Likewise.
3436 2018-05-29 Martin Liska <mliska@suse.cz>
3438 PR gcov-profile/85759
3439 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
3442 2018-05-29 Jakub Jelinek <jakub@redhat.com>
3444 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
3446 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
3450 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3451 VEC_PACK_FLOAT_EXPR): New tree codes.
3452 * tree-pretty-print.c (op_code_prio): Handle
3453 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3454 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3455 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3456 * tree-inline.c (estimate_operator_cost): Likewise.
3457 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
3458 * fold-const.c (const_binop): Likewise.
3459 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
3460 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3461 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3462 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
3463 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3464 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3465 * expr.c (expand_expr_real_2): Likewise.
3466 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
3467 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
3468 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
3470 * optabs.c (expand_widen_pattern_expr): For
3471 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
3472 sign from result type rather than operand's type.
3473 (expand_binop_directly): For vec_packu_float_optab and
3474 vec_packs_float_optab allow result type to be different from operand's
3476 * optabs-tree.c (optab_for_tree_code): Handle
3477 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3478 VEC_PACK_FLOAT_EXPR. Formatting fixes.
3479 * tree-vect-generic.c (expand_vector_operations_1): Handle
3480 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3481 VEC_PACK_FLOAT_EXPR.
3482 * tree-vect-stmts.c (supportable_widening_operation): Handle
3484 (supportable_narrowing_operation): Handle FLOAT_EXPR.
3485 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
3486 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
3487 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
3488 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
3490 (vec_pack<floatprefix>_float_<mode>): New expander.
3491 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
3493 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
3494 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
3495 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
3496 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
3497 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
3499 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
3500 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
3501 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3502 VEC_PACK_FLOAT_EXPR): Document.
3504 2018-05-29 Richard Biener <rguenther@suse.de>
3506 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
3508 (stmt_vec_info_vec): Make pointer.
3509 (init_stmt_vec_info_vec): Remove.
3510 (free_stmt_vec_info_vec): Likewise.
3511 (set_stmt_vec_info_vec): New function.
3512 (free_stmt_vec_infos): Likewise.
3513 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
3514 (set_vinfo_for_stmt): Likewise.
3515 (get_earlier_stmt): Likewise.
3516 (get_later_stmt): Likewise.
3517 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
3518 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
3519 (vec_info::~vec_info): Free stmt_vec_infos.
3520 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3521 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3522 (pass_slp_vectorize::execute): Likewise.
3523 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
3524 (free_stmt_vec_info_vec): Likewise.
3525 (set_stmt_vec_info_vec): New function.
3526 (free_stmt_vec_infos): Likewise.
3527 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3528 the global stmt_vec_info_vec.
3529 * tree-parloops.c (gather_scalar_reductions): Use
3530 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3533 2018-05-29 Richard Biener <rguenther@suse.de>
3535 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
3537 2018-05-29 Martin Liska <mliska@suse.cz>
3538 David Malcolm <dmalcolm@redhat.com>
3540 * vec.c (test_reverse): New.
3541 (vec_c_tests): Add new test.
3542 * vec.h (vl_ptr>::reverse): New function.
3544 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
3546 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
3548 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3551 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3553 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3555 2018-05-28 Richard Biener <rguenther@suse.de>
3557 PR tree-optimization/85933
3558 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3559 look at stmts marked as vectorizable.
3561 2018-05-28 Richard Biener <rguenther@suse.de>
3563 PR tree-optimization/85934
3564 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3565 vector boolean check before scalar optimization.
3567 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3569 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3572 2018-05-28 Mark Wielaard <mark@klomp.org>
3574 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3575 if it is an expression containing a minus sign.
3577 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3579 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3581 2018-05-27 Paul Koning <ni1d@arrl.net>
3583 * config/pdp11/pdp11.md (truncsihi2): Remove.
3585 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3586 Chung-Ju Wu <jasonwucj@gmail.com>
3588 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3590 (unaligned_store_dw): Ditto.
3591 * config/nds32/nds32-memory-manipulation.c
3592 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3593 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3594 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3595 (emit_setmem_word_loop): Rename to ...
3596 (emit_setmem_doubleword_loop): ... this.
3597 (nds32_gen_dup_4_byte_to_word_value): New function.
3598 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3599 (nds32_expand_setmem_loop): Refine implementation.
3600 (nds32_expand_setmem_loop_v3m): Ditto.
3601 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3604 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3606 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3608 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3610 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3611 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3612 machine->attr_no_prologue_p.
3613 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3614 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3615 (nds32_expand_epilogue): Consider attr_naked_p.
3616 (nds32_expand_epilogue_v3pop): Likewise.
3617 (nds32_can_use_return_insn): Likewise.
3618 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3619 attr_no_prologue_p fields.
3620 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3622 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3625 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3627 * config/i386/sse.md
3628 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3630 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3632 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3634 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3636 (*<floatsuffix>floatv2div2sf2): Rename to ...
3637 (*float<floatunssuffix>v2div2sf2): ... this.
3638 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3639 (float<floatunssuffix>v2div2sf2_mask): ... this.
3640 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3641 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3642 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3644 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3646 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3648 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3650 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3652 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3654 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3655 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3656 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3657 gen_ufix_truncv8dfv8si2.
3658 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3659 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3660 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3661 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3662 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3663 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3664 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3665 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3667 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3671 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3673 2018-05-25 Jim Wilson <jimw@sifive.com>
3675 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3676 * config/riscv/riscv.c (struct machine_function): Add
3677 interrupt_handler_p and attribute_checked_p fields.
3678 (riscv_attribute_table): Add interrupt.
3679 (riscv_interrupt_type_p): New.
3680 (riscv_save_reg_p): Save extra regs for interrupt handler.
3681 (riscv_use_save_libcall): Return false for interrupt handler.
3682 (riscv_first_stack_step): Add forward declaration.
3683 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3684 for interrupt handler with large frame. Use it for saved reg list.
3685 (riscv_expand_prologue): Move flag_stack_usage_info support to
3686 eliminate duplication.
3687 (riscv_expand_epilogue): Generate mret for interrupt handler.
3688 (riscv_epilogue_uses): New.
3689 (riscv_can_use_return_insn): Return false for interrupt handler.
3690 (riscv_function_ok_for_sibcall): Likewise.
3691 (riscv_set_current_function): Add interrupt handler support.
3692 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3693 * config/riscv/riscv.md (UNSPECV_MRET): New.
3695 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3697 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3699 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3701 PR tree-optimization/85712
3702 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3703 this candidate has already been replaced in-situ by a copy.
3705 2018-05-25 Jason Merrill <jason@redhat.com>
3707 PR c++/80485 - inline function non-zero address.
3708 * symtab.c (nonzero_address): Check DECL_COMDAT.
3710 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3713 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3714 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3715 extension of SImode operation. Use const123_operand predicate.
3716 (*saddsi_1): Remove.
3717 (*saddl_se_1): Ditto.
3719 (*ssubl_se_1): Ditto.
3720 * config/alpha/predicates.md (const123_operand): New predicate.
3721 * config/alpha/constraints.md (P): Use IN_RANGE.
3723 2018-05-25 Richard Biener <rguenther@suse.de>
3725 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3727 (ref_maybe_used_by_stmt_p): Likewise.
3728 (stmt_may_clobber_ref_p): Likewise.
3729 (stmt_may_clobber_ref_p_1): Likewise.
3730 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3732 (ref_maybe_used_by_stmt_p): Likewise.
3733 (stmt_may_clobber_ref_p): Likewise.
3734 (stmt_may_clobber_ref_p_1): Likewise.
3735 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3736 the alias oracle to disambiguate DRs with stmts DR analysis
3738 (vect_analyze_data_refs): Do not give up on not analyzable
3739 DRs for BB vectorization. Remove code truncating the dataref
3742 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3745 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3746 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3747 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3749 2018-05-25 Richard Biener <rguenther@suse.de>
3751 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3752 function, combining stmt data ref gathering and fatal analysis
3754 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3755 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3756 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3757 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3758 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3760 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3762 PR tree-optimization/85720
3763 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3764 SCC if all partitions are builtins.
3765 (version_loop_by_alias_check): New parameter. Generate cancelable
3766 runtime alias check if all partitions are builtins.
3767 (distribute_loop): Update call to above function.
3769 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3771 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3772 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3773 (parm_default_def_partition_arg): Ditto.
3774 (set_parm_default_def_partition): Ditto.
3775 (get_parm_default_def_partitions): Ditto and make it static.
3776 (get_undefined_value_partitions): Ditto and make it static.
3777 (remove_ssa_form): Refactor call to init_var_map here.
3778 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3779 computation for loop region.
3780 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3781 (register_default_def): Delete.
3782 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3783 (parm_default_def_partition_arg): Ditto.
3784 (set_parm_default_def_partition): Ditto.
3785 (get_parm_default_def_partitions): Ditto and make it static.
3786 (get_undefined_value_partitions): Ditto and make it static.
3787 (coalesce_with_default, coalesce_with_default): Update comment.
3788 (create_coalesce_list_for_region): New func factored out from
3789 create_outofssa_var_map.
3790 (populate_coalesce_list_for_outofssa): New func factored out from
3791 create_outofssa_var_map and coalesce_ssa_name.
3792 (create_outofssa_var_map): Delete.
3793 (coalesce_ssa_name): Refactor to support live range computation.
3794 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3795 (get_parm_default_def_partitions): Delete.
3796 (get_undefined_value_partitions): Ditto.
3797 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3798 computation for loop region.
3799 (new_tree_live_info, loe_visit_block): Ditto.
3800 (live_worklist, set_var_live_on_entry): Ditto.
3801 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3802 * tree-ssa-live.h (struct _var_map): New fields.
3803 (init_var_map): Change decl.
3804 (region_contains_p): New.
3806 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3808 * tree-ssa-live.h (live_merge_and_clear): Delete.
3810 2018-05-25 Richard Biener <rguenther@suse.de>
3813 * tree-dump.c (dequeue_and_dump): Remove access to removed
3814 operand 2 of a SWITCH_EXPR.
3816 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3818 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3819 multiplication and division.
3820 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3821 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3822 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3823 (cond_udiv_optab, cond_umod_optab): New optabs.
3824 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3825 (IFN_COND_RDIV): New internal functions.
3826 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3827 TRUNC_MOD_EXPR and RDIV_EXPR.
3828 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3829 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3831 (SVE_INT_BINARY): Include mult.
3832 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3833 (optab, sve_int_op): Handle mult.
3834 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3836 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3837 for SVE_INT_BINARY_SD.
3839 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3841 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3842 (optab, sve_int_op): Handle div and udiv.
3843 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3844 for SVE_INT_BINARY_SD.
3845 (*<optab><mode>3): New insn for the same.
3847 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3849 * tree-vect-patterns.c: Include predict.h.
3850 (vect_recog_divmod_pattern): Restrict check for division support
3851 to when optimizing for size.
3853 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3855 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3856 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3857 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3858 (gimple_match_op::set_op): Likewise.
3859 (gimple_resimplify4): Declare.
3860 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3861 (expr::gen_transform): Likewise.
3862 (decision_tree::gen): Generate a simplification routine for 4 operands.
3863 * gimple-match-head.c (gimple_simplify): Add an overload for
3864 4 operands. In the top-level function, handle up to 4 call
3865 arguments and call gimple_resimplify4.
3866 (gimple_resimplify4): New function.
3867 (build_call_internal): Pass a fourth operand.
3868 (maybe_push_to_seq): Likewise.
3869 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3870 Fold VEC_COND_EXPRs of an operation and a default value into
3871 an IFN_COND_* function if possible.
3872 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3874 (SVE_COND_FP_BINARY): Include them.
3875 (optab, sve_fp_op): Handle them.
3876 (SVE_INT_BINARY_REV): New code iterator.
3877 (SVE_COND_FP_BINARY_REV): New int iterator.
3878 (commutative): New int attribute.
3879 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3881 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3883 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3884 (*cond_<optab><mode>): New patterns for reversed operands.
3886 2018-05-25 Richard Biener <rguenther@suse.de>
3888 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3889 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3890 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3891 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3892 * tree-vect-data-refs.c (everywhere): Adjust users.
3893 * tree-vect-loop.c (everywhere): Likewise.
3894 * tree-vect-slp.c (everywhere): Likewise.
3895 * tree-vect-stmts.c (everywhere): Likewise.
3896 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3898 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3900 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3902 (gcc_cv_as_section_exclude): ... this.
3903 Try Solaris as #exclude syntax.
3904 * configure: Regenerate.
3905 * config.in: Regenerate.
3906 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3908 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3909 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3911 * varasm.c (default_elf_asm_named_section): Don't check if
3912 HAVE_GAS_SECTION_EXCLUDE is defined.
3914 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3916 * doc/md.texi: Update the documentation of the cond_* optabs
3917 to mention the new final operand. Fix GET_MODE_NUNITS call.
3918 Describe the scalar case too.
3919 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3920 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3922 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3923 (get_conditional_internal_fn): Update comment.
3924 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3925 accumulator value as a final argument to conditional functions.
3926 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3927 a define_expand and add an "else" operand. Assert for now that
3928 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3929 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3930 (*cond_<optab><mode>): New patterns.
3931 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3932 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3933 (UNSPEC_COND_EOR): Delete.
3934 (optab): Remove associated mappings.
3935 (SVE_INT_BINARY): New code iterator.
3936 (sve_int_op): Remove int attribute and add "minus" to the code
3938 (SVE_COND_INT_OP): Delete.
3939 (SVE_COND_FP_OP): Rename to...
3940 (SVE_COND_FP_BINARY): ...this.
3942 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3944 * optabs.c (can_reuse_operands_p): New function.
3945 (maybe_legitimize_operands): Try to reuse the results for
3948 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3950 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3951 Add {q} suffix to insn mnemonic.
3953 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3955 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3956 (msp430_warn_func_return): New.
3958 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3960 * fold-const.c (tree_nonzero_bits): New function.
3961 * fold-const.h (tree_nonzero_bits): Likewise.
3962 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3963 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3965 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3969 * varasm.c (assemble_alias): Check ifunc_resolver only on
3972 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3975 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3976 when memory input operand is handled.
3978 2018-05-24 Luis Machado <luis.machado@linaro.org>
3980 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3982 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3984 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3986 * match.pd: Delay FMA folds until after vectorization.
3988 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3991 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3992 address check not strict.
3994 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3996 * gimple-match.h (gimple_match_op): New class.
3997 (mprts_hook): Replace parameters with a gimple_match_op *.
3998 (maybe_build_generic_op): Likewise.
3999 (gimple_simplified_result_is_gimple_val): Replace parameters with
4000 a const gimple_match_op *.
4001 (gimple_simplify): Replace code_helper * and tree * parameters with
4002 a gimple_match_op * parameter.
4003 (gimple_resimplify1): Replace code_helper *, tree and tree *
4004 parameters with a gimple_match_op * parameter.
4005 (gimple_resimplify2): Likewise.
4006 (gimple_resimplify3): Likewise.
4007 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
4008 parameters with a gimple_match_op * parameter.
4009 * gimple-match-head.c (gimple_simplify): Change prototypes of
4010 auto-generated functions to take a gimple_match_op * instead of
4011 separate code_helper * and tree * parameters. Make the same
4012 change in the top-level overload and update calls to the
4013 gimple_resimplify routines. Update calls to the auto-generated
4014 functions and to maybe_push_res_to_seq in the publicly-facing
4015 operation-specific gimple_simplify overloads.
4016 (gimple_match_op::MAX_NUM_OPS): Define.
4017 (gimple_resimplify1): Replace rcode and ops with a single res_op
4018 parameter. Update call to gimple_simplify.
4019 (gimple_resimplify2): Likewise.
4020 (gimple_resimplify3): Likewise.
4021 (mprts_hook): Replace parameters with a gimple_match_op *.
4022 (maybe_build_generic_op): Likewise.
4023 (build_call_internal): Replace type, nargs and ops with
4024 a gimple_match_op *.
4025 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
4026 with a single gimple_match_op *. Update calls to mprts_hook,
4027 build_call_internal and gimple_simplified_result_is_gimple_val.
4028 Factor out code that is common to the tree_code and combined_fn cases.
4029 * genmatch.c (expr::gen_transform): Replace tem_code and
4030 tem_ops with a gimple_match_op called tem_op. Update calls
4031 to the gimple_resimplify functions and maybe_push_res_to_seq.
4032 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
4033 res_ops. Update call to the gimple_resimplify functions.
4034 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
4035 (decision_tree::gen): Make the functions take a gimple_match_op *
4036 called res_op instead of separate res_code and res_ops parameters.
4037 Update call accordingly.
4038 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
4039 and ops with a single res_op parameter. Update calls to
4040 maybe_build_generic_op and maybe_push_res_to_seq.
4041 (fold_stmt_1): Update calls to gimple_simplify and
4042 replace_stmt_with_simplification.
4043 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
4044 and gimple_simplified_result_is_gimple_val.
4045 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
4047 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
4048 with a gimple_match_op *.
4049 (vn_nary_build_or_lookup): Likewise. Update call to
4050 vn_nary_build_or_lookup_1.
4051 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
4052 gimple_match_op *. Update calls to the gimple_resimplify routines
4053 and to gimple_simplified_result_is_gimple_val.
4054 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
4055 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
4056 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
4057 (visit_nary_op): Likewise.
4058 (visit_reference_op_load): Likewise.
4060 2018-05-23 Luis Machado <luis.machado@linaro.org>
4062 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
4063 modifier for printing the step amount.
4065 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
4068 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
4071 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
4073 * doc/sourcebuild.texi (Endianness): New subsubsection.
4075 2018-05-23 Luis Machado <luis.machado@linaro.org>
4077 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4078 <prefetch_dynamic_strides>: New const bool field.
4079 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4080 prefetch_dynamic_strides.
4081 (exynosm1_prefetch_tune): Likewise.
4082 (thunderxt88_prefetch_tune): Likewise.
4083 (thunderx_prefetch_tune): Likewise.
4084 (thunderx2t99_prefetch_tune): Likewise.
4085 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
4087 (aarch64_override_options_internal): Update to set
4088 PARAM_PREFETCH_DYNAMIC_STRIDES.
4089 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4090 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4091 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4092 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4093 prefetch-dynamic-strides setting.
4095 2018-05-23 Luis Machado <luis.machado@linaro.org>
4097 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4098 <minimum_stride>: New const int field.
4099 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4100 minimum_stride field defaulting to -1.
4101 (exynosm1_prefetch_tune): Likewise.
4102 (thunderxt88_prefetch_tune): Likewise.
4103 (thunderx_prefetch_tune): Likewise.
4104 (thunderx2t99_prefetch_tune): Likewise.
4105 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
4106 <default_opt_level>: Set to 3.
4107 (aarch64_override_options_internal): Update to set
4108 PARAM_PREFETCH_MINIMUM_STRIDE.
4109 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4110 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4111 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4112 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4113 stride is constant and is below the minimum stride threshold.
4115 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4117 * config/arm/arm-cpus.in (mode26): Delete.
4118 (armv4): Delete mode26 reference.
4119 * config/arm/arm.c (arm_configure_build_target): Delete use of
4122 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4124 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
4126 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
4127 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
4129 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
4130 (floatunsdidf2): Ditto.
4132 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4134 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
4135 (fixuns_trunc<mode>si2_avx512f): Ditto.
4136 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
4137 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
4138 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
4140 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4142 PR rtl-optimization/79985
4143 * df-scan.c (df_insn_refs_collect): Remove special case for
4144 global registers and asm statements.
4146 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4148 * extend.texi (Global Register Variables): Rewrite the bullet list.
4149 Note that the register is available for allocation. Note that access
4150 via inline asm must use constraints. Add note about async-signal
4151 handlers. Remove paragraph about automagic register selection.
4153 2018-05-23 Richard Biener <rguenther@suse.de>
4155 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
4156 of fixed offset from memset VN.
4158 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
4160 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
4162 (alloc_cand_and_find_basis): Initialize first_interp field.
4163 (slsr_process_mul): Modify first_interp field.
4164 (slsr_process_add): Likewise.
4165 (slsr_process_cast): Modify first_interp field for each new
4167 (slsr_process_copy): Likewise.
4168 (dump_candidate): Dump first_interp field.
4169 (replace_mult_candidate): Process all interpretations, not just
4171 (replace_rhs_if_not_dup): Likewise.
4172 (replace_one_candidate): Likewise.
4174 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
4176 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
4178 (aarch64_needs_frame_chain): New function.
4179 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
4181 2018-05-23 Sudakshina Das <sudi.das@arm.com>
4184 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
4185 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
4186 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
4187 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
4188 as true for strict-align.
4189 (aarch64_can_inline_p): Perform checks even when callee has no
4190 attributes to check for strict alignment.
4191 * doc/extend.texi (AArch64 Function Attributes): Document
4193 * doc/invoke.texi: (AArch64 Options): Likewise.
4195 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
4197 PR tree-optimization/85853
4198 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
4199 the handling of the root of the node to...
4200 (vect_slp_analyze_node_operations_1): ...this new function,
4201 and run the whole thing with the child nodes' def types
4202 set according to their SLP node's def type.
4204 2018-05-23 Richard Biener <rguenther@suse.de>
4207 * tree-data-ref.c (create_runtime_alias_checks): Defer
4208 and ignore overflow warnings.
4210 2018-05-23 Yury Gribov <tetra2005@gmail.com>
4212 PR tree-optimization/85822
4213 * tree-vrp.c (is_masked_range_test): Fix handling of negative
4216 2018-05-23 Richard Biener <rguenther@suse.de>
4218 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4219 memset constants via native_interpret_expr.
4221 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
4224 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
4226 (cgraph_node::create_alias): Likewise.
4227 (cgraph_node::get_availability): Check ifunc_resolver instead
4228 of looking up ifunc attribute.
4229 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4230 * varasm.c (do_assemble_alias): Likewise.
4231 (assemble_alias): Likewise.
4232 (default_binds_local_p_3): Likewise.
4233 * cgraph.h (cgraph_node): Add ifunc_resolver.
4234 (cgraph_node::only_called_directly_or_aliased_p): Return false
4236 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
4238 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
4239 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
4240 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
4241 instead of looking up ifunc attribute.
4243 2018-05-22 Luis Machado <luis.machado@linaro.org>
4245 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
4247 2018-05-22 Martin Sebor <msebor@redhat.com>
4250 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
4251 only when expasion succeeds.
4252 (expand_builtin_strcmp): Same.
4253 (expand_builtin_strncmp): Same.
4255 2018-05-22 Martin Sebor <msebor@redhat.com>
4257 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
4259 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4260 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4262 * config/aarch64/aarch64-ldpstp.md: Replace uses of
4263 aarch64_mem_pair_operand with memory_operand and delete operand swapping
4265 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4266 Add check for legitimate_address.
4267 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
4268 (aarch64_swap_ldrstr_operands): New.
4269 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
4272 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4273 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4275 * config/aarch64/aarch64.md: New patterns to generate stp
4277 (store_pair_sw, store_pair_dw): New patterns to generate stp for
4278 single words and double words.
4279 (load_pair_sw, load_pair_dw): Likewise.
4280 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
4282 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
4284 * config/aarch64/aarch64-ldpstp.md: Modify peephole
4285 for different mode ldpstp and add peephole for merged zero stores.
4287 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4289 (aarch64_gen_store_pair): Rename calls to match new patterns.
4290 (aarch64_gen_load_pair): Rename calls to match new patterns.
4291 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
4292 (load_pair<DREG:mode><DREG2:mode>): ... This.
4293 (store_pair<mode>): Rename to...
4294 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
4295 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
4297 (V_INT_EQUIV): Handle SImode.
4298 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
4301 2018-05-22 Martin Sebor <msebor@redhat.com>
4304 * calls.c (maybe_warn_nonstring_arg): Use string length to set
4305 or ajust the presumed bound on an operation to avoid unnecessary
4308 2018-05-22 Martin Sebor <msebor@redhat.com>
4310 PR tree-optimization/85826
4311 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
4312 assuming that a DECL necesarily has a constant size.
4314 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
4317 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
4319 2018-05-22 Richard Biener <rguenther@suse.de>
4321 PR tree-optimization/85834
4322 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
4323 non-constant and non-zero memset arguments.
4325 2018-05-22 Martin Liska <mliska@suse.cz>
4328 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
4330 2018-05-22 Richard Biener <rguenther@suse.de>
4332 PR tree-optimization/85863
4333 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
4334 comparisons when vectype is specified.
4335 (vectorizable_condition): Do not specify vectype for
4336 vect_is_simple_cond when SLP vectorizing.
4338 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
4341 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
4342 define __ibm128 as long double.
4343 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
4344 as a distinct type when IEEE 128-bit support is enabled.
4345 (init_float128_ieee): Fix up conversions between IFmode and IEEE
4346 128-bit types to use the correct functions.
4347 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
4348 convert between 128-bit floating point types that have different
4349 modes but the same representation, instead of using gen_lowpart to
4351 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
4353 (IFKF_reg): New attributes to give the register constraints for
4355 (extend<mode>tf2_internal): New insns to mark an explicit
4356 conversion between 128-bit floating point types that have a
4357 different mode but share the same representation.
4359 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
4361 PR tree-optimization/85814
4362 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
4363 a null return from get_strinfo when unsharing the next
4364 strinfo in the chain.
4366 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4369 * varasm.c (weak_finish): Clean up weak_decls.
4371 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4373 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
4374 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
4375 UNSPEC_UADALP values.
4376 * config/aarch64/iterators.md (ABAL): New int iterator.
4379 (sur): Add mappings for the above.
4380 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
4382 (aarch64_<sur>abal<mode>_4): Likewise.
4383 (aarch64_<sur>adalp<mode>_3): Likewise.
4384 (<sur>sadv16qi): New define_expand.
4386 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
4388 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
4389 (*movdf_internal): Ditto.
4390 (*rcpsf2_sse): Ditto.
4391 (*rsqrtsf2_sse): Ditto.
4392 (*sqrt<mode>2_sse): Ditto.
4394 2018-05-21 Tamar Christina <tamar.christina@arm.com>
4396 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
4398 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
4399 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
4400 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
4401 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
4403 * config/aarch64/arm_neon.h: Likewise.
4404 * config/aarch64/iterators.md (VQ_I): New.
4406 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
4408 * config.gcc: Add arc/t-multilib-linux to tmake_file for
4410 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
4413 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4415 * config/nds32/constraints.md (S): New constraint.
4416 * config/nds32/nds32.md (call_internal): Use constraint S.
4417 (call_value_internal): Likewise.
4418 (sibcall_internal): Likewise.
4419 (sibcall_value_internal): Likewise.
4421 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4422 Chung-Ju Wu <jasonwucj@gmail.com>
4424 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
4427 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4428 Chung-Ju Wu <jasonwucj@gmail.com>
4430 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
4431 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
4432 (nds32_rtx_costs_impl): Simplify.
4433 (nds32_address_cost_impl): Simplify.
4434 (nds32_init_rtx_costs): New function.
4435 (nds32_rtx_costs_speed_prefer): Likewise.
4436 (nds32_rtx_costs_size_prefer): Likewise.
4437 (nds32_address_cost_speed_prefer): Likewise.
4438 (nds32_address_cost_speed_fwprop): Likewise.
4439 (nds32_address_cost_size_prefer): Likewise.
4440 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
4441 * config/nds32/nds32.c (nds32_option_override): Use
4442 nds32_init_rtx_costs function.
4444 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4446 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
4447 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
4448 (TARGET_PIPELINE_N8): Likewise.
4449 (TARGET_PIPELINE_N10): Likewise.
4450 (TARGET_PIPELINE_N13): Likewise.
4451 (TARGET_PIPELINE_GRAYWOLF): Likewise.
4453 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
4455 * config/nds32/nds32-fpu.md: Update copyright year.
4457 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4459 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
4461 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4463 * config/nds32/nds32.c
4464 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
4465 * config/nds32/nds32.opt (minline-asm-r15): New option.
4467 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4469 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
4471 * config/nds32/nds32.md (abssi2): New pattern.
4473 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
4475 * config/i386/i386.md (rex64namesuffix): New mode attribute.
4476 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
4477 Merge insn pattern from sse_cvtsi2ss<round_name> and
4478 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
4479 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
4480 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
4481 using SWI48 mode iterator.
4482 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
4483 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
4484 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
4485 pattern from sse_cvttss2si<round_saeonly_name>
4486 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
4487 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
4488 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
4489 using SWI48 mode iterator.
4490 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
4491 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
4492 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
4493 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
4494 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
4495 using SWI48 mode iterator.
4496 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
4497 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
4498 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
4499 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
4500 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
4501 SWI48 mode iterator.
4502 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
4503 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
4504 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
4505 pattern from sse_cvttsd2si<round_saeonly_name>
4506 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
4508 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4510 * config/nds32/nds32-md-auxiliary.c
4511 (nds32_valid_smw_lwm_base_p): Refine.
4512 (nds32_output_smw_single_word): Refine.
4513 (nds32_output_smw_double_word): New.
4514 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
4516 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4518 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
4519 (nds32_output_stack_pop): Refine.
4520 (nds32_expand_unaligned_load): Refine.
4521 (nds32_expand_unaligned_store): Refine.
4523 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4524 Chung-Ju Wu <jasonwucj@gmail.com>
4526 * config/nds32/constants.md: Add TP_REGNUM constant.
4527 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4528 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4530 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4531 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4532 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4533 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4534 and PIC code generation.
4535 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
4537 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
4539 * config/nds32/nds32.md: Support TLS and PIC.
4540 * config/nds32/nds32.c: Support TLS and PIC.
4541 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
4542 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
4545 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4547 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
4548 mode with E_ prefix.
4550 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4551 Chung-Ju Wu <jasonwucj@gmail.com>
4553 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4554 * config/nds32/nds32-md-auxiliary.c
4555 (symbolic_reference_mentioned_p): New.
4556 (nds32_legitimize_ict_address): New.
4557 (nds32_expand_ict_move): New.
4558 (nds32_indirect_call_referenced_p): New.
4559 (nds32_symbol_binds_local_p): Delete.
4560 (nds32_long_call_p): Modify.
4561 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4562 * config/nds32/nds32-protos.h
4563 (symbolic_reference_mentioned_p): Declare.
4564 (nds32_legitimize_ict_address): Declare.
4565 (nds32_expand_ict_move): Declare.
4566 (nds32_indirect_call_referenced_p): Declare.
4567 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4568 (nds32_relax_group): Use nds32_ict_const_p as condition.
4569 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4570 (nds32_asm_file_start): Output ict_model directive in asm code.
4571 (nds32_legitimate_address_p): Consider indirect call.
4572 (nds32_print_operand): Consider indirect call.
4573 (nds32_print_operand_address): Consider indirect call.
4574 (nds32_insert_attributes): Handle "indirect_call" attribute.
4575 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4576 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4577 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4578 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4579 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4580 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4581 (TARGET_ICT_MODEL_SMALL): Define.
4582 (TARGET_ICT_MODEL_LARGE): Define.
4583 * config/nds32/nds32.md (movsi): Consider ict model.
4584 (call, call_value): Consider ict model.
4585 (sibcall, sibcall_value): Consider ict model.
4586 * config/nds32/nds32.opt (mict-model): New option.
4587 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4590 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4591 Monk Chiang <sh.chiang04@gmail.com>
4592 Jim Wilson <jimw@sifive.com>
4594 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4595 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4596 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4597 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4598 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4599 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4600 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4601 compute save_libcall_adjustment properly.
4602 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4603 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4604 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4605 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4606 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4607 (ABI_SPEC): Handle mabi=ilp32e.
4608 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4609 (RVE): Add RVE mask.
4610 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4611 <-march>: Add rv32e as an example.
4613 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4616 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4617 (intra_create_variable_infos): Handle C++ constructors.
4619 2018-05-18 Martin Liska <mliska@suse.cz>
4621 * passes.def: Remove a redundant pass.
4623 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4626 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4628 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4630 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4632 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4634 (armv2, armv2a, armv3, armv3m): Delete architectures.
4635 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4636 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4637 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4639 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4640 (*mulsidi3adddi): Likewise.
4641 (mulsidi3): Likewise.
4642 (*mulsidi3_nov6): Likewise.
4643 (umulsidi3): Likewise.
4644 (umulsidi3_nov6): Likewise.
4645 (umaddsidi4): Likewise.
4646 (*umulsidi3adddi): Likewise.
4647 (smulsi3_highpart): Likewise.
4648 (*smulsi3_highpart_nov6): Likewise.
4649 (umulsi3_highpart): Likewise.
4650 (*umulsi3_highpart_nov6): Likewise.
4651 * config/arm/arm.h (arm_arch3m): Delete.
4652 * config/arm/arm.c (arm_arch3m): Delete.
4653 (arm_option_override_internal): Update armv3-related comment.
4654 (arm_configure_build_target): Delete use of isa_bit_mode32.
4655 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4656 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4657 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4659 (mulusa3): Likewise.
4660 * config/arm/arm-protos.h (arm_arch3m): Delete.
4661 * config/arm/arm-tables.opt: Regenerate.
4662 * config/arm/arm-tune.md: Likewise.
4663 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4664 deleted architectures.
4666 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4668 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4669 (armv5t, armv5te): New features.
4670 (ARMv5, ARMv5e): Delete fgroups.
4671 (ARMv5t, ARMv5te): Adjust for above changes.
4673 (armv5, armv5e): Delete arches.
4674 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4676 (*call_reg_arm): Likewise.
4677 (*call_value_reg_armv5): Likewise.
4678 (*call_value_reg_arm): Likewise.
4679 (*call_symbol): Likewise.
4680 (*call_value_symbol): Likewise.
4681 (*sibcall_insn): Likewise.
4682 (*sibcall_value_insn): Likewise.
4684 (prefetch): Likewise.
4685 (define_split and define_peephole2 dependent on arm_arch5):
4687 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4689 (TARGET_ARM_QBIT): Likewise.
4690 (TARGET_DSP_MULTIPLY): Likewise.
4691 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4692 (arm_arch5, arm_arch5e): Delete.
4693 (arm_arch5t, arm_arch5te): Declare.
4694 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4695 (arm_arch5t): Declare.
4696 (arm_option_reconfigure_globals): Update for the above.
4697 (arm_options_perform_arch_sanity_checks): Update comment, replace
4698 use of arm_arch5 with arm_arch5t.
4699 (use_return_insn): Likewise.
4700 (arm_emit_call_insn): Likewise.
4701 (output_return_instruction): Likewise.
4702 (arm_final_prescan_insn): Likewise.
4703 (arm_coproc_builtin_available): Likewise.
4704 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4705 arm_arch5e with arm_arch5t and arm_arch5te.
4706 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4707 (arm_arch5t, arm_arch5te): Declare.
4708 * config/arm/arm-tables.opt: Regenerate.
4709 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4710 * config/arm/t-multilib: Likewise.
4711 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4712 instead of arm_arch5.
4713 (*call_reg_thumb1): Likewise.
4714 (*call_value_reg_thumb1_v5): Likewise.
4715 (*call_value_reg_thumb1): Likewise.
4716 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4718 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4720 2018-05-18 Martin Liska <mliska@suse.cz>
4722 PR gcov-profile/84846
4723 * doc/gcov.texi: Document -t option of gcov tool.
4725 2018-05-18 Martin Liska <mliska@suse.cz>
4727 PR gcov-profile/84846
4728 * gcov.c (print_usage): Add new -t option.
4729 (process_args): Handle the option.
4730 (generate_results): Use stdout as output when requested by
4733 2018-05-18 Martin Liska <mliska@suse.cz>
4735 PR gcov-profile/84846
4736 * coverage.c (coverage_init): Write PWD to .gcno file.
4737 * doc/gcov.texi: Document how working directory is printed.
4738 * gcov-dump.c (dump_gcov_file): Print PWD.
4739 * gcov.c (output_intermediate_file): Likewise.
4740 (read_graph_file): Read PWD string.
4741 (output_lines): Print PWD.
4743 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4746 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4747 for retval and return false if all args to phi are zero.
4749 2018-05-18 Richard Biener <rguenther@suse.de>
4751 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4753 (evrp_dom_walker::before_dom_children): Call it.
4755 2018-05-18 Richard Biener <rguenther@suse.de>
4757 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4758 results when processing array refs with variable index.
4760 2018-05-18 Toon Moene <toon@moene.org>
4762 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4763 directly after that of -floop-interchange. Indicate that both
4764 options are enabled by default when specifying -O3.
4766 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4768 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4769 iterator. Delete separate integer-mode vec_set<mode> expander.
4770 (aarch64_simd_vec_setv2di): Delete.
4771 (vec_setv2di): Delete.
4772 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4773 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4774 the "w, r" alternative.
4776 2018-05-18 Martin Liska <mliska@suse.cz>
4778 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4779 * tree-pass.h (make_pass_lower_switch_O0): New function.
4780 * tree-switch-conversion.c (node_has_low_bound): Remove.
4781 (node_has_high_bound): Likewise.
4782 (node_is_bounded): Likewise.
4783 (class pass_lower_switch): Make it a template type and create
4785 (pass_lower_switch::execute): Add template argument.
4786 (make_pass_lower_switch): New function.
4787 (make_pass_lower_switch_O0): New function.
4788 (do_jump_if_equal): Remove.
4789 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4790 all the hard work to tree optimization passes.
4792 2018-05-18 Martin Liska <mliska@suse.cz>
4794 * dbgcnt.c (limit_low): Renamed from limit.
4795 (limit_high): New variable.
4796 (dbg_cnt_is_enabled): Check for upper limit.
4797 (dbg_cnt): Adjust dumping.
4798 (dbg_cnt_set_limit_by_index): Add new argument for high
4800 (dbg_cnt_set_limit_by_name): Likewise.
4801 (dbg_cnt_process_single_pair): Parse new format.
4802 (dbg_cnt_process_opt): Use strtok.
4803 (dbg_cnt_list_all_counters): Remove 'value' and add
4805 * doc/invoke.texi: Document changes.
4807 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4809 * doc/sourcebuild.texi (scalar_all_fma): Document.
4810 * tree.def (FMA_EXPR): Delete.
4811 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4812 * internal-fn.c (ternary_direct): New macro.
4813 (expand_ternary_optab_fn): Likewise.
4814 (direct_ternary_optab_supported_p): Likewise.
4815 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4816 * builtins.c (fold_builtin_fma): Delete.
4817 (fold_builtin_3): Don't call it.
4818 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4819 * expr.c (expand_expr_real_2): Likewise.
4820 * fold-const.c (operand_equal_p): Likewise.
4821 (fold_ternary_loc): Likewise.
4822 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4823 * gimple.c (DEFTREECODE): Likewise.
4824 * gimplify.c (gimplify_expr): Likewise.
4825 * optabs-tree.c (optab_for_tree_code): Likewise.
4826 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4827 * tree-eh.c (operation_could_trap_p): Likewise.
4828 (stmt_could_throw_1_p): Likewise.
4829 * tree-inline.c (estimate_operator_cost): Likewise.
4830 * tree-pretty-print.c (dump_generic_node): Likewise.
4831 (op_code_prio): Likewise.
4832 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4833 * tree-ssa-operands.c (get_expr_operands): Likewise.
4834 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4835 * fold-const-call.h (fold_fma): Delete.
4836 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4837 CFN_FNMA and CFN_FNMS.
4839 * genmatch.c (combined_fn): New enum.
4840 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4841 (commutative_op): New function.
4842 (commutate): Use it. Handle more than 2 operands.
4843 (dt_operand::gen_gimple_expr): Use commutative_op.
4844 (parser::parse_expr): Allow :c to be used with non-binary
4845 operators if the commutative operand is known.
4846 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4847 CFN_FMS, CFN_FNMA and CFN_FNMS.
4848 (backprop::process_assign_use): Remove FMA_EXPR handling.
4849 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4850 (gen_hsa_fma): New function.
4851 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4852 IFN_FNMA and IFN_FNMS.
4853 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4854 * gimple-fold.h (follow_all_ssa_edges): Declare.
4855 * gimple-fold.c (follow_all_ssa_edges): New function.
4856 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4857 gimple_build interface and use follow_all_ssa_edges to fold the result.
4858 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4859 instead of checking for optabs directly.
4860 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4861 rather than FMA_EXPRs.
4862 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4863 call to IFN_FMA instead of an FMA_EXPR.
4865 2018-05-17 Jim Wilson <jimw@sifive.com>
4867 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4868 sign extended promoted subreg, and the range does not have the sign bit
4869 set, then do a sign extend.
4871 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4872 test, check for sign extended subreg and/or constant operands, and
4873 do a sign extend in that case.
4875 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4877 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4878 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4880 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4881 Change logics_shift_reg to logics_shift_imm.
4882 (thunderx2t99_fp_loadpair_basic): Delete.
4883 (thunderx2t99_fp_storepair_basic): Delete.
4884 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4885 (thunderx2t99_asimd_polynomial): Delete.
4886 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4887 and neon_fp_mul_d_scalar_q.
4888 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4889 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4890 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4891 (thunderx2t99_asimd_lut): Add missing tbl types.
4892 (thunderx2t99_asimd_ext): Delete.
4893 (thunderx2t99_asimd_load1_1_mult): Delete.
4894 (thunderx2t99_asimd_load1_2_mult): Delete.
4895 (thunderx2t99_asimd_load1_ldp): New.
4896 (thunderx2t99_asimd_load1): New.
4897 (thunderx2t99_asimd_load2): Add missing *load2* types.
4898 (thunderx2t99_asimd_load3): New.
4899 (thunderx2t99_asimd_load4): New.
4900 (thunderx2t99_asimd_store1_1_mult): Delete.
4901 (thunderx2t99_asimd_store1_2_mult): Delete.
4902 (thunderx2t99_asimd_store2_mult): Delete.
4903 (thunderx2t99_asimd_store2_onelane): Delete.
4904 (thunderx2t99_asimd_store_stp): New.
4905 (thunderx2t99_asimd_store1): New.
4906 (thunderx2t99_asimd_store2): New.
4907 (thunderx2t99_asimd_store3): New.
4908 (thunderx2t99_asimd_store4): New.
4910 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4912 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4913 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4915 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4916 Segher Boessenkool <segher@kernel.crashing.org>
4919 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4922 2018-05-17 Richard Biener <rguenther@suse.de>
4924 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4925 for pruning loop and prune defs feeding only already visited PHIs.
4927 2018-05-17 Richard Biener <rguenther@suse.de>
4929 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4931 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4932 Richard Biener <rguenther@suse.de>
4934 PR tree-optimization/85793
4935 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4936 for VMAT_ELEMENTWISE.
4938 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4940 * internal-fn.h (lookup_internal_fn): Declare
4941 * internal-fn.c (lookup_internal_fn): New function.
4942 * gimple.c (gimple_build_call_from_tree): Handle calls to
4944 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4945 internal function names.
4946 * tree-pretty-print.c (dump_generic_node): Likewise.
4947 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4949 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4951 * gimple-fold.h (gimple_build): Make the function forms take
4952 combined_fn rather than built_in_function.
4953 (gimple_simplify): Likewise.
4954 * gimple-match-head.c (gimple_simplify): Likewise.
4955 * gimple-fold.c (gimple_build): Likewise.
4956 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4957 rather than gimple_build_call_internal.
4958 (get_initial_defs_for_reduction): Likewise.
4959 (vect_create_epilog_for_reduction): Likewise.
4960 (vectorizable_live_operation): Likewise.
4962 2018-05-17 Martin Liska <mliska@suse.cz>
4964 * gimple-ssa-sprintf.c (format_directive): Do not use
4965 space in between 'G_' and '('.
4967 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4970 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4971 even if the mask is not all ones.
4974 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4976 (ix86_gimple_fold_builtin): Likewise.
4979 * config/i386/i386.c: Include tree-vector-builder.h.
4980 (ix86_vector_shift_count): New function.
4981 (ix86_fold_builtin): Fold shift builtins by scalar count.
4982 (ix86_gimple_fold_builtin): Likewise.
4984 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4985 _mm512_setzero): New intrinsics.
4987 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4988 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4990 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4991 code generation for cases where splatting a value is not useful.
4992 * simplify-rtx.c (simplify_ternary_operation): Simplify
4993 vec_merge across a vec_duplicate and a paradoxical subreg forming
4994 a vector mode to a vec_concat.
4996 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4998 * config.gcc: Support "goldmont-plus".
4999 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5001 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5002 PROCESSOR_GOLDMONT_PLUS.
5003 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
5004 (processor_target_table): Add "goldmont-plus".
5005 (PTA_GOLDMONT_PLUS): Define.
5006 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
5007 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
5008 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
5009 (fold_builtin_cpu): Add "goldmont-plus".
5010 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
5011 (ix86_option_override_internal): Add "goldmont-plus".
5012 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
5013 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
5014 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
5015 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
5017 2018-05-17 Richard Biener <rguenther@suse.de>
5019 PR tree-optimization/85757
5020 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
5021 remove defs that only feed that PHI from further processing.
5023 2018-05-16 Jim Wilson <jimw@sifive.com>
5025 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
5027 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
5029 2018-05-16 Mark Wielaard <mark@klomp.org>
5031 * dwarf2out.c (count_index_strings): New function.
5032 (output_indirect_strings): Call count_index_strings and generate
5033 header for dwarf_version >= 5.
5035 2018-05-16 Mark Wielaard <mark@klomp.org>
5037 * dwarf2out.c (dwarf_FORM): New function.
5038 (set_indirect_string): Use dwarf_FORM.
5039 (reset_indirect_string): Likewise.
5040 (size_of_die): Likewise.
5041 (value_format): Likewise.
5042 (output_die): Likewise.
5043 (add_skeleton_AT_string): Likewise.
5044 (output_macinfo_op): Likewise.
5045 (index_string): Likewise.
5046 (output_index_string_offset): Likewise.
5047 (output_index_string): Likewise.
5048 (count_index_strings): Likewise.
5050 2018-05-16 Carl Love <cel@us.ibm.com>
5052 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5053 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
5055 2018-05-16 Martin Jambor <mjambor@suse.cz>
5057 * ipa-prop.c (ipa_free_all_edge_args): Remove.
5058 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
5060 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
5062 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
5063 (fnma<mode>4): Likewise.
5064 (fms<mode>4): Likewise.
5065 (fnms<mode>4): Likewise.
5066 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
5067 (aarch64_fnma<mode>4): Likewise.
5068 (aarch64_fms<mode>4): Likewise.
5069 (aarch64_fnms<mode>4): Likewise.
5070 (aarch64_fnmadd<mode>4): Likewise.
5072 2018-05-16 Jason Merrill <jason@redhat.com>
5074 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
5076 2018-05-16 Richard Biener <rguenther@suse.de>
5078 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
5079 (dump_stmt_cost): Declare.
5080 (add_stmt_cost): Dump cost we add.
5081 (add_stmt_costs): New function.
5082 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
5084 (vect_analyze_stmt): Adjust prototype.
5085 (vectorizable_condition): Likewise.
5086 (vectorizable_live_operation): Likewise.
5087 (vectorizable_reduction): Likewise.
5088 (vectorizable_induction): Likewise.
5089 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
5090 cost vector to pass to vectorizable_ and record afterwards.
5091 (vect_model_reduction_cost): Take cost vector argument and adjust.
5092 (vect_model_induction_cost): Likewise.
5093 (vectorizable_reduction): Likewise.
5094 (vectorizable_induction): Likewise.
5095 (vectorizable_live_operation): Likewise.
5096 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
5097 SLP_TREE_NUMBER_OF_VEC_STMTS.
5098 (vect_analyze_slp_cost_1): Remove.
5099 (vect_analyze_slp_cost): Likewise.
5100 (vect_slp_analyze_node_operations): Take visited args and
5101 a target cost vector. Avoid processing already visited stmt sets.
5102 (vect_slp_analyze_operations): Use a local cost vector to gather
5103 costs and register those of non-discarded instances.
5104 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
5105 (vect_schedule_slp_instance): Remove copying of
5106 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
5108 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
5109 adding cost. Record cost entry location.
5110 (vect_prologue_cost_for_slp_op): Function to compute cost of
5111 a constant or invariant generated for SLP vect in the prologue,
5112 split out from vect_analyze_slp_cost_1.
5113 (vect_model_simple_cost): Make static. Adjust for SLP costing.
5114 (vect_model_promotion_demotion_cost): Likewise.
5115 (vect_model_store_cost): Likewise, make static.
5116 (vect_model_load_cost): Likewise.
5117 (vectorizable_bswap): Add cost vector arg and adjust.
5118 (vectorizable_call): Likewise.
5119 (vectorizable_simd_clone_call): Likewise.
5120 (vectorizable_conversion): Likewise.
5121 (vectorizable_assignment): Likewise.
5122 (vectorizable_shift): Likewise.
5123 (vectorizable_operation): Likewise.
5124 (vectorizable_store): Likewise.
5125 (vectorizable_load): Likewise.
5126 (vectorizable_condition): Likewise.
5127 (vectorizable_comparison): Likewise.
5128 (can_vectorize_live_stmts): Likewise.
5129 (vect_analyze_stmt): Likewise.
5130 (vect_transform_stmt): Adjust calls to vectorizable_*.
5131 * tree-vectorizer.c: Include gimple-pretty-print.h.
5132 (dump_stmt_cost): New function.
5134 2018-05-16 Richard Biener <rguenther@suse.de>
5136 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
5137 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
5138 * tree-ssa-dse.c: Include tree-ssa-loop.h.
5139 (check_name): New callback.
5140 (dse_classify_store): Track cycles via a visited bitmap of PHI
5141 defs and simplify handling of in-loop and across loop dead stores
5142 and properly fail for loop-variant refs. Handle byte-tracking with
5143 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
5146 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
5148 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
5149 (vect_get_mask_type_for_stmt): Likewise.
5150 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
5152 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
5153 to determine the statement's vector type and the vector type that
5154 should be used for calculating nunits. Deal with cases in which
5155 the type has to be deferred.
5156 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
5157 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
5158 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
5159 (vect_determine_vf_for_stmt): New functions, split out from...
5160 (vect_determine_vectorization_factor): ...here.
5161 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
5162 (vect_get_mask_type_for_stmt): New functions, split out from
5163 vect_determine_vectorization_factor.
5165 2018-05-16 Richard Biener <rguenther@suse.de>
5167 * tree-cfg.c (verify_gimple_assign_ternary): Properly
5168 verify the [VEC_]COND_EXPR embedded comparison.
5170 2018-05-15 Martin Sebor <msebor@redhat.com>
5172 PR tree-optimization/85753
5173 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
5174 RECORD_TYPE in addition to ARRAY_TYPE.
5176 2018-05-15 Martin Sebor <msebor@redhat.com>
5179 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
5181 2018-05-15 Richard Biener <rguenther@suse.de>
5183 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
5184 add by_clobber_p one. Change algorithm to collect all defs
5185 representing uses we need to walk and try reducing them to
5186 a single one before failing.
5187 (dse_dom_walker::dse_optimize_stmt): Adjust.
5189 2018-05-13 Mark Wielaard <mark@klomp.org>
5191 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
5192 (size_of_loc_descr): Likewise.
5193 (output_loc_operands): Likewise.
5194 (output_loc_operands_raw): Likewise.
5195 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
5196 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
5197 (hash_loc_operands): Likewise.
5198 (compare_loc_operands): Likewise.
5200 2018-05-14 Mark Wielaard <mark@klomp.org>
5202 * dwarf2out.c (count_index_addrs): New function.
5203 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
5205 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5207 PR tree-optimization/83648
5208 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
5209 return value as malloc candidate.
5211 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5214 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
5215 param as true in call to suggest_attribute.
5217 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
5219 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
5220 -mreadonly-in-sdata.
5222 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5224 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
5226 (aarch64_crypto_aesd_fused): Likewise.
5228 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
5230 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
5231 (movsi_aarch64): Likewise.
5232 (load_pairsi): Likewise.
5233 (load_pairdi): Likewise.
5234 (store_pairsi): Likewise.
5235 (store_pairdi): Likewise.
5236 (load_pairsf): Likewise.
5237 (load_pairdf): Likewise.
5238 (store_pairsf): Likewise.
5239 (store_pairdf): Likewise.
5240 (zero_extend): Likewise.
5241 (trunc): Swap alternatives.
5242 (fcvt_target): Add '?' to prefer w over r.
5244 2018-05-14 Jakub Jelinek <jakub@redhat.com>
5247 * config/i386/i386.md: Disallow non-commutative arithmetics in
5248 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
5249 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
5250 in the peephole2 before it.
5252 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
5254 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
5255 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
5256 (ix86_handle_option): Handle -mcldemote.
5257 * config.gcc: New header.
5258 * config/i386/cldemoteintrin.h: New file.
5259 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
5260 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5262 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5263 OPTION_MASK_ISA_CLDEMOTE.
5264 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
5265 (ix86_valid_target_attribute_inner_p): Ditto.
5266 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
5267 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
5268 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
5269 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
5270 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
5272 * config/i386/i386.opt: Add -mcldemote.
5273 * config/i386/x86intrin.h: New header.
5274 * doc/invoke.texi: Add -mcldemote.
5276 2018-05-14 Richard Biener <rguenther@suse.de>
5278 * doc/match-and-simplify.texi: Adjust :s documentation.
5280 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
5282 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
5283 intended memcpy size.
5284 (REORDER_45): Likewise.
5286 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
5288 * sort.cc: New file.
5289 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
5290 * vec.c (qsort_chk): Use gcc_qsort.
5291 * Makefile.in (OBJS-libcommon): Add sort.o.
5292 (build/sort.o): New target. Use it...
5293 (BUILD_RTL): ... here, and...
5294 (build/gencfn-macros): ... here, and...
5295 (build/genmatch): ... here.
5297 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5298 Chung-Ju Wu <jasonwucj@gmail.com>
5300 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
5301 * config/nds32/nds32-graywolf.md: New file.
5302 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
5303 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
5305 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
5306 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
5307 * config/nds32/nds32.md (pipeline_model): Add graywolf.
5308 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
5309 * config/nds32/pipelines.md: Include n15 settings.
5311 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5312 Chung-Ju Wu <jasonwucj@gmail.com>
5314 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
5315 * config/nds32/nds32-n13.md: New file.
5316 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
5317 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
5319 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
5320 * config/nds32/nds32.md (pipeline_model): Add n13.
5321 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
5322 * config/nds32/pipelines.md: Include n13 settings.
5324 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5325 Chung-Ju Wu <jasonwucj@gmail.com>
5327 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
5328 * config/nds32/nds32-n10.md: New file.
5329 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
5330 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
5332 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
5333 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
5334 * config/nds32/nds32.md (pipeline_model): Add n10.
5335 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
5336 * config/nds32/pipelines.md: Include n10 settings.
5338 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
5339 Kito Cheng <kito.cheng@gmail.com>
5340 Chung-Ju Wu <jasonwucj@gmail.com>
5342 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
5343 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5344 Add enum values for DSP extension instructions.
5345 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
5347 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
5348 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
5350 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
5351 * config/nds32/nds32-dspext.md: New file for DSP implementation.
5352 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
5353 * config/nds32/nds32-intrinsic.md: Likewise.
5354 * config/nds32/nds32_intrinsic.h: Likewise.
5355 * config/nds32/nds32-md-auxiliary.c: Likewise.
5356 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
5357 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
5358 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
5359 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
5360 * config/nds32/nds32-protos.h: New declarations for DSP extension.
5361 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
5362 TYPE_DMAC in switch statement.
5363 * config/nds32/nds32.c: New checking and implementation for DSP
5364 extension instructions.
5365 * config/nds32/nds32.h: Likewise.
5366 * config/nds32/nds32.md: Likewise.
5367 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
5368 * config/nds32/predicates.md: Implement new predicates for DSP
5371 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5373 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
5374 Reformat alternatives and attributes so it is easier to identify
5375 which constraints/attributes go with which instruction.
5376 (mov<mode>_hardfloat32, FMOVE64): Likewise.
5377 (mov<mode>_softfloat32, FMOVE64): Likewise.
5378 (mov<mode>_hardfloat64, FMOVE64): Likewise.
5379 (mov<mode>_softfloat64, FMOVE64): Likewise.
5381 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
5383 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5385 (Basic PowerPC Built-in Functions): The new name of the
5386 subsection previously known as "PowerPC Built-in Functions".
5387 (Basic PowerPC Built-in Functions Available on all Configurations):
5389 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
5390 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
5391 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
5392 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
5394 2018-05-11 Martin Jambor <mjambor@suse.cz>
5397 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
5400 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5403 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
5405 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
5407 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
5408 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
5409 (ix86_handle_option): Handle -mwaitpkg.
5410 * config.gcc: New header.
5411 * config/i386/cpuid.h (bit_WAITPKG): New bit.
5412 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
5413 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
5415 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5416 OPTION_MASK_ISA_WAITPKG.
5417 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
5418 (ix86_option_override_internal): Add PTA_WAITPKG.
5419 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
5420 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
5421 IX86_BUILTIN_TPAUSE.
5422 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
5423 __builtin_ia32_umwait and __builtin_ia32_tpause.
5424 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
5425 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
5426 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
5427 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
5428 UNSPECV_TPAUSE): New.
5429 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
5430 * config/i386/i386.opt: Add -mwaitpkg.
5431 * config/i386/waitpkgintrin.h: New file.
5432 * config/i386/x86intrin.h: New header.
5433 * doc/invoke.texi: Add -mwaitpkg.
5435 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5438 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
5440 (cortex-m0): Use armv6s-m isa.
5441 (cortex-m0plus): Likewise.
5442 (cortex-m1): Likewise.
5443 (cortex-m0.small-multiply): Likewise.
5444 (cortex-m0plus.small-multiply): Likewise.
5445 (cortex-m1.small-multiply): Likewise.
5447 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
5448 Jakub Jelinek <jakub@redhat.com>
5450 PR tree-optimization/85692
5451 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
5452 source permute as well.
5454 2018-05-11 Martin Liska <mliska@suse.cz>
5457 * doc/extend.texi: Document LLVM style format for no_sanitize
5460 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5462 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
5463 mode_supports_vsx_dform_quad to mode_supports_dq_form.
5464 (mode_supports_vsx_dform_quad): Likewise.
5465 (mode_supports_vmx_dform): Move these functions to be next to the
5466 other mode_supports functions.
5467 (mode_supports_dq_form): Likewise.
5468 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
5469 mode_supports_dq_form.
5470 (reg_offset_addressing_ok_p): Likewise.
5471 (offsettable_ok_by_alignment): Likewise.
5472 (rs6000_legitimate_offset_address_p): Likewise.
5473 (legitimate_lo_sum_address_p): Likewise.
5474 (rs6000_legitimize_address): Likewise.
5475 (rs6000_legitimize_reload_address): Likewise.
5476 (rs6000_secondary_reload_inner): Likewise.
5477 (rs6000_preferred_reload_class): Likewise.
5478 (rs6000_output_move_128bit): Likewise.
5480 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5482 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
5483 Generate SImode target register for null target.
5484 <case IX86_BUILTIN_XGETBV>: Ditto.
5485 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
5486 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
5488 2018-05-10 Carl Love <cel@us.ibm.com>
5490 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5491 dcbtt and dcbtstt if operands[2] is 0.
5493 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5496 * config/i386/sse.md (usadv64qi): New expander.
5498 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
5500 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
5501 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
5502 -maltivec=be support.
5503 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
5504 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
5505 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
5506 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
5507 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
5508 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
5509 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
5510 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
5511 altivec_vsumsws): Adjust.
5512 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
5513 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
5514 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
5516 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
5517 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
5518 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
5519 (altivec_lve<VI_char>x): Delete expand.
5520 (*altivec_lve<VI_char>x_internal): Rename to...
5521 (altivec_lve<VI_char>x): ... this.
5522 (altivec_lvxl_<mode>): Delete expand.
5523 (*altivec_lvxl_<mode>_internal): Rename to ...
5524 (altivec_lvxl_<mode>): ... this.
5525 (altivec_stvxl_<mode>): Delete expand.
5526 (*altivec_stvxl_<mode>_internal): Rename to ...
5527 (altivec_stvxl_<mode>): ... this.
5528 (altivec_stve<VI_char>x): Delete expand.
5529 (*altivec_stve<VI_char>x_internal): Rename to ...
5530 (altivec_stve<VI_char>x): ... this.
5531 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5532 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5533 reduc_plus_scal_<mode>): Adjust.
5534 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
5536 (rs6000_cpu_cpp_builtins): Adjust.
5537 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
5538 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
5539 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
5540 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
5541 -maltivec=be support.
5542 (rs6000_split_vec_extract_var): Adjust.
5543 (rs6000_split_v4si_init): Adjust.
5544 (swap_selector_for_mode): Delete.
5545 (altivec_expand_lvx_be, altivec_expand_stvx_be,
5546 altivec_expand_stvex_be): Delete.
5547 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
5548 -maltivec=be support.
5549 (rs6000_gimple_fold_builtin): Ditto.
5550 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5552 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5553 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5554 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5555 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5556 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5557 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5558 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5559 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5560 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5561 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5562 anonymous split): Adjust.
5563 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5564 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5566 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5568 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5569 when --with-gxx-include-dir is also specified.
5570 * configure: Regenerate.
5572 2018-05-09 Jim Wilson <jimw@sifive.com>
5575 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5576 * config/riscv/t-withmultilib: New.
5577 * config/riscv/withmultilib.h: New.
5578 * doc/install.texi: Document RISC-V --with-multilib-list support.
5580 2018-05-09 Richard Biener <rguenther@suse.de>
5582 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5584 (vect_bb_vectorization_profitable_p): Adjust. Compute
5585 actual scalar cost using the cost vector and the add_stmt_cost
5588 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5590 PR rtl-optimization/85645
5591 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5592 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5594 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5596 PR rtl-optimization/85645
5597 * shrink-wrap.c (spread_components): Return a boolean saying if
5598 anything was changed.
5599 (try_shrink_wrapping_separate): Iterate spread_components until
5600 nothing changes anymore.
5602 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5604 PR rtl-optimization/85645
5605 * regrename.c (build_def_use): Also kill the chains that include the
5606 destination of a REG_CFA_REGISTER note.
5608 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5610 PR rtl-optimization/85645
5611 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5612 insn that has a REG_CFA_REGISTER note.
5614 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5616 * cfgexpand.c (expand_clobber): New function.
5617 (expand_gimple_stmt_1): Use it.
5618 * tree-vect-stmts.c (vect_clobber_variable): New function,
5620 (vectorizable_simd_clone_call): ...here.
5621 (vectorizable_store): Emit a clobber either side of an
5622 IFN_STORE_LANES sequence.
5623 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5625 2018-05-09 Tom de Vries <tom@codesourcery.com>
5628 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5629 (define_insn "trap_if_false"): Add exit after trap.
5631 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5633 PR rtl-optimization/85638
5634 * bb-reorder.c: Include common/common-target.h.
5635 (create_forwarder_block): New function extracted from...
5636 (fix_up_crossing_landing_pad): ...here. Rename into...
5637 (dw2_fix_up_crossing_landing_pad): ...this.
5638 (sjlj_fix_up_crossing_landing_pad): New function.
5639 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5640 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5641 from both partitions and exit the loop after one iteration.
5643 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5646 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5648 (Basic PowerPC Built-in Functions): The new name of the
5649 subsection previously known as "PowerPC Built-in Functions".
5650 (Basic PowerPC Built-in Functions Available on all Configurations):
5652 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5654 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5656 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5658 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5661 2018-05-08 Jim Wilson <jimw@sifive.com>
5663 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5664 (LD_EMUL_SUFFIX): New.
5665 (LINK_SPEC): Use it.
5667 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5669 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5671 (Basic PowerPC Built-in Functions): The new name of the
5672 subsection previously known as "PowerPC Built-in Functions".
5673 (Basic PowerPC Built-in Functions Available on all Configurations):
5675 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5677 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5679 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5681 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5684 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5687 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5688 after cmpelim optimization.
5690 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5692 * config.gcc: Support "goldmont".
5693 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5694 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5696 * config/i386/i386.c (m_GOLDMONT): Define.
5697 (processor_target_table): Add "goldmont".
5698 (PTA_GOLDMONT): Define.
5699 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5700 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5701 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5702 (fold_builtin_cpu): Add "goldmont".
5703 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5704 (ix86_option_override_internal): Add "goldmont".
5705 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5706 (processor_type): Add PROCESSOR_GOLDMONT.
5707 * config/i386/i386.md: Add CPU "glm".
5708 * config/i386/glm.md: New file.
5709 * config/i386/x86-tune.def: Add m_GOLDMONT.
5710 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5712 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5715 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5717 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5718 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5719 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5722 * config/i386/i386.c (ix86_fold_builtin): Handle
5723 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5726 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5727 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5729 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5732 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5733 (check_arch): Likewise.
5734 (check_fpu): Return the result rather than printing it.
5735 (end arch): Fix operator precedence.
5736 (end cpu): Likewise.
5737 (END): Print the result from check_fpu.
5739 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5740 Alan Hayward <alan.hayward@arm.com>
5741 David Sherwood <david.sherwood@arm.com>
5743 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5744 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5745 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5746 (*fcmuo<mode>_and): New patterns.
5748 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5750 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5751 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5752 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5753 (cmp_op, sve_imm_con): New code attributes.
5754 (SVE_COND_INT_CMP, imm_con): Delete.
5755 (cmp_op): Remove above unspecs from int attribute.
5756 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5758 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5759 comparison-specific unspecs.
5760 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5761 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5762 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5763 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5764 (*vec_fcm<cmp_op><mode>): Rename to...
5765 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5766 (*vec_fcmuo<mode>): Rename to...
5767 (*fcmuo<mode>): ...this and adjust likewise.
5768 (*pred_fcm<cmp_op><mode>): New pattern.
5769 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5770 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5772 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5774 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5775 (aarch64_emit_sve_predicated_cond): New function.
5776 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5777 (aarch64_emit_unspec_cond_or): Replace with...
5778 (aarch64_emit_sve_or_conds): ...this new function. Use
5779 aarch64_emit_sve_ptrue_op for the individual comparisons and
5780 aarch64_emit_binop to OR them together.
5781 (aarch64_emit_inverted_unspec_cond): Replace with...
5782 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5783 aarch64_emit_sve_ptrue_op for the comparison and
5784 aarch64_emit_unop to invert the result.
5785 (aarch64_expand_sve_vec_cmp_float): Update after the above
5786 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5788 2018-05-07 Nathan Sidwell <nathan@acm.org>
5790 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5791 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5792 (Backwards Compatibility): Likewise.
5794 2018-05-07 Luis Machado <luis.machado@linaro.org>
5798 2018-05-07 Luis Machado <luis.machado@linaro.org>
5800 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5801 <prefetch_dynamic_strides>: New const bool field.
5802 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5803 prefetch_dynamic_strides.
5804 (exynosm1_prefetch_tune): Likewise.
5805 (thunderxt88_prefetch_tune): Likewise.
5806 (thunderx_prefetch_tune): Likewise.
5807 (thunderx2t99_prefetch_tune): Likewise.
5808 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5810 (aarch64_override_options_internal): Update to set
5811 PARAM_PREFETCH_DYNAMIC_STRIDES.
5812 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5813 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5814 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5815 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5816 prefetch-dynamic-strides setting.
5818 2018-05-07 Luis Machado <luis.machado@linaro.org>
5820 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5821 <minimum_stride>: New const int field.
5822 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5823 minimum_stride field.
5824 (exynosm1_prefetch_tune): Likewise.
5825 (thunderxt88_prefetch_tune): Likewise.
5826 (thunderx_prefetch_tune): Likewise.
5827 (thunderx2t99_prefetch_tune): Likewise.
5828 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5829 (aarch64_override_options_internal): Update to set
5830 PARAM_PREFETCH_MINIMUM_STRIDE.
5831 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5832 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5833 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5834 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5835 stride is constant and is below the minimum stride threshold.
5837 2018-05-07 Luis Machado <luis.machado@linaro.org>
5839 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5842 2018-05-07 Luis Machado <luis.machado@linaro.org>
5844 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5845 <prefetch_dynamic_strides>: New const bool field.
5846 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5847 prefetch_dynamic_strides.
5848 (exynosm1_prefetch_tune): Likewise.
5849 (thunderxt88_prefetch_tune): Likewise.
5850 (thunderx_prefetch_tune): Likewise.
5851 (thunderx2t99_prefetch_tune): Likewise.
5852 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5854 (aarch64_override_options_internal): Update to set
5855 PARAM_PREFETCH_DYNAMIC_STRIDES.
5856 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5857 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5858 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5859 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5860 prefetch-dynamic-strides setting.
5862 2018-05-07 Luis Machado <luis.machado@linaro.org>
5864 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5865 <minimum_stride>: New const int field.
5866 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5867 minimum_stride field.
5868 (exynosm1_prefetch_tune): Likewise.
5869 (thunderxt88_prefetch_tune): Likewise.
5870 (thunderx_prefetch_tune): Likewise.
5871 (thunderx2t99_prefetch_tune): Likewise.
5872 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5873 (aarch64_override_options_internal): Update to set
5874 PARAM_PREFETCH_MINIMUM_STRIDE.
5875 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5876 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5877 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5878 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5879 stride is constant and is below the minimum stride threshold.
5881 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5884 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5885 the type is addressable. Don't force op into register if it has
5888 2018-05-05 Roland McGrath <mcgrathr@google.com>
5891 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5892 any section for which we don't know a specific type it should have,
5893 regardless of name. Previously this was done only for the exact
5894 names ".init_array", ".fini_array", and ".preinit_array".
5895 (default_elf_asm_named_section): Add comment about
5896 relationship with default_section_type_flags and SECTION_NOTYPE.
5897 (get_section): Don't consider it a type conflict if one side has
5898 SECTION_NOTYPE and the other doesn't, as long as neither has the
5899 SECTION_BSS et al used in the default_section_type_flags logic.
5901 2018-05-05 Tom de Vries <tom@codesourcery.com>
5904 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5905 (workaround_barsyncs): New function.
5906 (nvptx_reorg): Use workaround_barsyncs.
5907 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5908 (define_expand "nvptx_membar_cta"): New define_expand.
5909 (define_insn "*nvptx_membar_cta"): New insn.
5911 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5913 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5914 To improve optimization opportunities.
5915 * builtin-types.def: The new needed builtin types for the above.
5917 2018-05-04 Richard Biener <rguenther@suse.de>
5919 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5920 * gimple-ssa-store-merging.c
5921 (imm_store_chain_info::output_merged_store): Remove redundant create,
5922 release split_store vector contents on failure.
5923 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5924 scalar stmt vector on cache hit.
5926 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5928 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5930 * config.gcc (powerpc-xilinx-eabi*): Remove.
5931 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5933 (fusion_addis_mem_combo_load): Ditto.
5934 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5936 (rs6000_cpu_cpp_builtins): Ditto.
5937 * config/rs6000/rs6000-linux.c
5938 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5939 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5940 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5942 (rs6000_setup_reg_addr_masks): Ditto.
5943 (rs6000_init_hard_regno_mode_ok): Ditto.
5944 (rs6000_option_override_internal): Ditto.
5945 (legitimate_lo_sum_address_p): Ditto.
5946 (rs6000_legitimize_address): Ditto.
5947 (rs6000_legitimize_reload_address): Ditto.
5948 (rs6000_legitimate_address_p): Ditto.
5949 (abi_v4_pass_in_fpr): Ditto.
5950 (setup_incoming_varargs): Ditto.
5951 (rs6000_gimplify_va_arg): Ditto.
5952 (rs6000_split_multireg_move): Ditto.
5953 (rs6000_savres_strategy): Ditto.
5954 (rs6000_emit_prologue_components): Ditto.
5955 (rs6000_emit_epilogue_components): Ditto.
5956 (rs6000_emit_prologue): Ditto.
5957 (rs6000_emit_epilogue): Ditto.
5958 (rs6000_elf_file_end): Ditto.
5959 (rs6000_function_value): Ditto.
5960 (rs6000_libcall_value): Ditto.
5961 * config/rs6000/rs6000.h: Ditto.
5962 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5963 (TARGET_MINMAX): ... this. New.
5964 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5965 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5966 (*movsi_internal1_single): Delete.
5967 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5968 mfpu=, mxilinx-fpu): Delete.
5969 * config/rs6000/singlefp.h: Delete.
5970 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5971 * config/rs6000/t-rs6000: Ditto.
5972 * config/rs6000/t-xilinx: Delete.
5973 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5974 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5975 (VStype_simple): Delete.
5976 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5977 * config/rs6000/xfpu.h: Delete.
5978 * config/rs6000/xfpu.md: Delete.
5979 * config/rs6000/xilinx.h: Delete.
5980 * config/rs6000/xilinx.opt: Delete.
5981 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5982 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5984 2018-05-04 Tom de Vries <tom@codesourcery.com>
5987 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5990 2018-05-04 Richard Biener <rguenther@suse.de>
5993 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5994 (expand_complex_div_wide): Likewise.
5995 (expand_complex_operations_1): Likewise.
5996 (expand_complex_libcall): Preserve EH info of the original stmt.
5997 (tree_lower_complex): Handle removed blocks.
5998 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5999 on complex multiplication and division libcall builtins.
6001 2018-05-04 Richard Biener <rguenther@suse.de>
6004 * fold-const.c (negate_expr_p): Restrict negation of operand
6005 zero of a division to when we know that can happen without
6007 (fold_negate_expr_1): Likewise.
6009 2018-05-04 Jakub Jelinek <jakub@redhat.com>
6012 * real.h (real_nextafter): Declare.
6013 * real.c (real_nextafter): New function.
6014 * fold-const-call.c (fold_const_nextafter): New function.
6015 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
6016 CASE_CFN_NEXTTOWARD.
6017 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
6018 even when arg1_mode is different from arg0_mode.
6020 2018-05-03 Nathan Sidwell <nathan@acm.org>
6022 * doc/extend.texi (Deprecated Features): Remove
6024 (Backwards Compatibility): Likewise.
6025 * doc/invoke.texi (C++ Language Options): Likewise.
6026 (C++ Dialect Options): Likewise.
6028 2018-05-03 Jakub Jelinek <jakub@redhat.com>
6031 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
6032 _mm512_mask_mullox_epi64): New intrinsics.
6034 2018-05-03 Tom de Vries <tom@codesourcery.com>
6037 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6038 dump files): Add offload-tree.
6040 2018-05-03 Richard Biener <rguenther@suse.de>
6042 PR tree-optimization/85615
6043 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
6044 to loops not nested in BBs loop father to avoid creating multi-entry
6047 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6049 PR tree-optimization/70291
6050 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
6051 arguments. Change return type to tree. Emit libcall as a new
6052 statement rather than replacing existing one when inplace_p is true.
6053 (expand_complex_multiplication_components): New function.
6054 (expand_complex_multiplication): Expand floating-point complex
6055 multiplication using the above.
6056 (expand_complex_division): Rename inner_type parameter to type.
6057 Update expand_complex_libcall call-site.
6058 (expand_complex_operations_1): Update expand_complex_multiplication
6059 and expand_complex_division call-sites.
6061 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6064 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6065 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6066 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
6067 the highest significant bit of the shift count mask is clear. In
6068 check whether and[sq]i3 is needed verify that all significant bits
6069 of the shift count other than the highest are set.
6071 2018-05-02 Tom de Vries <tom@codesourcery.com>
6074 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
6075 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
6076 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
6077 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
6078 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
6079 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
6080 __builtin_goacc_parlevel_size.
6082 2018-05-02 Richard Biener <rguenther@suse.de>
6084 PR tree-optimization/85597
6085 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
6086 do not use split vect_get_vec_defs call but call vect_get_slp_defs
6089 2018-05-02 Tom de Vries <tom@codesourcery.com>
6092 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6093 dump files): Add ltrans-tree.
6095 2018-05-02 Tom de Vries <tom@codesourcery.com>
6098 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6099 dump files): Add wpa-ipa.
6101 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
6103 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
6104 powerpc*-*-linux*paired* target.
6105 * config/rs6000/750cl.h: Delete.
6106 * config/rs6000/paired.h: Delete.
6107 * config/rs6000/paired.md: Delete.
6108 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
6110 * config/rs6000/rs6000-builtin.def: Remove paired float support.
6111 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
6112 comment. Remove paired float support.
6113 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
6114 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6116 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
6117 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
6119 * config/rs6000/rs6000.c: Remove paired float support.
6120 (paired_expand_vector_init, paired_expand_vector_move,
6121 paired_emit_vector_compare, paired_emit_vector_cond_expr,
6122 (paired_expand_lv_builtin, paired_expand_stv_builtin,
6123 paired_expand_builtin, paired_expand_predicate_builtin,
6124 paired_init_builtins): Delete.
6125 * config/rs6000/rs6000.h: Remove paired float support.
6126 * config/rs6000/rs6000.md: Remove paired float support.
6127 (move_from_CR_ov_bit): Delete.
6128 * config/rs6000/rs6000.opt (mpaired): Delete.
6129 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
6130 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
6132 2018-05-02 Richard Biener <rguenther@suse.de>
6135 * gimplify.c (gimplify_save_expr): When in SSA form allow
6136 SAVE_EXPRs to compute to SSA vars.
6138 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6141 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6142 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6143 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
6144 clobber operands[2], instead use a new pseudo. Formatting fixes.
6146 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
6148 PR tree-optimization/85586
6149 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
6150 exit early for statements in the same group if the accesses are
6153 2018-05-02 Tom de Vries <tom@codesourcery.com>
6156 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
6159 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
6161 PR tree-optimization/85143
6162 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
6164 2018-05-01 Tom de Vries <tom@codesourcery.com>
6167 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
6168 not found" error message.
6170 2018-05-01 Tom de Vries <tom@codesourcery.com>
6173 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
6174 * vec.c (test_ordered_remove_if): New function.
6175 (vec_c_tests): Call test_ordered_remove_if.
6176 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
6177 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
6178 * tree-vect-patterns.c (vect_pattern_recog_1): Use
6179 VEC_ORDERED_REMOVE_IF.
6181 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6183 PR tree-optimization/82665
6184 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
6185 pointer subtraction where arguments come from a memchr call.
6187 2018-05-01 Jakub Jelinek <jakub@redhat.com>
6189 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
6190 --push-state --as-needed and --pop-state instead of --as-needed and
6191 --no-as-needed if ld supports it.
6192 * configure: Regenerated.
6195 * doc/install.texi2html: Replace _002d with - and _002a with * in
6196 generated html files using sed.
6198 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6201 * gcc-rich-location.c (blank_line_before_p): New function.
6202 (use_new_line): New function.
6203 (gcc_rich_location::add_fixit_insert_formatted): New function.
6204 * gcc-rich-location.h
6205 (gcc_rich_location::add_fixit_insert_formatted): New function.
6207 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6209 * selftest.c (assert_streq): Rename "expected" and "actual" to
6210 "val1" and "val2". Extend NULL-handling to cover both inputs
6211 symmetrically, while still requiring both to be non-NULL for a pass.
6212 * selftest.h (assert_streq): Rename "expected" and "actual" to
6214 (ASSERT_EQ): Likewise.
6215 (ASSERT_EQ_AT): Likewise.
6216 (ASSERT_KNOWN_EQ): Likewise.
6217 (ASSERT_KNOWN_EQ_AT): Likewise.
6218 (ASSERT_NE): Likewise.
6219 (ASSERT_MAYBE_NE): Likewise.
6220 (ASSERT_MAYBE_NE_AT): Likewise.
6221 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
6222 the assertion to pass.
6223 (ASSERT_STREQ_AT): Likewise.
6225 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6227 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
6228 interaction with -pie.
6230 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6232 * selftest.h: Fix alphabetization of per-source-file selftest
6235 2018-04-30 Jason Merrill <jason@redhat.com>
6237 PR c++/61982 - dead stores to destroyed objects.
6238 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
6241 2018-04-30 Jason Merrill <jason@redhat.com>
6243 * tree.c (build_clobber): New.
6244 * tree.h: Declare it.
6245 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
6247 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6249 * diagnostic-show-locus.c (layout::layout): Update for
6250 location_get_source_line returning a char_span.
6251 (struct char_span): Move to input.h.
6252 (struct correction): Update for fields in char_span becoming
6254 (struct source_line): Update for location_get_source_line
6255 returning a char_span.
6256 (layout::print_line): Likewise.
6257 * edit-context.c (edited_file::print_content): Likewise.
6258 (edited_file::print_diff_hunk): Likewise.
6259 (edited_file::print_run_of_changed_lines): Likewise.
6260 (edited_file::get_num_lines): Likewise.
6261 (edited_line::edited_line): Likewise.
6262 * final.c (asm_show_source): Likewise.
6263 * input.c (location_get_source_line): Convert return type
6264 from const char * to char_span, losing the final "line_len"
6266 (dump_location_info): Update for the above.
6267 (get_substring_ranges_for_loc): Likewise. Use a char_span
6268 when handling the literal within the line.
6269 (test_reading_source_line): Update for location_get_source_line
6270 returning a char_span.
6271 * input.h (class char_span): Move here from
6272 diagnostic-show-locus.c, converting from a struct to a class.
6273 Make data members private.
6274 (char_span::operator bool): New.
6275 (char_span::length): New.
6276 (char_span::get_buffer): New.
6277 (char_span::operator[]): New.
6278 (char_span::subspan): Make const.
6279 (char_span::xstrdup): New.
6280 (location_get_source_line): Convert return type from const char *
6281 to char_span, losing the final "line_size" param.
6283 2018-04-30 Jan Hubicka <jh@suse.cz>
6285 * lto-wrapper.c (ltrans_priorities): New static var.
6286 (cmp_priority): New.
6287 (run_gcc): Read priorities and if doing parallel build order
6288 the Makefile by them.
6290 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6292 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
6294 2018-04-30 Richard Biener <rguenther@suse.de>
6296 * tree-cfg.c (verify_address): Remove base argument, add
6297 flag whether to check TREE_ADDRESSABLE and do that.
6298 (verify_expr): Remove.
6299 (verify_types_in_gimple_reference): Add pieces from verify_expr.
6300 (verify_gimple_assign_single): Likewise.
6301 (verify_gimple_switch): Likewise.
6302 (verify_expr_location_1): Dereference tp once. Add (disabled)
6303 piece from verify_expr.
6304 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
6306 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6308 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
6310 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6312 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
6313 (small_data_pattern): Likewise.
6314 (arc_rewrite_small_data): Likewise.
6315 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
6316 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
6317 (get_symbol_alignment): New function.
6318 (legitimate_small_data_address_p): Likewise.
6319 (legitimate_scaled_address): Update, call
6320 legitimate_small_data_address_p.
6321 (output_sdata): New static variable.
6322 (arc_print_operand): Update how we handle small data operands.
6323 (arc_print_operand_address): Likewise.
6324 (arc_legitimate_address_p): Update, use
6325 legitimate_small_data_address_p.
6326 (arc_rewrite_small_data_p): Remove.
6327 (arc_rewrite_small_data_1): Likewise.
6328 (arc_rewrite_small_data): Likewise.
6329 (small_data_pattern): Likewise.
6330 (compact_sda_memory_operand): Update to use
6331 legitimate_small_data_address_p and get_symbol_alignment.
6332 (prepare_move_operands): Don't rewite sdata pattern.
6333 (prepare_extend_operands): Remove.
6334 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
6336 (zero_extendqisi2): Likewise.
6337 (zero_extendhisi2): Likewise.
6338 (extendqihi2): Likewise.
6339 (extendqisi2): Likewise.
6340 (extendhisi2): Likewise.
6344 * config/arc/constraints.md (Usd): Change it to memory constraint.
6346 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6348 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
6349 as source of std instructions.
6350 * config/arc/arc.md (movsi_insn): Update pattern predicate to
6351 allow 6-bit constants as source for store instructions.
6352 (movdi_insn): Update instruction pattern to allow 6-bit constants
6353 as source for store instructions.
6355 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6357 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
6359 2018-04-30 Nathan Sidwell <nathan@acm.org>
6360 Sandra Loosemore <sandra@codesourcery.com>
6362 * dumpfile.c (dump_open): Allow '-' for stdout.
6363 * doc/invoke.texi (Developer Options): Document dump filename
6364 determination early. Document stdin/stdout selection.
6366 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
6368 Microblaze Target: PIC data text relative
6370 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
6371 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
6373 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
6374 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
6375 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
6376 New addressing mode for data-text relative position indepenedent code.
6377 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
6378 'ADDRESS_SYMBOLIC_TXT_REL'.
6379 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
6380 (microblaze_legitimate_pic_operand): Exclude function calls from
6381 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
6382 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
6384 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6385 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6386 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
6387 for 'address + offset'.
6388 (microblaze_expand_prologue): Add new function prologue call for
6390 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
6391 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
6392 table in case of TARGET_PIC_DATA_TEXT_REL.
6393 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
6394 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
6395 Add new macros 'UNSPEC_TEXT',
6396 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
6397 + exclude function calls from 'UNSPEC_PLT' in case of data text
6399 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6400 new target hook for generating address diff vector tables in case of
6402 * doc/tm.texi : Regenerate.
6403 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
6404 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
6405 of addr diff vector generation.
6406 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6407 target hook definition.
6408 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
6409 Add default function for generate_pic_addr_diff_vec -> flag_pic.
6410 * doc/invoke.texi (Add new pic option): Add new microblaze pic
6411 option for data text relative.
6413 2018-04-30 Richard Biener <rguenther@suse.de>
6415 * tree-chrec.h (evolution_function_is_constant_p): Remove
6417 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
6419 2018-04-30 Richard Biener <rguenther@suse.de>
6422 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
6424 2018-04-30 Richard Biener <rguenther@suse.de>
6426 PR tree-optimization/28364
6427 PR tree-optimization/85275
6428 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
6429 copying first exit test.
6431 2018-04-28 Mark Wielaard <mark@klomp.org>
6433 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
6435 (dwarf_AT): Handle DW_AT_addr_base.
6436 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
6438 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
6441 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
6442 (*ashl<dwi>3_doubleword_mask_1): Ditto.
6443 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
6444 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
6446 2018-04-28 Richard Biener <rguenther@suse.de>
6448 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
6449 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
6450 to reflect use. Only add interesting stmts.
6452 2018-04-27 Martin Jambor <mjambor@suse.cz>
6455 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
6456 the jump function allows for passing through aggregate values.
6458 2018-04-27 David Malcolm <dmalcolm@redhat.com>
6460 * input.h (in_system_header_at): Convert from macro to inline
6462 (from_macro_expansion_at): Likewise.
6463 (from_macro_definition_at): Likewise.
6465 2018-04-27 Jeff Law <law@redhat.com>
6467 * config.gcc: Mark tile* targets as deprecated/obsolete.
6469 2018-04-27 Richard Biener <rguenther@suse.de>
6471 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
6474 2018-04-27 Richard Biener <rguenther@suse.de>
6476 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
6478 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
6480 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
6481 with Yd constraint. Set "preferred_for_speed" attribute from
6482 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
6484 (*movdi_internal): Ditto.
6485 (movti_interunit splitters): Remove
6486 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
6487 (movdi_interunit splitters): Ditto.
6488 * config/i386/constraints.md (Ye): Remove.
6489 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
6491 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6494 * config/aarch64/constraints.md (Usg): Limit to 31.
6497 2018-04-27 Jakub Jelinek <jakub@redhat.com>
6499 PR tree-optimization/85529
6500 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
6501 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
6502 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
6503 zero extension or masking of the MSB bit.
6504 (optimize_range_tests): Add FIRST_BB argument, pass it through
6505 to optimize_range_tests_var_bound.
6506 (maybe_optimize_range_tests, reassociate_bb): Adjust
6507 optimize_range_tests callers.
6509 2018-04-26 Richard Biener <rguenther@suse.de>
6510 Jakub Jelinek <jakub@redhat.com>
6512 * cgraph.h (symbol_table): Just declare debug method here.
6513 * symtab.c (symbol_table::debug): Define.
6515 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
6517 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
6519 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6521 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6522 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6523 (*movdi_internal): Substitute Yi and Yj constraint with x
6524 and Ym and Yn constraint with y constraint. Update "isa"
6525 attribute and set "preferred_for_speed" attribute from
6526 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6527 (*movsi_internal): Ditto.
6528 (*movdf_internal): Ditto.
6529 (*movsf_internal): Ditto.
6530 (*zero_extendsidi2): Ditto.
6531 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6532 (sse2_loadld): Ditto.
6533 (*vec_extract<ssevecmodelower>_0): Ditto.
6534 (*vec_extractv4si_0_zext_sse4): Ditto.
6535 (vec_concatv2di): Ditto.
6536 (*vec_dup<mode>): Ditto.
6537 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
6538 * config/i386/constraints.md (Yi): Remove.
6543 2018-04-26 Nathan Sidwell <nathan@acm.org>
6545 * dumpfile.c (dump_open): New.
6546 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
6547 (dump_finish): Detect stdio/stderr by value not name.
6549 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6551 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6553 2018-04-26 Tom de Vries <tom@codesourcery.com>
6556 * config/nvptx/nvptx.c (verify_neutering_jumps)
6557 (verify_neutering_labels): New function
6558 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6560 2018-04-26 Tom de Vries <tom@codesourcery.com>
6563 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6564 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6567 2018-04-26 Richard Biener <rguenther@suse.de>
6568 Tom de Vries <tom@codesourcery.com>
6571 * lto-streamer-out.c (output_function): Fixup loops if required to match
6572 discovery done in the reader.
6574 2018-04-26 Richard Biener <rguenther@suse.de>
6576 PR tree-optimization/85116
6577 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6578 have a loop exit from the single latch predecessor. Remove
6579 case of header with just condition.
6580 (ch_base::copy_headers): Exclude infinite loops from any
6582 (pass_ch::execute): Record exits.
6584 2018-04-26 Richard Biener <rguenther@suse.de>
6586 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6587 prologue cost vector and pass it to vect_get_load_cost.
6588 (vect_get_peeling_costs_all_drs): Likewise.
6589 (vect_peeling_hash_get_lowest_cost): Likewise.
6590 (vect_enhance_data_refs_alignment): Likewise.
6592 2018-04-26 Richard Biener <rguenther@suse.de>
6595 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6596 checking of integer<->pointer conversions.
6597 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6598 sign-/zero-extending pointer types.
6599 (expand_omp_for_static_chunk): Likewise.
6601 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6602 Jean Lee <xiaoyur347@gmail.com>
6604 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6605 (TARGET_ASAN_SHADOW_OFFSET): Define.
6606 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6607 true for -fsanitize=address.
6609 2018-04-25 Mark Wielaard <mark@klomp.org>
6611 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6614 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6616 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6617 than "alu", remove explicit "memory" and "imm_disp" attributes.
6618 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6621 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6622 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6623 gen_lowpart_no_emit.
6625 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6628 * config/i386/i386.c (ix86_expand_builtin): Change memory
6629 operand to XI, extend p0 to Pmode.
6630 * config/i386/i386.md: Change unspec volatile and operand
6631 1 mode to XI, change operand 0 mode to P.
6633 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6635 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6636 GET_MODE_MASK before any checking.
6637 (nds32_can_use_bset_p): Likewise.
6638 (nds32_can_use_btgl_p): Likewise.
6640 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6642 * config/nds32/nds32-doubleword.md: New define_split pattern for
6643 illegal register number.
6645 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6647 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6649 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6651 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6653 2018-04-25 Richard Biener <rguenther@suse.de>
6655 * lto-streamer.h (LTO_major_version): Bump to 8.
6657 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6659 * BASE-VER: Set to 9.0.0.
6661 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6663 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6664 in __abskf2 and __powikf2.
6666 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6669 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6670 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6671 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6672 Use the above on operand 2. Reindent.
6673 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6675 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6678 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6680 * config/i386/i386.opt (mcet): Removed.
6681 * doc/install.texi: Remove -mcet documentation.
6682 * doc/invoke.texi: Likewise.
6684 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6687 * doc/install.texi: Remove -mcet from bootstrap-cet.
6689 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6692 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6693 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6697 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6698 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6699 containing a CONST_VECTOR.
6701 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6703 * doc/install.texi: Update newlib dependency for nvptx.
6705 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6708 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6709 instead of INTVAL when shifting x left.
6711 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6713 PR tree-optimization/85478
6714 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6715 vect_grouped_store_supported for single element vectors.
6717 2018-04-24 Richard Biener <rguenther@suse.de>
6720 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6721 load cost increase to the case of non-constant step.
6723 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6726 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6727 destination if any_malformed_asm.
6729 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6732 * expr.c (store_field): In the bitfield case, if the value comes from
6733 a function call and is returned in registers by means of a PARALLEL,
6734 do not change the mode of the temporary unless BLKmode and VOIDmode.
6736 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6738 PR rtl-optimization/85423
6739 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6740 dependencies to debug insns when the previous insn is non-debug.
6742 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6744 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6745 enums into a single definition.
6746 (fls): Fix predicates and printing.
6749 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6751 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6752 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6753 and short u6 immediate.
6754 (check_if_valid_sleep_operand): Remove.
6755 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6757 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6759 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6760 flag_always_save_lp condition.
6761 * config/nds32/nds32.opt (malways-save-lp): New option.
6763 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6765 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6766 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6767 * config/nds32/nds32.h
6768 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6769 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6771 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6773 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6774 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6776 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6777 Chung-Ju Wu <jasonwucj@gmail.com>
6779 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6781 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6782 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6784 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6786 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6788 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6790 * config/nds32/nds32-protos.h (nds32_data_alignment,
6791 nds32_local_alignment): Declare.
6792 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6793 nds32_local_alignment): New functions.
6794 (TARGET_CONSTANT_ALIGNMENT): Define.
6795 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6797 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6799 * config/nds32/nds32.c
6800 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6801 (TARGET_MODES_TIEABLE_P): Likewise.
6803 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6805 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6808 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6809 Chung-Ju Wu <jasonwucj@gmail.com>
6811 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6812 for unaligned access.
6813 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6815 * config/nds32/nds32-intrinsic.md: Likewise.
6816 * config/nds32/nds32_intrinsic.h: Likewise.
6817 * config/nds32/nds32.h (nds32_builtins): Likewise.
6818 * config/nds32/nds32.opt (munaligned-access): New option.
6819 * config/nds32/nds32.c (nds32_asm_file_start): Display
6820 flag_unaligned_access status.
6822 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6824 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6825 -mno-relax is present.
6826 * config/riscv/linux.h (LINK_SPEC): Ditto.
6828 2018-04-20 Martin Sebor <msebor@redhat.com>
6831 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6833 (gimple_fold_builtin_stxcpy_chk): Same.
6834 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6836 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6839 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6840 __powikf2 when long double is IEEE 128-bit.
6842 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6844 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6845 step to make sure stack always aligned.
6847 2018-04-20 Carl Love <cel@us.ibm.com>
6850 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6851 size check for arg0.
6853 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6854 Tom de Vries <tom@codesourcery.com>
6857 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6858 Emit insns for calls too.
6859 (nvptx_find_par): Always look for worker-level predecessor insn.
6860 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6862 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6863 (nvptx_process_pars): Propagate frames for calls.
6865 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6868 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6870 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6871 (ix86_handle_option): Don't handle OPT_mibt.
6872 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6874 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6876 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6877 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6878 (ix86_target_macros): Define __CET__ with flag_cf_protection
6879 for -fcf-protection.
6880 * config/i386/i386.c (isa2_opts): Remove -mibt.
6881 * config/i386/i386.h (TARGET_IBT): Removed.
6882 (TARGET_IBT_P): Likewise.
6883 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6884 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6885 * config/i386/i386.opt (mcet): Update help message.
6888 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6889 -mcet as an alias for -mshstk.
6891 2018-04-20 Richard Biener <rguenther@suse.de>
6894 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6895 complexity by forcing a single use of the multiply operand.
6897 2018-04-20 Martin Jambor <mjambor@suse.cz>
6900 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6901 recursion dependency to only apply to non-clones.
6903 2018-04-20 Martin Jambor <mjambor@suse.cz>
6906 * ipa-cp.c (create_specialized_node): Check that clones of
6907 self-recursive edges exist during IPA-CP.
6909 2018-04-19 Toon Moene <toon@moene.org>
6911 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6914 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6916 PR tree-optimization/85467
6917 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6918 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6919 VECTOR_CST element to type.
6921 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6924 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6925 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6926 (builtin_longjmp): Likewise.
6927 (save_stack_nonlocal): New pattern.
6928 (restore_stack_nonlocal): Likewise.
6930 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6933 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6934 Replace ASM_OUTPUT_LABEL with fprintf.
6936 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6939 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6940 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6941 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6942 define __IBT__ and __SHSTK__ for -fcf-protection.
6943 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6945 (ix86_trampoline_init): Likewise.
6946 (x86_output_mi_thunk): Likewise.
6947 (ix86_notrack_prefixed_insn_p): Likewise.
6948 (ix86_option_override_internal): Don't disallow -fcf-protection.
6949 * config/i386/i386.md (rdssp<mode>): Also enable for
6951 (incssp<mode>): Likewise.
6952 (nop_endbr): Likewise.
6953 * config/i386/i386.opt (mcet): Change help message to built-in
6957 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6958 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6959 enable CET built-in functions.
6961 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6963 * common/config/i386/i386-common.c
6964 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6965 OPTION_MASK_ISA_MOVDIRI_UNSET,
6966 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6967 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6968 * config.gcc (movdirintrin.h): New header.
6969 * config/i386/cpuid.h (bit_MOVDIRI,
6970 bit_MOVDIR64B): New bits.
6971 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6973 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6974 (VOID, PVOID, PCVOID)): New function types.
6975 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6976 __builtin_ia32_directstoreu_u64,
6977 __builtin_ia32_movdir64b): New builtins.
6978 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6979 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6981 (ix86_valid_target_attribute_inner_p): Ditto.
6982 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6983 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6984 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6985 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6986 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6987 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6988 (movdiri<mode>, movdir64b_<mode>): New.
6989 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6990 * config/i386/immintrin.h: Include movdirintrin.h.
6991 * config/i386/movdirintrin.h: New file.
6992 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6994 2018-04-19 Richard Biener <rguenther@suse.de>
6997 * cfg.c (clear_bb_flags): When loop state says we have
6998 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
7000 2018-04-19 Richard Biener <rguenther@suse.de>
7002 PR tree-optimization/84737
7003 * tree-vect-data-refs.c (vect_copy_ref_info): New function
7004 copying restrict info.
7005 (vect_setup_realignment): Use it.
7006 * tree-vectorizer.h (vect_copy_ref_info): Declare.
7007 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
7008 the first DR to all generated stores.
7009 (vectorizable_load): Likewise for loads.
7011 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7013 PR tree-optimization/85446
7014 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
7015 the integral and pointer types to have the same precision.
7017 * doc/install.texi: Document --disable-cet being the default and
7020 2018-04-18 Martin Liska <mliska@suse.cz>
7022 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
7025 2018-04-18 Martin Liska <mliska@suse.cz>
7028 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7031 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7032 arguments if they are comparable.
7034 2018-04-18 Martin Liska <mliska@suse.cz>
7037 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7040 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7043 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
7046 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
7047 ENDBR after calling __morestack.
7049 2018-04-18 David Malcolm <dmalcolm@redhat.com>
7052 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
7053 by using gcc_base_ver to generate a gcc_driver_version, and use
7054 it when generating GCC_DRIVER_NAME.
7055 * configure: Regenerate.
7057 2018-04-18 Jakub Jelinek <jakub@redhat.com>
7060 * config.gcc: Obsolete powerpc*-*-*spe*.
7062 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7065 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
7066 (stabstr_D): Change type of unum from unsigned int to
7067 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
7070 2018-04-17 Jim Wilson <jimw@sifive.com>
7073 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
7074 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
7075 Set arg_pointer_offset after using pretend_args_size.
7077 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7079 PR rtl-optimization/85431
7080 * dse.c (record_store): Ignore zero width stores.
7083 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
7084 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
7085 __builtin_stack_restore rather than after it.
7086 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
7087 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
7088 argument instead of virtual_dynamic_stack_rtx.
7090 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
7092 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
7094 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
7095 Add note to error message to explain internal mapping of overloaded
7096 built-in function name to non-overloaded built-in function name.
7097 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
7100 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7103 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
7104 where the inputs overlap with the output.
7106 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7108 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
7109 (=v, v) alternative and explicit "memory" attribute.
7110 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
7111 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7113 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
7114 "sselog1" type instead of "sselog".
7115 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
7116 "sselog". Remove explicit "memory" attribute.
7117 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
7118 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7120 (vec_extract_hi_v32hi): Merge all alternatives into one, use
7121 "sselog1" type instead of "sselog". Remove explicit "memory"
7123 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
7124 use "sselog1" type instead of "sselog". Remove explicit "memory"
7126 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
7127 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7129 (vec_extract_hi_v64qi): Merge all alternatives into one, use
7130 "sselog1" type instead of "sselog". Remove explicit "memory"
7132 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
7133 use "sselog1" type instead of "sselog". Remove explicit "memory"
7137 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
7140 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
7143 2018-04-17 Martin Jambor <mjambor@suse.cz>
7146 * ipa-cp.c (create_specialized_node): Call
7147 expand_all_artificial_thunks if necessary.
7149 2018-04-17 Martin Liska <mliska@suse.cz>
7152 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
7153 in message, remote space in between '_G' and '('.
7155 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7158 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
7159 avx512f_vmcmp<mode>3<round_saeonly_name>,
7160 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
7161 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
7162 avx512f_rndscale<mode><round_saeonly_name>,
7163 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
7164 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7165 Use %<iptr>2 instead of %2 for -masm=intel.
7166 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
7167 avx512f_vcvttss2usi<round_saeonly_name>,
7168 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
7170 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
7171 avx512f_vcvttsd2usi<round_saeonly_name>,
7172 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
7173 Use %q1 instead of %1 for -masm=intel.
7174 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
7175 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
7176 of %3 for -masm=intel.
7177 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
7179 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
7181 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
7183 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
7184 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
7186 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
7188 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
7189 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
7190 %g1 and one with %0 and %1.
7191 (avx512er_vmrcp28<mode><round_saeonly_name>,
7192 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
7194 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
7195 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
7196 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
7197 of %0 and %{%4%} for -masm=intel.
7198 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
7199 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
7200 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
7201 order of %0 and %{%5%}%{z%} for -masm=intel.
7203 2018-04-17 Jan Hubicka <jh@suse.cz>
7206 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
7208 2018-04-17 Martin Liska <mliska@suse.cz>
7211 * multiple_target.c (create_dispatcher_calls): Set apostrophes
7212 for target_clone error message. Make default implementation
7213 clone to be a local declaration.
7214 (separate_attrs): Add new argument and check for an empty
7216 (expand_target_clones): Handle it.
7217 (ipa_target_clone): Make redirection just for target_clones
7220 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
7221 Tom de Vries <tom@codesourcery.com>
7224 * omp-expand.c (expand_oacc_for): Add dummy false branch for
7225 tiled basic blocks without omp continue statements.
7227 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
7230 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
7231 vec_extract expression as having side effects to make sure it gets
7234 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
7237 * config/i386/i386.c (get_builtin_code_for_version): Check
7240 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
7243 * gcc/config.gcc: Support "skylake".
7244 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
7246 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
7247 (processor_target_table): Add "skylake".
7248 (ix86_option_override_internal): Add "skylake".
7249 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
7250 PROCESSOR_CANNONLAKE.
7251 (get_builtin_code_for_version): Fix priority for
7252 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
7253 PROCESSOR_SKYLAKE-AVX512.
7254 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
7255 (processor_type): Add PROCESSOR_SKYLAKE.
7257 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
7258 Jason Merrill <jason@redhat.com>
7261 * convert.c (convert_to_integer_1): Use direct recursion for
7262 enumeral types and types with a precision less than the number
7263 of bits in their mode.
7265 2018-04-16 Julia Koval <julia.koval@intel.com>
7268 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
7269 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
7271 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
7274 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
7275 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
7276 and -mno-direct-move.
7278 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
7281 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
7282 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
7283 Ensure negative shifts result in {0}.
7285 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
7287 PR rtl-optimization/79916
7288 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
7289 regs (if any) to define how to gnerate SD moves when LRA is in
7292 2018-04-13 Jakub Jelinek <jakub@redhat.com>
7294 PR rtl-optimization/85393
7295 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
7296 * except.c (expand_dw2_landing_pad_for_region): Make static.
7297 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
7298 a label and unconditional jump to old_bb, rather than
7299 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
7302 PR rtl-optimization/85376
7303 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
7304 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
7305 instead of a specific value.
7307 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7308 Bin Cheng <bin.cheng@arm.com>
7310 PR tree-optimization/82965
7311 PR tree-optimization/83991
7312 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
7313 by_profile_only parameter.
7314 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
7315 information if the loop was predicted to iterate too many times.
7316 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
7318 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7321 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
7324 2018-04-13 Martin Liska <mliska@suse.cz>
7325 Jakub Jelinek <jakub@redhat.com>
7328 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
7329 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
7330 * builtins.c (expand_builtin_memory_copy_args): Use
7331 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
7332 handle dest_addr == pc_rtx.
7334 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
7337 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
7338 asked to not generate direct moves.
7339 (fix_trunc<mode>si2_stfiwx): Similar.
7340 (fix_trunc<mode>si2_internal): Similar.
7342 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7345 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
7346 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
7347 lookup if dest in some wider mode is known to be const0_rtx and
7348 if so, record permanent equivalence for it to be ZERO_EXTEND of
7349 the narrower mode destination.
7351 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7353 * lto-streamer-out.c (output_function): Revert 259346.
7354 * omp-expand.c (expand_oacc_for): Likewise.
7356 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
7358 PR rtl-optimization/85354
7359 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
7360 * sel-sched.c (sel_global_init): ... here.
7362 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
7365 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
7366 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7367 mode for PE-COFF targets.
7368 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
7369 (i386_pe_asm_lto_end): Likewise.
7370 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
7371 (TARGET_ASM_LTO_END): Likewise.
7372 * config/i386/winnt.c (saved_debug_info_level): New static variable.
7373 (i386_pe_asm_lto_start): New function.
7374 (i386_pe_asm_lto_end): Likewise.
7376 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7377 Richard Biener <rguenther@suse.de>
7380 * lto-streamer-out.c (output_function): Fix CFG loop state before
7382 * omp-expand.c (expand_oacc_for): Handle calls to internal
7383 functions like regular functions.
7385 2018-04-12 Richard Biener <rguenther@suse.de>
7388 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
7389 for the early LTO debug to properly generate references to it
7390 during DIE emission. Do not re-use that for the skeleton for
7392 (dwarf2out_early_finish): Likewise.
7394 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7397 * config/i386/sse.md
7398 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
7399 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
7400 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
7401 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
7402 and output is a reg, avoid creating invalid lowpart subreg, but
7403 instead split into a 512-bit move. Don't split if not AVX512VL,
7404 input is xmm16+ reg and output is a mem.
7405 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
7406 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
7407 xmm16+ reg and output is a mem.
7409 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7411 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
7412 also for flag_dwarf2_cfi_asm.
7414 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7416 PR rtl-optimization/85342
7417 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
7418 a bool scalar var inside of the loop instead. Don't try to update
7419 recog_data.operand after failed apply_change_group.
7421 2018-04-12 Tom de Vries <tom@codesourcery.com>
7424 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
7425 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
7426 array with flexible array member as array without given dimension.
7427 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
7428 argument for undefined param to true.
7430 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
7433 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7434 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
7435 from PowerPC section.
7436 * config/rs6000/sysv4.opt (mcall-): Improve help text.
7437 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
7438 help text that is too long.
7439 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
7440 help text that is too long.
7441 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
7442 help text that is too long.
7444 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
7446 * config/alpha/alpha.md (stack_probe_internal): Rename
7447 from "probe_stack". Update all callers.
7449 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7451 PR rtl-optimization/84566
7452 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
7453 sched_macro_fuse_insns.
7455 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7458 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
7459 (compute_block_dependences): ... from here.
7461 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7463 PR tree-optimization/85331
7464 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
7465 from int to HOST_WIDE_INT.
7467 2018-04-11 Martin Jambor <mjambor@suse.cz>
7470 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
7471 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
7472 not the same as the source val.
7473 (cgraph_edge_brings_value_p): New parameter.
7474 (gather_edges_for_value): Pass destination value to
7475 cgraph_edge_brings_value_p.
7476 (perhaps_add_new_callers): Likewise.
7477 (get_info_about_necessary_edges): Likewise and exclude values brought
7478 only by self-recursive edges.
7479 (create_specialized_node): Redirect only clones of self-calling edges.
7480 (+self_recursive_pass_through_p): New function.
7481 (find_more_scalar_values_for_callers_subset): Use it.
7482 (find_aggregate_values_for_callers_subset): Likewise.
7483 (known_aggs_to_agg_replacement_list): Removed.
7484 (decide_whether_version_node): Re-calculate known constants for all
7485 remaining context clones.
7487 2018-04-11 Richard Biener <rguenther@suse.de>
7490 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
7491 from early DWARF output.
7492 (dwarf2out_early_finish): Output line info unconditionally into
7493 early DWARF and add reference to it.
7495 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7498 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
7499 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
7500 other than V2DFmode using iptr mode attribute.
7501 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
7503 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7505 PR rtl-optimization/84659
7506 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
7508 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7511 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
7513 (output_loc_list): Pass address of a dummy size variable even in the
7514 locview handling loop.
7515 (index_location_lists): Add comment on why skip_loc_list_entry can't
7518 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7521 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7524 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7527 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7528 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7529 and -mstring-compare-inline-limit.
7531 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7534 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7535 for stack clash protection in a register whenever we need it to be in
7538 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7540 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
7541 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
7543 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7546 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
7548 (mlong-double-): Ditto.
7549 * config/rs6000/sysv4.opt (msdata=): Ditto.
7550 (mtls-size=): Ditto.
7552 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7554 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7555 erroneous entries for
7556 "vector int vec_ldl (int, long int *)", and
7557 "vector unsigned int vec_ldl (int, unsigned long int *)".
7558 Add comments and entries for
7559 "vector bool char vec_ldl (int, bool char *)",
7560 "vector bool short vec_ldl (int, bool short *)",
7561 "vector bool int vec_ldl (int, bool int *)",
7562 "vector bool long long vec_ldl (int, bool long long *)",
7563 "vector pixel vec_ldl (int, pixel *)",
7564 "vector long long vec_ldl (int, long long *)",
7565 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7566 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7567 type tree bool_long_long_type_node and correct definition of
7568 bool_V2DI_type_node to make reference to this new type tree.
7569 (rs6000_mangle_type): Replace erroneous reference to
7570 bool_long_type_node with bool_long_long_type_node.
7571 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7572 comments to emphasize sign distinctions for char and int types and
7573 replace RS6000_BTI_bool_long constant with
7574 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7575 use of RS6000_BTI_pixel.
7576 (bool_long_type_node): Remove this macro definition.
7577 (bool_long_long_type_node): New macro definition
7579 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7581 PR rtl-optimization/85300
7582 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7583 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7584 simplify_unary_operation fails.
7586 2018-04-10 Martin Liska <mliska@suse.cz>
7588 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7589 cgraph_edge and ipa_ref.
7591 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7595 * config/i386/sse.md
7596 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7597 computation of the VEC_MERGE selector from mask.
7598 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7599 Fix decoding of the VEC_MERGE selector into mask.
7601 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7603 PR tree-optimization/85286
7604 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7606 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7608 * final.c (final_1): Set insn_last_address as well as
7609 insn_current_address.
7611 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7614 * explow.c (emit_stack_probe): Call validize_mem on memory location
7615 before passing it to gen_probe_stack. Create address operand and
7616 legitimize it for the probe_stack_address case.
7618 2018-04-09 Jan Hubicka <jh@suse.cz>
7621 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7622 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7623 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7625 (free_lang_data): Rebuild type inheritance graph.
7627 2018-04-09 Martin Sebor <msebor@redhat.com>
7629 * invoke.texi (-finline-small-functions): Mention other optimization
7631 (-findirect-inlining, -fpartial-inlining): Same.
7632 (-finline-functions-called-once): Same.
7633 (-freorder-blocks-and-partition): Same.
7635 2018-04-09 Jan Hubicka <jh@suse.cz>
7638 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7639 jumps; choose last target that matches the criteria (i.e.
7640 no partition changes for non-crossing jumps).
7641 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7642 support for redirecting crossing jumps to non-crossing.
7644 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7646 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7647 also for naked functions.
7649 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7651 * config/arc/arc.md (add_shift): New pattern.
7652 (add_shift2): Likewise.
7653 (sub_shift): Likewise.
7654 (sub_shift_cmp0_noout): Likewise.
7655 (compare_si_ashiftsi): Likewise.
7656 (xbfu_cmp0_noout): New combine pattern.
7657 (xbfu_cmp0"): Likewise.
7658 (movsi_set_cc_insn): Place the predicable variant first.
7659 (commutative_binary_cmp0_noout): Remove clobber.
7660 (commutative_binary_cmp0): New pattern.
7661 (noncommutative_binary_cmp0): Likewise.
7662 (noncommutative_binary_cmp0_noout): Likewise.
7663 (noncommutative_binary_comparison_result_used): Removed.
7664 (rsub_cmp0): New pattern.
7665 (rsub_cmp0_noout): Likewise.
7666 (extzvsi): Changed, keep only meaningful variants.
7667 (SQH, SEZ): New iterators.
7668 (SQH_postfix): New mode attribute.
7669 (SEZ_prefix): New code attribute.
7670 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7671 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7672 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7674 (noncommutative_operator): Check the availability of barrel
7677 2018-04-09 Richard Biener <rguenther@suse.de>
7679 PR tree-optimization/85284
7680 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7681 Only use the niter constraining form of simple_iv when the exit
7684 2018-04-09 Tom de Vries <tom@codesourcery.com>
7687 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7688 (define_expand "*memory_barrier"): New define_expand.
7689 (define_insn "memory_barrier"): New insn.
7691 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7693 PR rtl-optimization/80463
7694 PR rtl-optimization/83972
7695 PR rtl-optimization/83480
7697 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7698 correct producer for the insn.
7699 (tidy_control_flow): Fixup seqnos in case of debug insns.
7701 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7703 PR rtl-optimization/83913
7705 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7706 different sched-times when merging exprs.
7708 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7710 PR rtl-optimization/83962
7712 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7713 tidy_fallthru_edge and tidy_control_flow.
7715 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7717 PR rtl-optimization/83530
7719 * sel-sched.c (force_next_insn): New global variable.
7720 (remove_insn_for_debug): When force_next_insn is true, also leave only
7721 next insn in the ready list.
7722 (sel_sched_region): When the region wasn't scheduled, make another pass
7723 over it with force_next_insn set to 1.
7725 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7727 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7729 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7730 for interrupt control.
7731 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7732 functions for interrupt control.
7733 * config/nds32/nds32-intrinsic.md: Likewise.
7734 * config/nds32/nds32_intrinsic.h: Likewise.
7735 * config/nds32/nds32.h (nds32_builtins): Likewise.
7737 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7739 * config/nds32/nds32.c (nds32_init_machine_status,
7740 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7741 strict_aligned_p field.
7742 (nds32_expand_to_rtl_hook): New function.
7743 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7744 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7746 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7747 Chung-Ju Wu <jasonwucj@gmail.com>
7749 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7750 * config/nds32/nds32-n7.md: New file.
7751 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7752 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7754 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7755 * config/nds32/nds32.md (pipeline_model): Add n7.
7756 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7757 * config/nds32/pipelines.md: Include n7 settings.
7759 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7760 Chung-Ju Wu <jasonwucj@gmail.com>
7762 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7763 * config/nds32/nds32-e8.md: New file.
7764 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7765 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7767 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7768 * config/nds32/nds32.md (pipeline_model): Add e8.
7769 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7770 * config/nds32/pipelines.md: Include e8 settings.
7772 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7773 Chung-Ju Wu <jasonwucj@gmail.com>
7775 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7776 * config/nds32/nds32-n8.md: New file.
7777 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7778 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7780 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7781 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7782 * config/nds32/nds32.md (pipeline_model): Add n8.
7783 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7784 * config/nds32/pipelines.md: Include n8 settings.
7786 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7787 Chung-Ju Wu <jasonwucj@gmail.com>
7789 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7790 * config/nds32/nds32-n9-2r1w.md: New file.
7791 * config/nds32/nds32-n9-3r2w.md: New file.
7792 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7793 nds32_register_ports): New or modify for cpu n9.
7794 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7796 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7797 * config/nds32/nds32-utils.c: New file.
7798 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7799 TARGET_MUL_SLOW): Define.
7800 * config/nds32/nds32.md (pipeline_model): New attribute.
7801 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7802 New options that support cpu n9.
7803 * config/nds32/pipelines.md: Include n9 settings.
7804 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7806 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7808 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7809 information if necessary.
7810 (output_cond_branch_compare_zero): Likewise.
7811 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7812 (nds32_target_alignment): Refine for alignment.
7813 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7814 (FUNCTION_BOUNDARY): Modify.
7815 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7817 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7819 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7821 * config/nds32/constants.md (unspec_volatile_element): Add values for
7822 TLB operation and data prefetch.
7823 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7824 functions for TLB operation and data prefetch.
7825 * config/nds32/nds32-intrinsic.md: Likewise.
7826 * config/nds32/nds32_intrinsic.h: Likewise.
7827 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7828 (nds32_print_operand): Likewise.
7829 * config/nds32/nds32.h (nds32_builtins): Likewise.
7831 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7832 Andrew Pinski <pinsika@gcc.gnu.org>
7835 * match.pd: Use constant_boolean_node of correct type instead of
7836 boolean_true_node or boolean_false_node for simplifying
7837 pointer comparisons to zero.
7839 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7841 PR tree-optimization/80021
7842 * tree.c (verify_type_variant): Make error call in verify_variant_match
7843 translatable and remove final full stop.
7845 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7847 * config/nds32/constants.md (unspec_volatile_element): Add
7848 UNSPEC_VOLATILE_EH_RETURN.
7849 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7850 nds32_output_stack_pop): Support dwarf exception handling process.
7851 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7852 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7853 exception handling process.
7854 (nds32_compute_stack_frame): Likewise.
7855 (nds32_return_addr_rtx): Likewise.
7856 (nds32_initial_elimination_offset): Likewise.
7857 (nds32_expand_prologue): Likewise.
7858 (nds32_expand_epilogue): Likewise.
7859 (nds32_dynamic_chain_address): New function.
7860 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7862 (DYNAMIC_CHAIN_ADDRESS): Define.
7863 (EH_RETURN_DATA_REGNO): Define.
7864 (EH_RETURN_STACKADJ_RTX): Define.
7865 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7866 patterns for dwarf exception handling.
7868 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7870 * config/nds32/nds32.h: Clean up obsolete macros.
7872 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7874 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7875 Add enum values for particular instructions.
7876 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7877 particular intrinsic functions.
7878 * config/nds32/nds32-intrinsic.md: Likewise.
7879 * config/nds32/nds32_intrinsic.h: Likewise.
7880 * config/nds32/nds32.h (nds32_builtins): Likewise.
7881 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7882 (btst, ave): New patterns for particular instructions.
7884 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7886 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7887 Add enum values for atomic load/store and memory sync.
7888 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7890 * config/nds32/nds32-intrinsic.md: Likewise.
7891 * config/nds32/nds32_intrinsic.h: Likewise.
7892 * config/nds32/nds32.h (nds32_builtins): Likewise.
7894 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7896 PR tree-optimization/85257
7897 * fold-const.c (native_encode_vector): If not all elts could fit
7898 and off is -1, return 0 rather than offset.
7899 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7900 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7901 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7902 adjust buffer in native_interpret_expr call.
7904 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7906 * config/nds32/constants.md (unspec_volatile_element): Add cache
7907 control enum values.
7908 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7909 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7910 * config/nds32/nds32.c (nds32_cctl_names): New.
7911 (nds32_print_operand): Handle cache control register names.
7912 * config/nds32/nds32.h (nds32_builtins): New enum values.
7913 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7915 * config/nds32/nds32.md (type): Add mmu.
7916 * config/nds32/pipelines.md (simple_insn): Add mmu.
7918 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7920 * config/nds32/nds32.md (type): Remove call.
7921 * config/nds32/pipelines.md (simple_insn): Likewise.
7923 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7925 * config/nds32/constants.md (unspec_volatile_element): Add
7926 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7927 UNSPEC_VOLATILE_FMFCFG.
7928 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7929 description for fmfcfg and fmfcsr.
7930 (bdesc_1arg): Add fmtcsr.
7931 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7932 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7933 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7934 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7935 unspec_fmfcfg): New patterns.
7936 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7937 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7938 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7939 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7940 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7941 __nds32__fmfcfg): Define.
7943 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7945 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7946 intrinsic register names.
7947 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7948 intrinsic register enum values and macros.
7950 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7952 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7953 for load/store addressing form.
7954 (nds32_print_operand_address): Likewise.
7956 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7959 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7960 based on LABEL_REF. Remove useless assertion.
7961 (pic_address_needs_scratch): Fix formatting.
7962 (sparc_legitimize_pic_address): Minor tweaks.
7963 (sparc_delegitimize_address): Adjust assertion accordingly.
7964 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7965 into symbolic_operand.
7966 (movsi_high_pic_label_ref): Likewise.
7967 (movsi_lo_sum_pic_label_ref): Likewise.
7968 (movdi_pic_label_ref): Likewise.
7969 (movdi_high_pic_label_ref): Likewise.
7970 (movdi_lo_sum_pic_label_ref): Likewise.
7972 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7974 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7975 custom LIB_SPEC setup.
7977 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7978 Kito Cheng <kito.cheng@gmail.com>
7980 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7981 * config/riscv/freebsd.h: New.
7983 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7985 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7986 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7989 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7990 Kito Cheng <kito.cheng@gmail.com>
7992 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7993 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7994 * config/nds32/nds32-protos.h (nds32_output_call,
7995 nds32_output_return): Declare.
7996 * config/nds32/nds32.md: Refine all the call and return patterns.
7998 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8001 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
8002 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
8004 PR rtl-optimization/84872
8005 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
8006 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
8009 2018-04-06 Tamar Christina <tamar.christina@arm.com>
8011 * expr.c (copy_blkmode_to_reg): Revert 254862.
8012 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
8014 2018-04-06 Richard Biener <rguenther@suse.de>
8017 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
8018 after seeing a component reference with an adjacent field. Treat
8019 refs to arrays at struct end of external decls similar to
8020 refs to unconstrained commons.
8022 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8025 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
8026 look through SAVE_EXPRs with non-side-effects argument. Adjust
8028 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
8031 2018-04-06 Richard Biener <rguenther@suse.de>
8034 * alias.c (find_base_term): New wrapper around find_base_term
8035 unwinding CSELIB_VAL_PTR changes.
8036 (find_base_term): Do not restore CSELIB_VAL_PTR during the
8039 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8041 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
8043 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
8044 constant definitions.
8045 ("nop"): lr 0,0 -> nopr r0
8046 ("nop_lr0", "nop_lr1"): New insn definitions.
8048 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8050 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
8051 NDS32_V3PUSH_AVAILABLE_P macro.
8053 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
8054 Chung-Ju Wu <jasonwucj@gmail.com>
8056 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
8057 (nds32*-*-*): Add float and fpu_config into supported_defaults.
8058 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
8059 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
8060 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
8061 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
8062 * config/nds32/constraints.md: New constraints and checking for hard
8063 float configuration.
8064 * config/nds32/iterators.md: New mode iterator and attribute for hard
8065 float configuration.
8066 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
8068 * config/nds32/nds32-fpu.md: New file.
8069 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
8070 deal with hard float code generation.
8071 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
8073 (abi_type, float_reg_number): New enum type.
8074 * config/nds32/nds32-predicates.c: New predicates for hard float.
8075 * config/nds32/nds32-protos.h: Declare functions for hard float.
8076 * config/nds32/nds32.c: Implementation for hard float configuration.
8077 * config/nds32/nds32.h: Definitions for hard float configuration.
8078 * config/nds32/nds32.md: Include hard float machine description and
8079 modify patterns for hard float configuration.
8080 * config/nds32/nds32.opt: New options for hard float configuration.
8081 * config/nds32/predicates.md: New predicates for hard float
8084 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8086 * common/config/nds32/nds32-common.c
8087 (nds32_option_optimization_table): Enable -mreleax-hint by default.
8089 2018-04-05 Jakub Jelinek <jakub@redhat.com>
8092 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
8093 CONSTRUCTOR_ELT (ctor, ...)->value.
8095 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
8098 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
8100 2018-04-05 Tom de Vries <tom@codesourcery.com>
8103 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
8106 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8107 Kito Cheng <kito.cheng@gmail.com>
8109 * config/nds32/constraints.md (U33): Fine-tune checking condition.
8110 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
8111 * config/nds32/nds32.h (nds32_16bit_address_type): Add
8112 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
8114 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8115 Kito Cheng <kito.cheng@gmail.com>
8117 * config/nds32/constraints.md (Ufe): New memory constraint.
8118 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
8119 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
8120 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
8122 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
8123 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
8125 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8127 * config/nds32/nds32.md: Use optimize_size in the condition for
8128 alu-shift instructions.
8130 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8132 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
8134 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8136 * config/nds32/nds32.md (negsi2): Refine pattern.
8138 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8139 Chung-Ju Wu <jasonwucj@gmail.com>
8141 * config/nds32/iterators.md (shift_rotate): New code iterator.
8142 (shift): New code attribute.
8143 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
8144 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
8145 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
8146 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
8147 bit-wise operations.
8148 (andsi3, *andsi3): Ditto.
8149 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
8150 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
8151 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
8152 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
8153 nds32_ior_operand, nds32_xor_operand): New predicates.
8155 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8157 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
8158 (addsi3, subsi3): ... this.
8160 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8162 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
8164 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8166 * config/nds32/nds32.md: Adjust indention.
8168 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8170 * config/nds32/nds32.md (feature): New attribute.
8172 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8174 * config/nds32/nds32.md (subtype): New attribute.
8176 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
8179 * config/arm/arm-builtins.c (arm_expand_builtin): Change
8180 expansion to perform a bitwise AND of the argument followed by a
8181 boolean negation of the result.
8183 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
8185 PR rtl-optimization/84878
8186 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
8187 the basic block. Assert the use reference is not artificial and that
8188 it has an associated insn.
8190 2018-04-04 Michael Matz <matz@suse.de>
8192 * builtins.c (compute_objsize): Pass correct operand
8193 to array_at_struct_end_p.
8195 2018-04-04 Richard Biener <rguenther@suse.de>
8198 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
8199 from contexts for DINFO_LEVEL_TERSE and below.
8201 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8203 * config/nds32/nds32-doubleword.md (move_<mode>): Require
8204 resiter_operand condition.
8205 * config/nds32/nds32.md (*move<mode>): Ditto.
8207 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8208 Monk Chiang <sh.chiang04@gmail.com>
8210 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
8212 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8214 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
8216 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8217 Kito Cheng <kito.cheng@gmail.com>
8219 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
8220 nds32_cond_code_str, output_cond_branch,
8221 output_cond_branch_compare_zero, nds32_expand_cbranch,
8222 nds32_expand_cstore, nds32_expand_movcc,
8223 nds32_output_cbranchsi4_equality_zero,
8224 nds32_output_cbranchsi4_equality_reg,
8225 nds32_output_cbranchsi4_equality_reg_or_const_int,
8226 nds32_output_cbranchsi4_greater_less_zero: New functions.
8227 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
8228 nds32_expand_cstore, nds32_expand_movcc,
8229 nds32_output_cbranchsi4_equality_zero,
8230 nds32_output_cbranchsi4_equality_reg,
8231 nds32_output_cbranchsi4_equality_reg_or_const_int,
8232 nds32_output_cbranchsi4_greater_less_zero): Declare.
8233 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
8234 nds32_rimm11s_operand): New predicates.
8235 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
8236 * config/nds32/nds32.md: Rewrite all the branch and conditional move
8239 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8241 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
8242 * config/nds32/nds32.md: Ditto.
8243 * config/nds32/pipelines.md: Ditto.
8245 2018-04-04 Richard Biener <rguenther@suse.de>
8247 PR tree-optimization/85168
8248 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
8249 propagating abnormals.
8251 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8253 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
8255 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8256 Kito Cheng <kito.cheng@gmail.com>
8258 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
8259 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
8260 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
8261 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
8262 * config/nds32/nds32.md (sibcall_internal): New.
8263 (sibcall_register): Remove.
8264 (sibcall_immediate): Remove.
8265 (sibcall_value_internal): New.
8266 (sibcall_value_register): Remove.
8267 (sibcall_value_immediate): Remove.
8268 * config/nds32/predicates.md (nds32_general_register_operand): New.
8269 (nds32_call_address_operand): New.
8271 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8273 PR rtl-optimization/85167
8274 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
8275 bb_defs if *split_p, instead preinitialize it to NULL.
8277 PR tree-optimization/85156
8278 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
8279 evaluating the argument multiple times.
8281 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
8283 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
8285 (_mm_cvtpd_ps): Likewise.
8286 (_mm_cvttpd_epi32): Likewise.
8287 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
8288 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
8289 vector, pixel, and bool following altivec.h include.
8291 2018-04-03 Martin Sebor <msebor@redhat.com>
8293 * doc/extend.texi (Common Function Attributes): Clarify.
8294 (const attribute): Likewise.
8295 (pure attribute): Likewise.
8297 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8300 * config/i386/i386.c (ix86_expand_vector_set): Use
8301 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
8303 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
8305 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
8306 instructions when changing rounding bits to preserve precision bits
8307 in the x87 control word.
8309 2018-04-03 Martin Liska <mliska@suse.cz>
8311 PR tree-optimization/82491
8312 * rtl.h (strip_offset_and_add): Replace += suboffset with
8313 poly_uint64 () + suboffset.
8315 2018-03-29 Martin Liska <mliska@suse.cz>
8316 Martin Jambor <mjambor@suse.cz>
8319 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
8320 param_type is not an integral or pointer type.
8322 2018-04-03 Richard Biener <rguenther@suse.de>
8324 * sese.h (recompute_all_dominators): Remove.
8326 2018-04-02 Martin Sebor <msebor@redhat.com>
8328 * doc/invoke.texi (-Wrestrict): Fix typos.
8330 2018-04-02 Jim Wilson <jimw@sifive.com>
8332 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
8333 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
8334 (<optab>di3, <optab>si3_extend): Likewise.
8335 (<optab>si3_mask, <optab>si3_mask_1): New.
8336 (<optab>di3_mask, <optab>di3_mask_1): New.
8337 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
8338 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
8339 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
8341 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
8343 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
8346 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
8348 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
8349 (nds32_canonicalize_comparison): New function.
8351 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8352 Kito Cheng <kito.cheng@gmail.com>
8353 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8355 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
8356 * config/nds32/constants.md (unspec_volatile_element): Add
8357 UNSPEC_VOLATILE_RELAX_GROUP.
8358 * config/nds32/nds32-relax-opt.c: New file.
8359 * config/nds32/nds32-predicates.c
8360 (nds32_symbol_load_store_p): New function.
8361 * config/nds32/nds32-protos.h
8362 (nds32_symbol_load_store_p): Declare function.
8363 (make_pass_nds32_relax_opt): Declare new rtl pass function.
8364 * config/nds32/nds32.c
8365 (nds32_register_pass): New function to register pass.
8366 (nds32_register_passes): New function to register passes.
8367 * config/nds32/nds32.md (relax_group): New pattern.
8368 * config/nds32/nds32.opt (mrelax-hint): New option.
8369 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
8371 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
8373 * config/nds32/t-nds32: Modify files dependency.
8375 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8377 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
8378 (PROFILE_HOOK): Define its implementation.
8380 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8382 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
8383 type and 32-bit size.
8385 2018-04-01 Jakub Jelinek <jakub@redhat.com>
8388 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
8389 (V_128_256): New mode iterator.
8390 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
8391 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
8392 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
8394 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
8395 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
8397 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
8400 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
8401 NaN inputs correctly.
8403 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
8406 * config/rs6000/vsx.md (??r): New mode attribute.
8407 (*vsx_mov<mode>_64bit): Use it.
8408 (*vsx_mov<mode>_32bit): Likewise.
8410 2018-03-30 Martin Sebor <msebor@redhat.com>
8412 PR tree-optimization/84818
8413 * builtins.c (check_access): Use warning_n.
8415 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8418 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
8420 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
8423 2018-03-30 Julia Koval <julia.koval@intel.com>
8426 * x86-tune.def (movx, partial_reg_dependency): Enable for
8429 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
8432 * lra-constraints.c (process_alt_operands): Move setting
8433 this_alternative_matches below.
8435 2018-03-29 Martin Liska <mliska@suse.cz>
8438 * doc/invoke.texi: Document how LTO works with debug info.
8439 Describe auto-load support of binutils. Mention 'x86-64'
8440 as valid option value of -march option.
8442 2018-03-29 Jakub Jelinek <jakub@redhat.com>
8444 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
8447 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
8448 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
8449 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
8452 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8455 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
8456 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
8457 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
8458 for RS6000_BTM_POWERPC64.
8459 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
8460 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
8461 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
8466 2018-03-28 Carl Love <cel@us.ibm.com>
8469 2017-09-27 Carl Love <cel@us.ibm.com>
8471 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
8472 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
8473 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
8476 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8478 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
8479 instead of __vector bool.
8480 (_mm_max_pu8): Likewise.
8481 (_mm_min_pi16): Likewise.
8483 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8486 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
8487 (DIVWEUO): Likewise.
8489 (DIVDEUO): Likewise.
8490 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
8491 DIVWEUO and DIVDEUO.
8492 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
8493 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
8494 (div_extend): Likewise.
8495 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
8497 (__builtin_divweuo): Likewise.
8498 (__builtin_divdeo): Likewise.
8499 (__builtin_divdeuo): Likewise.
8501 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8504 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
8505 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
8507 PR tree-optimization/82004
8508 * gimple-match-head.c (optimize_pow_to_exp): New function.
8509 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
8510 Don't fold to exp if optimize_pow_to_exp is false.
8512 2018-03-28 Martin Liska <mliska@suse.cz>
8515 * calls.c (initialize_argument_information): Fix trailing space.
8516 * common.opt: Fix typo and provide better explanation for
8517 -fsanitize-coverage option.
8518 * config/i386/i386.opt: Fix typo.
8520 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8521 Martin Liska <mliska@suse.cz>
8524 * gimplify.c (asan_poison_variable): Don't do the check for
8525 gimplify_omp_ctxp here.
8526 (gimplify_decl_expr): Do it here.
8527 (gimplify_target_expr): Likewise.
8529 2018-03-28 Martin Liska <mliska@suse.cz>
8532 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8533 chkp_type_bounds_count if MPX is not enabled.
8535 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
8537 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
8539 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8542 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
8543 function to create the function decl for complex long double
8544 multiply and divide for -mabi=ieeelongdouble.
8545 (init_float128_ieee): Call it.
8547 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8550 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8551 -fcf-protection=branch -mibt.
8552 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8554 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8557 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8558 arm_word_relocations.
8560 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8563 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8564 extern array declarations.
8566 2018-03-27 Richard Biener <rguenther@suse.de>
8569 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8570 explicit single_use checks.
8572 2018-03-27 Richard Biener <rguenther@suse.de>
8574 PR tree-optimization/85082
8575 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8578 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8580 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8581 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8582 Turn on fasynchronous-unwind-tables and funwind-tables.
8584 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8587 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8588 (*bmi_blsr_<mode>_ccz): Ditto.
8590 2018-03-26 Tom de Vries <tom@codesourcery.com>
8592 PR tree-optimization/85063
8593 * omp-general.c (offloading_function_p): New function. Factor out
8595 * omp-offload.c (pass_omp_target_link::gate): ... here.
8596 * omp-general.h (offloading_function_p): Declare.
8597 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8598 with attribute omp declare target for offloading functions.
8600 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8602 PR tree-optimization/84005
8603 * tree-data-ref.h (get_base_for_alignment): Declare.
8604 * tree-data-ref.c (get_base_for_alignment_1): New function.
8605 (get_base_for_alignment): Likewise.
8606 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8607 get_base_for_alignment to find a suitable base object, instead
8608 of always using drb->base_address.
8610 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8613 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8614 known size by default.
8616 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8619 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8620 and non BLKmode operands.
8622 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8625 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8626 Clean up attributes.
8628 2018-03-23 Richard Biener <rguenther@suse.de>
8631 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8632 we are going to emit early debug for LTO.
8634 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8637 * function.c (match_asm_constraints_1): Don't optimize if input
8638 doesn't satisfy general_operand predicate for output's mode.
8641 * alias.c (write_dependence_p): Don't require for x_canonicalized
8642 non-VOIDmode if x has VOIDmode.
8645 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8646 just don't try to optimize it rather than assert it never happens.
8648 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8650 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8651 macro expansions for definition of ST_INTERNAL_<mode> and
8652 LD_INTERNAL_<mode> builtins.
8653 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8655 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8657 (altivec_expand_st_builtin): Likewise.
8658 (altivec_expand_builtin): Remove calls to deleted functions.
8659 (rs6000_address_for_altivec): Delete this function.
8660 * config/rs6000/vector.md: Remove expands for
8661 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8663 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8666 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8667 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8668 re-computing once computed.
8669 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8670 (arm_init_machine_status): Initialize
8671 machine->static_chain_stack_bytes.
8673 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8676 * doc/extend.texi: Add four new prototypes for vec_ld.
8677 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8678 definitions for more logical presentation.
8679 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8680 entries for V1TI variants of __builtin_altivec_ld builtin.
8681 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8682 handling of V1TI variant of LVX icode pattern.
8683 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8684 (rs6000_gimple_fold_builtin): Likewise.
8685 (altivec_init_builtins): Add code to define
8686 __builtin_altivec_lvx_v1ti function.
8688 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8691 * function.c (match_asm_constraints_1): Don't do the optimization
8692 if input isn't a REG, SUBREG, MEM or constant.
8694 2018-03-22 Tom de Vries <tom@codesourcery.com>
8696 PR tree-optimization/84956
8697 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8698 bb_has_abnormal_pred.
8700 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8703 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8704 DECL_INITIAL (decl) to decl at the end.
8705 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8708 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8710 * doc/extend.texi (__builtin_tgmath): Document when complex
8711 integer types are treated as _Complex _Float64.
8713 2018-03-21 Tom de Vries <tom@codesourcery.com>
8715 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8717 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8719 PR tree-optimization/84960
8720 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8721 if it is ENTRY block, move them into single succ of ENTRY in that case.
8723 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8725 PR tree-optimization/84811
8726 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8727 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8728 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8729 (known_subrange_p): Update accordingly. Cast each value involved
8730 in the size comparison, rather than casting the result of the
8733 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8735 PR tree-optimization/84982
8736 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8737 by flipping the least significant bit rather than all bits from
8738 bitpos to bitpos + bitsize - 1.
8740 2018-03-21 Nathan Sidwell <nathan@acm.org>
8742 * doc/extend.texi (Deprecated Features): Remove mention of
8743 long-deleted deprecations.
8745 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8748 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8749 * configure: Regenerate.
8751 2018-03-21 Tom de Vries <tom@codesourcery.com>
8753 PR tree-optimization/83126
8754 * tree-parloops.c (num_phis): New function.
8755 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8757 2018-03-21 Nathan Sidwell <nathan@acm.org>
8759 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8760 mention anon-struct/union members and trailing attributes.
8762 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8764 PR tree-optimization/84969
8765 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8766 builtin memset partitions if they set different rhs values.
8768 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8770 PR rtl-optimization/84989
8771 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8772 VEC_DUPLICATE with scalar result mode.
8774 2018-03-21 Martin Liska <mliska@suse.cz>
8777 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8778 not intended return statement.
8780 2018-03-21 Martin Liska <mliska@suse.cz>
8783 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8784 (chkp_find_bound_slots_1): Limit number of iterations.
8786 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8789 * Minor grammar fixes for x86 options.
8791 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8794 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8795 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8797 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8800 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8801 (altivec_lvx_<mode>_1op): Likewise.
8802 (altivec_stvx_<mode>_2op): Likewise.
8803 (altivec_stvx_<mode>_1op): Likewise.
8804 (altivec_lvx_<VM2:mode>): New define_expand.
8805 (altivec_stvx_<VM2:mode>): Likewise.
8806 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8807 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8808 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8809 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8810 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8811 (rs6000_gen_lvx): Likewise.
8812 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8813 (altivec_expand_stv_builtin): Likewise.
8814 (altivec_expand_builtin): Likewise.
8815 * config/rs6000/vector.md: Likewise.
8817 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8820 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8823 2018-03-20 Richard Biener <rguenther@suse.de>
8826 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8827 sign-conversions as zero, fall back to standard scalar_stmt
8830 2018-03-20 Martin Liska <mliska@suse.cz>
8833 * predict.c (rebuild_frequencies): Handle case when we have
8834 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8836 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8839 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8840 flag_section_anchors.
8841 * varasm.c (use_blocks_for_decl_p): Remove hack for
8842 dw2_force_const_mem.
8845 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8847 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8848 be created, use lowpart_subreg of operands[0] rather than operands[0]
8850 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8851 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8852 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8853 and n constraint instead of aarch64_shift_imm_di and Usd.
8854 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8855 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8857 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8860 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8861 to favor GPR over NEON registers.
8862 (<shift>di3_neon): Likewise.
8864 2018-03-20 Tom de Vries <tom@codesourcery.com>
8867 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8868 (nvptx_process_pars): Emit bar.sync asap and alap.
8870 2018-03-20 Tom de Vries <tom@codesourcery.com>
8873 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8874 seen_label if seen_label is already set.
8876 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8879 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8880 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8881 Use 1U instead of 1. Formatting fixes.
8884 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8885 instead of TREE_TYPE (s1) for the return value.
8887 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8889 PR tree-optimization/84946
8890 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8891 bitsize + bitsize in poly_uint64 rather than poly_int64.
8894 * dwarf2asm.c: Include fold-const.c.
8895 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8896 of decl rather than decl itself.
8898 PR rtl-optimization/84643
8899 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8901 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8904 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8905 calling assemble_variable.
8907 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8910 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8911 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8913 2018-03-19 Jim Wilson <jimw@sifive.com>
8916 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8917 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8918 (riscv_first_stack_step): Likewise.
8919 (riscv_option_override): Use STACK_BOUNDARY instead of
8921 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8923 (BIGGEST_ALIGNMENT): Set to 128.
8924 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8925 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8928 2018-03-19 Richard Biener <rguenther@suse.de>
8930 PR tree-optimization/84933
8931 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8932 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8934 2018-03-19 Richard Biener <rguenther@suse.de>
8936 PR tree-optimization/84859
8937 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8938 (cond_if_else_store_replacement): Perform sinking operation on
8939 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8940 Generalize what a BB with a single eligible store is.
8942 2018-03-19 Richard Biener <rguenther@suse.de>
8944 PR tree-optimization/84929
8945 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8946 chrec_is_positive against non-chrec arg.
8948 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8951 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8953 2018-03-18 Martin Liska <mliska@suse.cz>
8955 PR rtl-optimization/84635
8956 * regrename.c (build_def_use): Use matches_mode only when
8959 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8961 PR tree-optimization/84913
8962 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8963 vectorize chains of COND_EXPRs.
8965 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8967 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8969 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8971 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8973 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8975 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8977 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8978 Kito Cheng <kito.cheng@gmail.com>
8980 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8981 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8982 (nds32_adjust_reg_alloc_order): New function.
8983 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8985 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8987 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8988 nds32_print_operand, nds32_print_operand_address): Use
8989 HOST_WIDE_INT_PRINT_DEC instead.
8991 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8993 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8995 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8998 * config/i386/i386.c (initial_ix86_tune_features,
8999 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
9001 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
9002 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
9003 rather than 1u << ix86_tune. Formatting fix.
9004 (ix86_option_override_internal): Change ix86_arch_mask from
9005 unsigned int to unsigned HOST_WIDE_INT, initialize to
9006 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
9007 (ix86_function_specific_restore): Likewise.
9009 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9012 * postreload.c (reload_combine_recognize_pattern): Perform
9013 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
9014 truncate_int_for_mode the result for the destination's mode.
9017 * hsa-gen.c (mem_type_for_type): Fix comment typo.
9018 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
9020 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9023 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
9026 * lra-assigns.c (lra_split_hard_reg_for): Don't use
9027 regno_allocno_class_array and sorted_pseudos.
9028 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
9029 insns where regno is used.
9031 2018-03-16 Martin Liska <mliska@suse.cz>
9034 * multiple_target.c (create_dispatcher_calls): Redirect
9035 reference in the symbol table.
9037 2018-03-16 Martin Liska <mliska@suse.cz>
9040 * multiple_target.c (create_dispatcher_calls): Redirect also
9043 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9047 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
9048 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
9049 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
9051 2018-03-16 Julia Koval <julia.koval@intel.com>
9053 * doc/invoke.texi (Skylake Server): Add CLWB.
9054 Cannonlake): Remove CLWB.
9056 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9058 PR tree-optimization/84841
9059 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
9061 (FLOAT_ONE_CONST_TYPE): Define.
9062 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
9063 (sort_by_operand_rank): Put entries with higher constant_type last
9064 rather than first to match comments.
9066 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
9068 * config/nios2/nios2.md (movsi_internal): Fix thinko in
9071 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9074 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
9075 check and use address of target always.
9077 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
9080 * config/i386/i386.c (indirect_thunk_needed): Update comments.
9081 (indirect_thunk_bnd_needed): Likewise.
9082 (indirect_thunks_used): Likewise.
9083 (indirect_thunks_bnd_used): Likewise.
9084 (indirect_return_needed): New.
9085 (indirect_return_bnd_needed): Likewise.
9086 (output_indirect_thunk_function): Add a bool argument for
9088 (output_indirect_thunk_function): Don't generate alias for
9089 function return thunk.
9090 (ix86_code_end): Call output_indirect_thunk_function to generate
9091 function return thunks.
9092 (ix86_output_function_return): Set indirect_return_bnd_needed
9093 and indirect_return_needed instead of indirect_thunk_bnd_needed
9094 and indirect_thunk_needed.
9096 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
9098 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
9099 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
9100 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
9102 2018-03-15 David Malcolm <dmalcolm@redhat.com>
9103 Paul Hua <paul.hua.gm@gmail.com>
9106 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
9108 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
9110 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
9111 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
9114 2018-03-15 Tamar Christina <tamar.christina@arm.com>
9117 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9118 instead of GET_MODE_SIZE when comparing Units.
9120 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9123 * varasm.c (hash_section): Return an unchangeble hash value
9124 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
9125 Return !aarch64_can_use_per_function_literal_pools_p ().
9127 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9130 * optabs.c (emit_conditional_move): Pass address of cmode's copy
9131 rather than address of cmode as last argument to prepare_cmp_insn.
9133 2018-03-15 Julia Koval <julia.koval@intel.com>
9135 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
9136 F_AVX512VNNI, F_AVX512BITALG): New.
9138 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
9141 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
9142 insn for floating-point loads and stores.
9144 2018-03-14 Carl Love <cel@us.ibm.com>
9146 * config/rs6000/rs6000-c.c: Add macro definitions for
9147 ALTIVEC_BUILTIN_VEC_PERMXOR.
9148 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
9149 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
9150 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
9151 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
9153 * config/doc/extend.texi: Add prototypes for vec_permxor.
9155 2018-03-14 David Malcolm <dmalcolm@redhat.com>
9158 * diagnostic-show-locus.c (class layout_point): Convert m_line
9159 from int to linenum_type.
9160 (line_span::comparator): Use linenum "compare" function when
9161 comparing line numbers.
9162 (test_line_span): New function.
9163 (layout_range::contains_point): Convert param "row" from int to
9165 (layout_range::intersects_line_p): Likewise.
9166 (layout::will_show_line_p): Likewise.
9167 (layout::print_source_line): Likewise.
9168 (layout::should_print_annotation_line_p): Likewise.
9169 (layout::print_annotation_line): Likewise.
9170 (layout::print_leading_fixits): Likewise.
9171 (layout::annotation_line_showed_range_p): Likewise.
9172 (struct line_corrections): Likewise for field m_row.
9173 (line_corrections::line_corrections): Likewise for param "row".
9174 (layout::print_trailing_fixits): Likewise.
9175 (layout::get_state_at_point): Likewise.
9176 (layout::get_x_bound_for_row): Likewise.
9177 (layout::print_line): Likewise.
9178 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
9179 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
9180 * input.c (selftest::test_linenum_comparisons): New function.
9181 (selftest::input_c_tests): Call it.
9182 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
9183 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
9184 * selftest.h (ASSERT_GT): New macro.
9185 (ASSERT_GT_AT): New macro.
9186 (ASSERT_LT): New macro.
9187 (ASSERT_LT_AT): New macro.
9189 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
9191 PR rtl-optimization/84780
9192 * combine.c (distribute_links): Don't make a link based on pc_rtx.
9194 2018-03-14 Martin Liska <mliska@suse.cz>
9196 * tree.c (record_node_allocation_statistics): Use
9197 get_stats_node_kind.
9198 (get_stats_node_kind): New function extracted from
9199 record_node_allocation_statistics.
9200 (free_node): Use get_stats_node_kind.
9202 2018-03-14 Richard Biener <rguenther@suse.de>
9204 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
9205 that the value-set of ANTIC_IN doesn't grow.
9208 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9210 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9211 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9213 2018-03-14 Julia Koval <julia.koval@intel.com>
9215 * config.gcc (icelake-client, icelake-server): New.
9217 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
9218 (initial_ix86_arch_features): Ditto.
9219 (PTA_SKYLAKE): Add SGX.
9220 (PTA_ICELAKE): Remove.
9221 (PTA_ICELAKE_CLIENT): New.
9222 (PTA_ICELAKE_SERVER): New.
9223 (ix86_option_override_internal): Split up icelake on icelake client and
9225 (get_builtin_code_for_version): Ditto.
9226 (fold_builtin_cpu): Ditto.
9227 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
9228 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
9229 * config/i386/i386.h (processor_type): Ditto.
9230 * doc/invoke.texi: Ditto.
9232 2018-03-14 Jakub Jelinek <jakub@redhat.com>
9235 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
9236 INTEGER_CST offset, add it together with bitpos / 8 and
9237 sign extend based on POINTER_SIZE.
9241 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
9244 * config/i386/constraints.md (Yc): New register constraint.
9245 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
9246 Use Yc constraint for alternative 2 of operand 0. Remove
9247 preferred_for_speed attribute.
9249 2018-03-14 Richard Biener <rguenther@suse.de>
9251 PR tree-optimization/84830
9252 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
9253 with the old one to avoid oscillations.
9255 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
9258 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
9260 (assign_by_spills): Return a flag of reload assignment failure.
9261 Do not process the reload assignment failures. Do not spill other
9262 reload pseudos if they has the same reg class. Update n if
9264 (lra_assign): Add a return arg. Set up from the result of
9265 assign_by_spills call.
9266 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9267 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9268 usage_insns if it is not NULL.
9269 (spill_hard_reg_in_range): New function.
9270 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9271 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9272 function prototypes.
9273 (lra_assign): Change prototype.
9274 * lra.c (lra): Add code to deal with fails by splitting hard reg
9277 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
9279 * config/riscv/riscv.opt (mrelax): New option.
9280 * config/riscv/riscv.c (riscv_file_start): Emit ".option
9281 "norelax" when riscv_mrelax is disabled.
9282 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
9284 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9287 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
9288 reassociation for int modes.
9290 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9292 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
9293 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
9295 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
9296 * config/aarch64/aarch64-sve.md
9297 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
9298 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
9299 (*extend<mode><Vwide>2): Rename to...
9300 (aarch64_sve_extend<mode><Vwide>2): ...this.
9301 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
9302 renaming the old pattern to...
9303 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
9305 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
9306 define_expand, renaming the old pattern to...
9307 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
9308 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
9309 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
9310 account when deciding which SVE instruction the optab should use.
9311 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
9313 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9315 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
9316 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
9317 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
9318 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
9319 (tlsdesc_small_<mode>): Turn a define_expand and use
9320 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
9321 (tlsdesc_small_advsimd_<mode>): ...this.
9322 (tlsdesc_small_sve_<mode>): New pattern.
9324 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9326 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
9327 (UNSPEC_UMUL_HIGHPART): New constants.
9328 (MUL_HIGHPART): New int iteraor.
9329 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
9330 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
9332 (*<su>mul<mode>3_highpart): New define_insn.
9334 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9337 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9340 2018-03-13 Martin Liska <mliska@suse.cz>
9343 * (sem_item_optimizer::sem_item_optimizer): Initialize new
9345 (sem_item_optimizer::~sem_item_optimizer): Release it.
9346 (sem_item_optimizer::merge_classes): Register variable aliases.
9347 (sem_item_optimizer::fixup_pt_set): New function.
9348 (sem_item_optimizer::fixup_points_to_sets): Likewise.
9349 * ipa-icf.h: Declare new variables and functions.
9351 2018-03-13 Jakub Jelinek <jakub@redhat.com>
9354 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
9355 integer_pow2p@2 and test integer_pow2p in condition.
9356 (A < 0 ? C : 0): Similarly for @1.
9359 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
9360 characters starting at p contain '\0' character, don't look beyond
9364 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
9365 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
9368 * reg-stack.c (change_stack): Change update_end var from int to
9369 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
9370 also call set_block_for_insn on the newly added insns and rescan.
9373 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
9374 on the last operand.
9377 * tree.c (stabilize_reference_1): Return save_expr (e) for
9378 STATEMENT_LIST even if it doesn't have side-effects.
9380 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
9382 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
9384 2018-03-12 Renlin Li <renlin.li@arm.com>
9386 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
9387 aarch64_output_scalar_simd_mov_immediate.
9389 2018-03-12 Martin Sebor <msebor@redhat.com>
9391 PR tree-optimization/83456
9392 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
9393 for perfectly overlapping calls to memcpy.
9394 (gimple_fold_builtin_memory_chk): Same.
9395 (gimple_fold_builtin_strcpy): Handle no-warning.
9396 (gimple_fold_builtin_stxcpy_chk): Same.
9397 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
9399 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9401 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
9402 parameter. Use it for SFmode.
9403 (rs6000_function_arg_advance_1): Adjust.
9404 (rs6000_function_arg): Adjust.
9405 (rs6000_gimplify_va_arg): Pass false for that new parameter.
9407 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9409 PR rtl-optimization/84169
9410 PR rtl-optimization/84780
9411 * combine.c (can_combine_p): Check for a 2-insn combination whether
9412 the destination register is used between the two insns, too.
9414 2018-03-12 Richard Biener <rguenther@suse.de>
9416 PR tree-optimization/84803
9417 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
9418 for refs DR analysis didn't process.
9420 2018-03-12 Richard Biener <rguenther@suse.de>
9422 PR tree-optimization/84777
9423 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
9424 force-vectorize loops ignore whether we are optimizing for size.
9426 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
9428 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
9429 (TARGET_MD_ASM_ADJUST): Define.
9431 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
9432 Kito Cheng <kito.cheng@gmail.com>
9433 Chung-Ju Wu <jasonwucj@gmail.com>
9435 * config/nds32/nds32.c (nds32_compute_stack_frame,
9436 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
9437 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
9438 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
9439 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
9440 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
9441 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
9442 * config/nds32/nds32.md (prologue, epilogue): Use macro
9443 NDS32_V3PUSH_AVAILABLE_P to do checking.
9445 2018-03-11 Jakub Jelinek <jakub@redhat.com>
9448 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
9449 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
9450 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
9451 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
9452 addition of most attributes on !orig_type_die or the attribute not
9453 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
9455 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9456 Chung-Ju Wu <jasonwucj@gmail.com>
9458 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
9460 * config/nds32/nds32.opt (mvh): New option.
9462 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9463 Chung-Ju Wu <jasonwucj@gmail.com>
9465 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
9467 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
9468 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
9471 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9472 Chung-Ju Wu <jasonwucj@gmail.com>
9474 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
9476 * config/nds32/nds32-multiple.md (strlensi): New pattern.
9477 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
9479 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
9480 Kito Cheng <kito.cheng@gmail.com>
9481 Chung-Ju Wu <jasonwucj@gmail.com>
9483 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
9484 UNSPEC_FFMISM and UNSPEC_FLMISM.
9485 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
9486 for ffb, ffmism and flmism.
9487 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
9488 (unspec_ffmism): Ditto.
9489 (unspec_flmism): Ditto.
9490 (nds32_expand_builtin_impl): Check if string extension is available.
9491 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
9492 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
9494 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
9497 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9500 * lra-assigns.c (assign_by_spills): Return a flag of reload
9501 assignment failure. Do not process the reload assignment
9502 failures. Do not spill other reload pseudos if they has the same
9504 (lra_assign): Add a return arg. Set up from the result of
9505 assign_by_spills call.
9506 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9507 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9508 usage_insns if it is not NULL.
9509 (spill_hard_reg_in_range): New function.
9510 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9511 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9512 function prototypes.
9513 (lra_assign): Change prototype.
9514 * lra.c (lra): Add code to deal with fails by splitting hard reg
9517 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9520 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9522 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9525 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9526 (dw_val_node): Add val_symbolic_view.
9527 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9528 (symview_upper_bound): New.
9529 (new_line_info_table): Initialize symviews_since_reset.
9530 (dwarf2out_source_line): Count symviews_since_reset and set
9531 symview_upper_bound.
9532 (dw_val_equal_p): Handle symview.
9533 (add_AT_symview): New.
9534 (print_dw_val): Handle symview.
9535 (attr_checksum, attr_checksum_ordered): Likewise.
9536 (same_dw_val_p, size_of_die): Likewise.
9537 (value_format, output_die): Likewise.
9538 (add_high_low_attributes): Use add_AT_symview for entry_view.
9539 (dwarf2out_finish): Reset symview_upper_bound, clear
9542 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
9545 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
9546 Add strict argument and use it.
9547 (rs6000_split_multireg_move): Update for new strict argument.
9548 (mem_operand_gpr): Disallow all non-offsettable addresses.
9549 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9551 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9554 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9555 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9556 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9559 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9560 decl, use remap_type if we want to use the type.
9562 2018-03-09 Martin Sebor <msebor@redhat.com>
9564 PR tree-optimization/84526
9565 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9567 (builtin_access::generic_overlap): Be prepared to handle non-array
9570 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9572 PR rtl-optimization/84682
9573 * lra-constraints.c (process_address_1): Check is_address flag
9574 for address constraints.
9575 (process_alt_operands): Likewise.
9576 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9577 preprocess_constraints.
9578 * recog.h (preprocess_constraints): Add oploc parameter.
9580 * recog.c (preprocess_constraints): Test address_operand for
9581 CT_ADDRESS constraints.
9583 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9586 * lra-assigns.c (assign_by_spills): Return a flag of reload
9587 assignment failure. Do not process the reload assignment
9588 failures. Do not spill other reload pseudos if they has the same
9590 (lra_assign): Add a return arg. Set up from the result of
9591 assign_by_spills call.
9592 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9593 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9594 usage_insns if it is not NULL.
9595 (spill_hard_reg_in_range): New function.
9596 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9597 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9598 function prototypes.
9599 (lra_assign): Change prototype.
9600 * lra.c (lra): Add code to deal with fails by splitting hard reg
9603 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9606 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9607 Accept complain bool parameter. Only emit errors if it is true.
9608 (arm_parse_cpu_option_name): Likewise.
9609 (arm_target_thumb_only): Adjust callers of the above.
9610 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9611 prototype to take a default true bool parameter.
9612 (arm_parse_arch_option_name): Likewise.
9614 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9615 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9619 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9620 * configure: Regenerate.
9621 * configure.ac ("linker --version-script option"): New.
9622 ("linker soname option"): New.
9624 2018-03-09 Richard Biener <rguenther@suse.de>
9626 PR tree-optimization/84775
9627 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9628 immediate uses of predicate stmts and mark them modified.
9631 PR tree-optimization/84178
9632 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9634 (version_loop_for_if_conversion): Delay update_ssa call.
9635 (tree_if_conversion): Delay update_ssa until after predicate
9638 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9641 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9642 when the function accesses prior frames.
9644 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9647 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9648 gen_llsym, otherwise call maybe_gen_llsym.
9651 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9652 has ',' character inside of it.
9654 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9657 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9658 as clobbering CC_REGNUM.
9660 2018-03-08 Richard Biener <rguenther@suse.de>
9663 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9664 (follow_copies_to_constant): Do not follow SSA names registered
9667 2018-03-08 Richard Biener <rguenther@suse.de>
9669 PR tree-optimization/84178
9670 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9672 (version_loop_for_if_conversion): Delay update_ssa call.
9673 (tree_if_conversion): Delay update_ssa until after predicate
9676 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9678 PR tree-optimization/84178
9679 * tree-if-conv.c (release_bb_predicate): Remove the
9680 the assertion that the stmts have NULL use_ops.
9681 Discard the statements, asserting that they haven't
9682 yet been added to a BB.
9684 2018-03-08 Richard Biener <rguenther@suse.de>
9686 PR tree-optimization/84746
9687 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9688 (phi_translate): Pass in destination ANTIC_OUT set.
9689 (phi_translate_1): Likewise. For a simplified result lookup
9690 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9691 (phi_translate_set): Adjust.
9692 (do_pre_regular_insertion): Likewise.
9693 (do_pre_partial_partial_insertion): Likewise.
9695 2018-03-08 Martin Liska <mliska@suse.cz>
9697 PR gcov-profile/84735
9698 * doc/gcov.texi: Document usage of profile files.
9699 * gcov-io.h: Document changes in the format.
9701 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9705 * dwarf2out.c (struct dw_line_info_table): Update comments for
9707 (FORCE_RESET_NEXT_VIEW): New.
9708 (FORCE_RESETTING_VIEW_P): New.
9709 (RESETTING_VIEW_P): Check for -1 too.
9710 (ZERO_VIEW_P): Likewise.
9711 (new_line_info_table): Force-reset next view.
9712 (dwarf2out_begin_function): Likewise.
9713 (dwarf2out_source_line): Simplify zero_view_p initialization.
9714 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9715 view directly. Omit view when omitting .loc at line 0.
9717 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9719 PR tree-optimization/84740
9720 * tree-switch-conversion.c (process_switch): Call build_constructors
9721 only if info.phi_count is non-zero.
9723 PR tree-optimization/84739
9724 * tree-tailcall.c (find_tail_calls): Check call arguments against
9725 DECL_ARGUMENTS (current_function_decl) rather than
9726 DECL_ARGUMENTS (func) when checking for tail recursion.
9728 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9730 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9731 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9732 Volker Reichelt's entry and add entries for people that perform
9733 GCC fuzzy testing and report numerous bugs.
9735 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9738 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9739 readonly data in sdata, if that is disabled.
9740 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9741 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9742 -mreadonly-in-sdata option.
9744 2018-03-07 Martin Sebor <msebor@redhat.com>
9746 PR tree-optimization/84468
9747 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9748 basic block when looking for nul assignment.
9750 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9753 * except.h (output_function_exception_table): Adjust prototype.
9754 * except.c (output_function_exception_table): Remove FNNAME parameter
9755 and add SECTION parameter. Ouput one part of the table at a time.
9756 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9757 the first part of the exception table and emit unwind directives.
9758 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9759 (i386_pe_seh_cold_init): Likewise.
9760 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9761 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9762 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9763 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9764 * config/i386/winnt.c: Include except.h.
9765 (struct seh_frame_state): Add reg_offset, after_prologue and
9766 in_cold_section fields.
9767 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9768 (i386_pe_seh_cold_init): New function.
9769 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9770 to seh->in_cold_section.
9771 (seh_emit_push): Record the offset of the push.
9772 (seh_emit_save): Record the offet of the save.
9773 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9774 Test seh->after_prologue to disregard the epilogue.
9775 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9776 (i386_pe_end_cold_function): New function.
9778 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9781 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9782 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9785 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9787 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9788 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9791 * multiple_target.c: Include tree-inline.h and intl.h.
9792 (expand_target_clones): Diagnose and fail if node->definition and
9793 !tree_versionable_function_p (node->decl).
9795 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9797 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9799 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9800 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9801 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9803 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9806 * combine.c (try_combine): Use reg_or_subregno instead of handling
9807 just paradoxical SUBREGs and REGs.
9809 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9811 * config/arc/arc.c (arc_finalize_pic): Remove function.
9812 (arc_must_save_register): We use single base PIC register, remove
9813 checks to save/restore the PIC register.
9814 (arc_expand_prologue): Likewise.
9815 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9817 (arc_verify_short): Likewise.
9818 (arc_attr_type): Likewise.
9819 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9820 (walk_stores): Likewise.
9821 (arc_address_cost): Make it static.
9822 (arc_verify_short): Likewise.
9823 (branch_dest): Likewise.
9824 (arc_attr_type): Likewise.
9825 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9826 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9827 (arc_final_prescan_insn): Remove inserting the nops due to
9828 hardware hazards. It is done in reorg step.
9829 (insn_length_variant_t): Remove.
9830 (insn_length_parameters_t): Likewise.
9831 (arc_insn_length_parameters): Likewise.
9832 (arc_get_insn_variants): Likewise.
9833 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9835 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9838 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9841 PR tree-optimization/84687
9842 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9844 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9846 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9848 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9849 Rename to ppc_speculation_barrier.
9850 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9851 __builtin_ppc_speculation_barrier.
9853 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9856 * combine.c (combine_simplify_rtx): Don't try to simplify if
9857 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9860 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9862 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9863 to 32 bytes when compiling for POWER9.
9865 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9868 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9869 regparm >= 3 with no arg reg available also for calls with
9870 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9873 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9875 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9877 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9880 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9882 2018-03-05 Richard Biener <rguenther@suse.de>
9884 PR tree-optimization/84486
9885 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9886 When inserting a __builtin_assume_aligned call set the LHS
9887 SSA name alignment info accordingly.
9889 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9891 PR tree-optimization/84114
9892 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9893 Avoid reassociation of FLOAT_MODE addition.
9895 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9897 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9898 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9899 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9900 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9901 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9902 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9903 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9905 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9906 __builtin_ia32_wbinvd): New builtins.
9907 (SPECIAL_ARGS2): New.
9908 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9909 (SPECIAL_ARGS2): New.
9910 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9911 (ix86_valid_target_attribute_inner_p): Ditto.
9912 (ix86_init_mmx_sse_builtins): Add special_args2.
9913 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9914 TARGET_WBNOINVD_P): New.
9915 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9916 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9917 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9918 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9919 * config/i386/pconfigintrin.h: New file.
9920 * config/i386/wbnoinvdintrin.h: Ditto.
9921 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9923 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9925 2018-03-05 Richard Biener <rguenther@suse.de>
9927 PR tree-optimization/84670
9928 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9930 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9931 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9932 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9933 doesn't grow if all successors (recursively) were visited at least
9936 2018-03-05 Richard Biener <rguenther@suse.de>
9938 PR tree-optimization/84650
9939 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9940 if executed in the loop pipeline.
9942 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9944 * doc/configfiles.texi (Configuration Files): Move info about
9945 conditionalizing $target-protos.h to...
9946 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9947 differs from $target-protos.h.
9949 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9950 Chung-Ju Wu <jasonwucj@gmail.com>
9952 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9953 * config/nds32/nds32-multiple.md (setmemsi): Define.
9954 * config/nds32/nds32-memory-manipulation.c
9955 (nds32_gen_dup_4_byte_to_word_value): New.
9956 (emit_setmem_word_loop): New.
9957 (emit_setmem_byte_loop): New.
9958 (nds32_expand_setmem_loop): New.
9959 (nds32_expand_setmem_loop_v3m): New.
9960 (nds32_expand_setmem_unroll): New.
9961 (nds32_expand_setmem): New.
9963 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9964 Chung-Ju Wu <jasonwucj@gmail.com>
9966 * config/nds32/nds32-memory-manipulation.c
9967 (nds32_emit_load_store): New.
9968 (nds32_emit_post_inc_load_store): New.
9969 (nds32_emit_mem_move): New.
9970 (nds32_emit_mem_move_block): New.
9971 (nds32_expand_movmemsi_loop_unknown_size): New.
9972 (nds32_expand_movmemsi_loop_known_size): New.
9973 (nds32_expand_movmemsi_loop): New.
9974 (nds32_expand_movmemsi_unroll): New.
9975 (nds32_expand_movmemqi): Rename ...
9976 (nds32_expand_movmemsi): ... to this.
9977 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9978 (movmemsi): ... to this.
9979 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9980 (nds32_expand_movmemsi): ... to this.
9982 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9983 Monk Chiang <sh.chiang04@gmail.com>
9984 Chung-Ju Wu <jasonwucj@gmail.com>
9986 * config/nds32/nds32-protos.h
9987 (nds32_expand_load_multiple): New arguments.
9988 (nds32_expand_store_multiple): Ditto.
9989 (nds32_valid_multiple_load_store): Rename ...
9990 (nds32_valid_multiple_load_store_p): ... to this.
9991 * config/nds32/nds32-memory-manipulation.c
9992 (nds32_expand_load_multiple): Refine implementation.
9993 (nds32_expand_store_multiple): Ditto.
9994 * config/nds32/nds32-multiple.md
9995 (load_multiple): Update nds32_expand_load_multiple interface.
9996 (store_multiple): Update nds32_expand_store_multiple interface.
9997 * config/nds32/nds32-predicates.c
9998 (nds32_valid_multiple_load_store): Rename ...
9999 (nds32_valid_multiple_load_store_p): ... to this and refine
10001 * config/nds32/predicates.md
10002 (nds32_load_multiple_and_update_address_operation): New predicate.
10003 (nds32_store_multiple_and_update_address_operation): New predicate.
10005 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10006 Chung-Ju Wu <jasonwucj@gmail.com>
10008 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
10009 (combo): New attribute.
10010 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
10012 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
10014 * config/nds32/nds32.opt: Change -mcmodel= default value.
10016 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
10017 Monk Chiang <sh.chiang04@gmail.com>
10018 Chung-Ju Wu <jasonwucj@gmail.com>
10020 * config/nds32/constants.md (unspec_element): New enum.
10021 * config/nds32/constraints.md (Umw): New constraint.
10022 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
10023 * config/nds32/nds32-intrinsic.md: Likewise.
10024 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
10025 (nds32_valid_smw_lwm_base_p): New.
10026 (nds32_output_smw_single_word): New.
10027 (nds32_output_lmw_single_word): New.
10028 (nds32_expand_unaligned_load): New.
10029 (nds32_expand_unaligned_store): New.
10030 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
10031 (nds32_output_smw_single_word): Declare.
10032 (nds32_output_lmw_single_word): Declare.
10033 (nds32_expand_unaligned_load): Declare.
10034 (nds32_expand_unaligned_store): Declare.
10035 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
10036 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
10037 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
10038 NDS32_BUILTIN_UASTORE_DW.
10039 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
10042 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
10043 Kito Cheng <kito.cheng@gmail.com>
10044 Chung-Ju Wu <jasonwucj@gmail.com>
10046 * config/nds32/nds32-intrinsic.c
10047 (nds32_expand_builtin_null_ftype_reg): Delete.
10048 (nds32_expand_builtin_reg_ftype_imm): Ditto.
10049 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
10050 (nds32_read_argument): New.
10051 (nds32_legitimize_target): Ditto.
10052 (nds32_legitimize_argument): Ditto.
10053 (nds32_check_constant_argument): Ditto.
10054 (nds32_expand_unop_builtin): Ditto.
10055 (nds32_expand_unopimm_builtin): Ditto.
10056 (nds32_expand_binop_builtin): Ditto.
10057 (nds32_builtin_decl_impl): Ditto.
10058 (builtin_description): Ditto.
10059 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
10060 (nds32_init_builtins_impl): Ditto.
10061 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
10062 (nds32_builtin_decl): New.
10063 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
10064 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
10066 2018-03-02 Jeff Law <law@redhat.com>
10068 * reorg.c (stop_search_p): Handle DEBUG_INSN.
10069 (redundant_insn, fill_simple_delay_slots): Likewise.
10070 (fill_slots_from_thread): Likewise.
10071 * resource.c (mark_referenced_resources): Likewise.
10072 (mark_set_resources, find_dead_or_set_registers): Likewise.
10074 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10076 * substring-locations.h (format_warning_va): Formatting fix for
10077 ATTRIBUTE_GCC_DIAG.
10078 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
10080 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
10081 * substring-locations.c: Include intl.h.
10082 (format_warning_va): Turned into small wrapper around
10083 format_warning_n_va, renamed to ...
10084 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
10085 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
10087 (format_warning_at_substring_n): New function.
10088 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
10089 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
10090 format_warning_at_substring with just a shorter name instead of
10091 const function pointer.
10092 (fmtwarn_n): New function.
10093 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
10094 appropriate, get rid of all the fmtstr temporaries, move conditionals
10095 with G_() wrapped string literals directly into fmtwarn arguments,
10096 cast dir.len to (int), formatting fixes.
10098 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
10100 * doc/invoke.texi: Remove "Cilk Plus" references.
10102 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10103 Richard Biener <rguenther@suse.de>
10106 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
10107 for error or warning attributes if CALL_FROM_THUNK_P is set.
10110 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10113 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
10114 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
10117 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
10118 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
10120 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
10121 instead of -1U in last predictors element's probability member.
10123 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10126 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10127 arguments if they are comparable.
10129 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10131 PR tree-optimization/84634
10132 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
10133 masks and masked_loop_p with a single loop_masks, making sure it's
10134 null for bb vectorization.
10136 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10138 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10139 (vect_analyze_data_ref_access): Use loop->safe_len rather than
10140 loop->force_vectorize to check whether there is no alias.
10142 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10145 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
10147 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
10149 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
10150 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
10151 instead of a loop around prev_real_insn.
10152 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
10155 PR inline-asm/84625
10156 * config/i386/i386.c (ix86_print_operand): Use conditional
10157 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
10160 2018-03-02 Richard Biener <rguenther@suse.de>
10162 PR tree-optimization/84427
10163 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
10164 (bitmap_set_subtract_values): Rewrite to handle multiple
10167 (prune_clobbered_mems): Likewise.
10168 (phi_translate): Take edge instead of pred/phiblock.
10169 (phi_translate_1): Likewise.
10170 (phi_translate_set): Likewise. Insert all translated
10171 exprs for a value into the set, keeping possibly multiple
10172 expressions per value.
10173 (compute_antic_aux): Adjust for phi_translate changes.
10174 When intersecting union the expressions and prune those
10175 not in the final value set, keeping possibly multiple
10176 expressions per value. Do not use value-insertion
10177 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
10178 all expressions. Add verification that the value-sets
10179 only shrink during iteration.
10180 (compute_partial_antic_aux): Adjust for the phi_translate changes.
10181 (do_pre_regular_insertion): Likewise.
10182 (do_pre_partial_partial_insertion): Likewise.
10184 2018-03-02 Richard Biener <rguenther@suse.de>
10187 * config/darwin.c (saved_debug_info_level): New static global.
10188 (darwin_asm_lto_start): Disable debug info generation for LTO out.
10189 (darwin_asm_lto_end): Restore debug info generation settings.
10191 2018-03-01 Martin Liska <mliska@suse.cz>
10194 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
10195 volatile arguments.
10197 2018-03-01 Richard Biener <rguenther@suse.de>
10200 * dwarf2out.c (gen_variable_die): Properly handle late VLA
10201 type annotation with LTO when debug was disabled at compile-time.
10203 2018-03-01 Matthew Fortune <mfortune@gmail.com>
10205 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
10207 (mips_final_postscan_insn): Likewise.
10209 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
10211 PR rtl-optimization/84528
10212 * alias.c (init_alias_target): Add commentary.
10213 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
10214 a unique base value if the frame pointer is not eliminated
10215 to the stack pointer.
10217 2018-03-01 Tom de Vries <tom@codesourcery.com>
10219 PR rtl-optimization/83327
10220 * lra-int.h (hard_regs_spilled_into): Declare.
10221 * lra.c (hard_regs_spilled_into): Define.
10222 (init_reg_info): Init hard_regs_spilled_into.
10223 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
10224 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
10225 (process_bb_lives): Handle hard_regs_spilled_into.
10226 (lra_create_live_ranges_1): Before doing liveness propagation, clear
10227 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
10229 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
10231 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
10232 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
10233 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
10234 * config/rs6000/aix72.h: New file.
10236 2018-02-28 Jakub Jelinek <jakub@redhat.com>
10238 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
10239 instead of warning_at with conditional singular and plural messages
10243 * stor-layout.c (update_alignment_for_field): For
10244 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
10245 && !DECL_PACKED (field), do the alignment update, just use
10246 only desired_align instead of MAX (type_align, desired_align)
10248 (place_field): Don't do known_align < desired_align handling
10249 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
10250 is non-NULL, instead do it after rli->prev_field handling and
10251 only if not within a bitfield word. For DECL_PACKED (field)
10252 use type_align of BITS_PER_UNIT.
10254 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
10256 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
10257 superfluous parentheses and trailing spaces.
10259 2018-02-28 Richard Biener <rguenther@suse.de>
10261 PR tree-optimization/84584
10262 * graphite-scop-detection.c (scop_detection::add_scop): Discard
10263 SCoPs with fake exit edge.
10265 2018-02-28 Martin Liska <mliska@suse.cz>
10268 * timevar.c (timer::print): Fix format to properly print 100%
10271 2018-02-28 Richard Biener <rguenther@suse.de>
10273 PR middle-end/84607
10274 * genmatch.c (capture_info::walk_match): Do not mark
10275 captured expressions without operands as expr_p given
10276 they act more like predicates and should be subject to
10277 "lost tail" side-effect preserving.
10279 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
10281 PR rtl-optimization/81611
10282 * auto-inc-dec.c (attempt_change): Move dead note from
10283 mem_insn if it's the next use of regno
10284 (find_address): Take address use of reg holding
10285 non-incremented value. Add parm to limit search to the named
10287 (merge_in_block): Attempt to use a mem insn that is the next
10288 use of the original regno.
10290 2018-02-27 Martin Sebor <msebor@redhat.com>
10293 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
10294 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
10296 2018-02-27 Martin Sebor <msebor@redhat.com>
10298 PR translation/84207
10299 * diagnostic-core.h (warning_n, error_n, inform_n): Change
10300 n argument to unsigned HOST_WIDE_INT.
10301 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
10302 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
10303 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
10304 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
10306 2018-02-27 Richard Biener <rguenther@suse.de>
10308 PR tree-optimization/84512
10309 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
10310 Do not use the estimate returned from record_stmt_cost for
10311 the scalar iteration cost but sum properly using add_stmt_cost.
10313 2018-02-27 Richard Biener <rguenther@suse.de>
10315 PR tree-optimization/84466
10316 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10317 Adjust last change to less strictly validate use operands.
10319 2018-02-27 Martin Liska <mliska@suse.cz>
10321 PR gcov-profile/84548
10322 * gcov.c (process_file): Allow partial overlap and consider it
10323 also as group functions.
10324 (output_lines): Properly calculate range of lines for a group.
10326 2018-02-27 Martin Liska <mliska@suse.cz>
10328 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
10329 'ggc' suffixes. Change first column width.
10330 (timer::print): Fix formatting of the column.
10332 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
10334 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
10335 preserve inline entry blocks for the sake of debug inline
10336 entry point markers alone.
10337 (remove_unused_locals): Suggest in comments a better place to
10338 force the preservation of inline entry blocks that are
10339 otherwise unused, but do not preserve them.
10341 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10343 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
10345 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10348 * config/i386/constraints.md (Bs): Replace
10349 ix86_indirect_branch_register with
10350 TARGET_INDIRECT_BRANCH_REGISTER.
10352 * config/i386/i386.md (indirect_jump): Likewise.
10353 (tablejump): Likewise.
10354 (*sibcall_memory): Likewise.
10355 (*sibcall_value_memory): Likewise.
10356 Peepholes of indirect call and jump via memory: Likewise.
10357 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
10358 (*sibcall_value_GOT_32): Likewise.
10359 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10360 (GOT_memory_operand): Likewise.
10361 (call_insn_operand): Likewise.
10362 (sibcall_insn_operand): Likewise.
10363 (GOT32_symbol_operand): Likewise.
10364 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
10366 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10368 PR rtl-optimization/83496
10369 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
10370 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
10371 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
10372 redundant insn, if any.
10373 (relax_delay_slots): Likewise.
10374 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
10376 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
10378 PR tree-optimization/83965
10379 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
10380 that grouped statements are part of a reduction chain. Return
10381 true if the statement is not marked as a reduction itself but
10382 is part of a group.
10383 (vect_recog_dot_prod_pattern): Don't check whether the statement
10384 is part of a group here.
10385 (vect_recog_sad_pattern): Likewise.
10386 (vect_recog_widen_sum_pattern): Likewise.
10388 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10391 * final.c (rest_of_clean_state): Also look for calls inside sequences.
10393 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10396 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
10398 (ix86_output_indirect_function_return): New prototype.
10399 (ix86_split_simple_return_pop_internal): Likewise.
10400 * config/i386/i386.c (indirect_return_via_cx): New.
10401 (indirect_return_via_cx_bnd): Likewise.
10402 (indirect_thunk_name): Handle return va CX_REG.
10403 (output_indirect_thunk_function): Create alias for
10404 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
10405 (ix86_output_indirect_jmp): Remove the bool argument.
10406 (ix86_output_indirect_function_return): New function.
10407 (ix86_split_simple_return_pop_internal): Likewise.
10408 * config/i386/i386.md (*indirect_jump): Don't pass false
10409 to ix86_output_indirect_jmp.
10410 (*tablejump_1): Likewise.
10411 (simple_return_pop_internal): Change it to define_insn_and_split.
10412 Call ix86_split_simple_return_pop_internal to split it for
10413 -mfunction-return=.
10414 (simple_return_indirect_internal): Call
10415 ix86_output_indirect_function_return instead of
10416 ix86_output_indirect_jmp.
10418 2018-02-26 Jakub Jelinek <jakub@redhat.com>
10421 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
10422 memset and value initialization afterwards.
10424 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
10426 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
10428 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10431 * common/config/aarch64/aarch64-common.c
10432 (aarch_option_optimization_table[]): Switch
10433 off fomit-frame-pointer
10435 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10436 Chung-Ju Wu <jasonwucj@gmail.com>
10438 * config/nds32/nds32-multiple.md (load_multiple): Disallow
10440 (store_multiple): Ditto.
10442 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10444 * config.gcc: Add --with-cpu support for nds32 target.
10445 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
10446 * config/nds32/nds32.opt: Add -mcpu= option.
10448 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
10450 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
10451 isel=yes): Warn for these deprecated options.
10453 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
10455 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
10456 ISA_2_5_MASKS_EMBEDDED.
10458 2018-02-23 Jakub Jelinek <jakub@redhat.com>
10460 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
10461 p->max as pointers rather than using iterative_hash_expr.
10463 2018-02-23 Carl Love <cel@us.ibm.com>
10465 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
10466 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
10468 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
10469 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
10470 P8V_BUILTIN_VEC_VUNSIGNED2.
10472 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
10475 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
10476 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
10478 * lra-constraints.c (curr_insn_transform): Set up
10479 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
10481 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
10482 * lra-eliminations.c (spill_pseudos): Ditto.
10483 (process_insn_for_elimination): Ditto.
10484 * lra-lives.c (reg_early_clobber_p): Use the new macros.
10485 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
10486 LRA_NON_CLOBBERED_ALT.
10488 2018-02-22 Martin Sebor <msebor@redhat.com>
10490 PR tree-optimization/84480
10491 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
10492 to maybe_diag_stxncpy_trunc. Call it.
10493 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
10494 from gimple_fold_builtin_strcpy. Print inlining stack.
10495 (handle_builtin_stxncpy): Print inlining stack.
10496 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
10498 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
10501 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
10502 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
10503 and -fcheck-pointer-bounds are used together.
10504 (indirect_thunk_prefix): New enum.
10505 (indirect_thunk_need_prefix): New function.
10506 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
10507 "_nt" instead of "_bnd" for NOTRACK prefix.
10508 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
10509 (output_indirect_thunk_function): Likewise.
10511 (ix86_code_end): Update output_indirect_thunk_function calls.
10512 (ix86_output_indirect_branch_via_reg): Replace
10513 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
10514 (ix86_output_indirect_branch_via_push): Likewise.
10515 (ix86_output_function_return): Likewise.
10516 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
10517 incompatible with -fcf-protection=branch and
10518 -fcheck-pointer-bounds.
10520 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10523 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10524 Change gcc_assert call to output_operand_lossage.
10526 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10528 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10530 2018-02-22 DJ Delorie <dj@redhat.com>
10531 Sebastian Perta <sebastian.perta@renesas.com>
10532 Oleg Endo <olegendo@gcc.gnu.org>
10534 * config/rx/rx.c (rx_rtx_costs): New function.
10535 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
10537 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
10539 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
10541 2018-02-22 Martin Liska <mliska@suse.cz>
10544 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
10545 Add "native" as a possible value.
10547 2018-02-22 Martin Liska <mliska@suse.cz>
10550 * config/i386/i386.c (ix86_option_override_internal):
10551 Add "native" as a possible value for -march and -mtune.
10553 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10556 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10557 to all type variants.
10559 PR tree-optimization/84503
10560 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10561 width as info->bitpos + info->bitsize - start.
10562 (merged_store_group::merge_overlapping): Simplify width computation.
10563 (check_no_overlap): New function.
10564 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10565 start + width and last_order of the group, fail if check_no_overlap
10567 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10568 to group if check_no_overlap fails.
10570 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10572 * config/rs6000/altivec.md: Delete contraint arguments to
10573 define_expand, define_split, and define_peephole2, and in
10574 define_insn_and_split if always unused.
10575 * config/rs6000/darwin.md: Ditto.
10576 * config/rs6000/dfp.md: Ditto.
10577 * config/rs6000/rs6000.md: Ditto.
10578 * config/rs6000/sync.md: Ditto.
10579 * config/rs6000/vector.md: Ditto.
10580 * config/rs6000/vsx.md: Ditto.
10582 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10584 * config/rs6000/altivec.md: Write output control strings as braced
10585 blocks instead of double-quoted strings.
10586 * config/rs6000/darwin.md: Ditto.
10587 * config/rs6000/rs6000.md: Ditto.
10588 * config/rs6000/vector.md: Ditto.
10589 * config/rs6000/vsx.md: Ditto.
10591 2018-02-21 Jason Merrill <jason@redhat.com>
10593 PR c++/84314 - ICE with templates and fastcall attribute.
10594 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10596 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10598 * ipa-cp.c (determine_versionability): Fix comment typos.
10600 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10603 * ipa-cp.c (determine_versionability): Do not version functions caling
10606 2018-02-21 Martin Liska <mliska@suse.cz>
10609 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10610 Add "native" as a possible value.
10611 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10612 the macro when native cpu detection is available.
10614 2018-02-21 Martin Liska <mliska@suse.cz>
10617 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10618 Add "native" as a possible value.
10619 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10620 when native cpu detection is available.
10622 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10623 Martin Sebor <msebor@redhat.com>
10625 PR tree-optimization/84478
10626 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10628 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10629 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10630 support which is conservatively correct, for 2 only stay conservative
10631 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10632 argument to the 2 argument get_range_strlen, adjust 6 arg
10633 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10635 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10636 (gimple_fold_builtin_strlen): Pass true as last argument to
10639 2018-02-20 Martin Sebor <msebor@redhat.com>
10641 PR middle-end/84095
10642 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10643 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10644 (builtin_memref::builtin_memref): Factor out parts into
10645 set_base_and_offset and call it.
10647 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10649 PR middle-end/84406
10650 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10651 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10652 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10653 search at the associated MODE_INT.
10655 2018-02-20 Jeff Law <law@redhat.com>
10657 PR middle-end/82123
10658 PR tree-optimization/81592
10659 PR middle-end/79257
10660 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10661 for range data rather than using global data.
10662 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10663 range data rather than using global data.
10664 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10665 pass it to children as needed.
10666 (struct directive::fmtresult): Similarly.
10667 (struct directive::set_width): Similarly.
10668 (struct directive::set_precision): Similarly.
10669 (format_integer, format_directive, parse_directive): Similarly.
10670 (format_none): Accept unnamed vr_values parameter.
10671 (format_percent, format_floating, format_character): Similarly.
10672 (format_string, format_plain): Similarly.
10673 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10674 the EVRP range analyzer for range data rather than using global data.
10675 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10676 gimple-ssa-evrp-analyze.h
10677 (class sprintf_dom_walker): Add after_dom_children member function.
10678 Add evrp_range_analyzer member.
10679 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10680 range analyzer as needed.
10681 (sprintf_dom_walker::after_dom_children): New member function.
10682 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10684 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10685 (evrp_range_analyzer::pop_to_marker): Likewise.
10687 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10689 PR tree-optimization/84419
10690 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10691 with the required type if its current type is compatible but
10694 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10696 PR middle-end/82004
10697 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10698 after vectorization.
10700 2018-02-20 Martin Liska <mliska@suse.cz>
10703 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10704 possible values if we don't have a hint.
10706 2018-02-20 Martin Liska <mliska@suse.cz>
10710 * final.c (shorten_branches): Build align_tab array with one
10712 * opts.c (finish_options): Add alignment option limit check.
10713 (MAX_CODE_ALIGN): Likewise.
10714 (MAX_CODE_ALIGN_VALUE): Likewise.
10715 * doc/invoke.texi: Document maximum allowed option value for
10716 all -falign-* options.
10718 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10721 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10722 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10723 * var-tracking.c (emit_note_insn_var_location): Remove all references
10724 to NOTE_INSN_CALL_ARG_LOCATION.
10725 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10726 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10727 Use copy_rtx_if_shared.
10728 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10729 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10730 (dwarf2out_var_location): Remove handling of
10731 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10733 * final.c (final_scan_insn): Remove all references to
10734 NOTE_INSN_CALL_ARG_LOCATION.
10735 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10736 before dumping final insns.
10737 * except.c (emit_note_eh_region_end): Remove all references to
10738 NOTE_INSN_CALL_ARG_LOCATION.
10739 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10740 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10741 * config/arc/arc.c (hwloop_optimize): Likewise.
10742 * config/arm/arm.c (create_fix_barrier): Likewise.
10743 * config/s390/s390.c (s390_chunkify_start): Likewise.
10744 * config/sh/sh.c (find_barrier): Likewise.
10745 * config/i386/i386.c (rest_of_insert_endbranch,
10746 ix86_seh_fixup_eh_fallthru): Likewise.
10747 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10748 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10749 * config/frv/frv.c (frv_function_prologue): Likewise.
10750 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10752 (note_outside_basic_block_p): Remove all references to
10753 NOTE_INSN_CALL_ARG_LOCATION.
10754 * gengtype.c (adjust_field_rtx_def): Likewise.
10755 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10757 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10758 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10761 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10764 PR tree-optimization/84452
10765 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10766 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10769 2018-02-19 Martin Liska <mliska@suse.cz>
10772 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10774 2018-02-19 Martin Liska <mliska@suse.cz>
10775 Richard Sandiford <richard.sandiford@linaro.org>
10777 PR tree-optimization/82491
10778 * gimple-fold.c (get_base_constructor): Make earlier bail out
10781 2018-02-19 Carl Love <cel@us.ibm.com>
10783 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10784 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10786 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10787 P8V_BUILTIN_VEC_NEG.
10789 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10791 * config/rl78/rl78.md (movdf): New define expand.
10793 2018-02-19 Martin Liska <mliska@suse.cz>
10796 * doc/invoke.texi: Fix typo.
10797 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10799 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10801 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10802 handle rs6000_single_float and rs6000_double_float specially for
10805 2018-02-16 Jeff Law <law@redhat.com>
10807 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10808 (pop_regs): Use it.
10810 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10813 * ipa-inline.c (inline_small_functions): Fix a typo.
10815 2018-02-16 Nathan Sidwell <nathan@acm.org>
10817 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10819 2018-02-16 Carl Love <cel@us.ibm.com>
10821 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10822 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10823 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10824 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10825 expansion to P8V_BUILTIN_VEC_FLOAT2.
10827 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10829 PR rtl-optimization/70023
10830 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10831 src_regno into account.
10833 2018-02-16 Carl Love <cel@us.ibm.com>
10835 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10836 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10837 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10838 * config/rs6000/rs6000.c: Remove case statements for
10839 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10840 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10841 and P9V_BUILTIN_VEC_VINSERT4B.
10842 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10843 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10844 * config/rs6000/vsx.md:
10845 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10848 2018-02-16 Carl Love <cel@us.ibm.com>
10850 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10852 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10854 * config/rs6000/rs6000-c.c: Add the definitions for
10855 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10856 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10857 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10858 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10859 definition for insert4b and define insn *insert3b_internal.
10860 * doc/extend.texi: Add documentation for vec_extract4b.
10862 2018-02-16 Nathan Sidwell <nathan@acm.org>
10864 * doc/extend.texi (Backwards Compatibility): Mention friend
10865 injection. Note for-scope is deprecated.
10866 * doc/invoke.texi (-ffriend-injection): Deprecate.
10868 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10870 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10871 that moved to I2, also allow destinations that are a paradoxical
10872 subreg (instead of a normal reg).
10874 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10877 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10880 2018-02-16 Richard Biener <rguenther@suse.de>
10882 PR tree-optimization/84037
10883 PR tree-optimization/84016
10885 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10886 Adjust vec_construct for the fact we need additional higher latency
10887 128bit inserts for AVX256 and AVX512 vector builds.
10888 (ix86_add_stmt_cost): Scale vector construction cost for
10891 2018-02-16 Richard Biener <rguenther@suse.de>
10893 PR tree-optimization/84417
10894 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10895 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10896 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10898 2018-02-16 Martin Liska <mliska@suse.cz>
10901 * internal-fn.def (ASAN_CHECK): Set proper flags.
10902 (ASAN_MARK): Likewise.
10904 2018-02-16 Julia Koval <julia.koval@intel.com>
10906 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10907 from PTA_CANNONLAKE.
10909 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10912 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10913 Use ++iter rather than iter++ for std::list iterators.
10914 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10915 defer deleting them until all nodes in the forest are processed. Do
10916 free even leaf nodes. Change to_process into auto_vec.
10919 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10920 * vec.h (vec_default_construct): Use memset instead of placement new
10921 if BROKEN_VALUE_INITIALIZATION is defined.
10922 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10923 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10926 PR rtl-optimization/83723
10927 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10928 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10929 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10931 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10933 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10935 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10937 PR rtl-optimization/81443
10938 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10939 from inner REGs to paradoxical SUBREGs.
10941 2018-02-16 Richard Biener <rguenther@suse.de>
10943 PR tree-optimization/84399
10944 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10945 For operands we can analyze at their definition make sure we can
10946 analyze them at each use as well.
10948 2018-02-16 Richard Biener <rguenther@suse.de>
10950 PR tree-optimization/84190
10951 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10952 volatile accesses if the decl isn't volatile.
10954 2018-02-15 Jason Merrill <jason@redhat.com>
10956 PR c++/84314 - ICE with templates and fastcall attribute.
10957 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10958 TYPE_CANONICAL on an existing type.
10960 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10962 PR tree-optimization/84383
10963 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10964 dstoff nor call operand_equal_p if dstbase is NULL.
10966 PR tree-optimization/84334
10967 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10968 also a CONSTANT_CLASS_P, punt.
10970 2018-02-14 Jim Wilson <jimw@sifive.com>
10972 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10973 first SMALL_OPERAND check. New local min_second_step. Move assert
10974 to where locals are set. Add TARGET_RVC support.
10975 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10977 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10979 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10981 2018-02-14 Martin Sebor <msebor@redhat.com>
10983 PR tree-optimization/83698
10984 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10985 arrays constrain the offset range to their bounds.
10986 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10987 (builtin_access::overlap): Avoid setting the size of overlap if it's
10989 (maybe_diag_overlap): Also consider arrays when deciding what values
10990 of offsets to include in diagnostics.
10992 2018-02-14 Martin Sebor <msebor@redhat.com>
10995 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10996 that correspond to the kind of a declaration.
10998 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
11001 * config/pa/pa.md: Load address of PIC label using the linkage table
11002 if the label is nonlocal.
11004 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
11006 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
11007 warning message if user requests -maltivec=be.
11008 * doc/invoke.texi: Document deprecation of -maltivec=be.
11010 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
11013 * config/rs6000/rs6000-c.c: Update definitions for
11014 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
11015 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
11017 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11020 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
11021 add _get_ssp intrinsics. Remove argument from
11022 __builtin_ia32_rdssp[d|q].
11023 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
11024 * config/i386/i386-builtin.def: Remove argument from
11025 __builtin_ia32_rdssp[d|q].
11026 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
11027 ix86_expand_special_args_builtin for _rdssp[d|q].
11028 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
11029 Clear register before usage.
11030 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
11031 Add documentation for new _get_ssp and _inc_ssp intrinsics.
11033 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
11035 PR tree-optimization/84357
11036 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
11037 operand 1 of an ARRAY_REF too.
11039 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
11042 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
11043 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11045 (set_of_reg): New struct.
11046 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
11047 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
11048 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11050 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
11051 Split into bitclr, bitset, bitinvert patterns if appropriate.
11052 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
11053 use rx_fuse_in_memory_bitop.
11054 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
11055 to named insn, correct maximum insn length.
11057 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
11060 * machmode.def: Define a complex mode for PARTIAL_INT.
11061 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
11063 * doc/rtl.texi: Document CSPImode.
11064 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
11066 (msp430_hard_regno_nregs_with_padding): Likewise.
11068 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
11071 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
11073 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
11075 PR rtl-optimization/84169
11076 * combine.c (try_combine): New variable split_i2i3. Set it to true if
11077 we generated a parallel as new i3 and we split that to new i2 and i3
11078 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
11079 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
11080 those to i2, not i1. Partially rewrite this scan code.
11082 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11085 * stor-layout.c (place_field): For variable length fields, adjust
11086 offset_align afterwards not just based on the field's alignment,
11087 but also on the size.
11089 PR middle-end/84309
11090 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
11091 of exps and logs in the use_exp2 case.
11093 2018-02-13 Jeff Law <law@redhat.com>
11095 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
11096 entry for "vector".
11098 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
11101 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
11105 * common.opt (gas-loc-support, gas-locview-support): New.
11106 (ginline-points, ginternal-reset-location-views): New.
11107 * doc/invoke.texi: Document them. Use @itemx where intended.
11108 (gvariable-location-views): Adjust.
11109 * target.def (reset_location_view): New.
11110 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
11111 (TARGET_RESET_LOCATION_VIEW): New.
11112 * doc/tm.texi: Rebuilt.
11113 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
11114 (dwarf2out_default_as_locview_support): New.
11115 (output_asm_line_debug_info): Use option variables.
11116 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
11117 (output_loc_list): Likewise.
11118 (add_high_low_attributes): Check option variables.
11119 Don't output entry view attribute in strict mode.
11120 (gen_inlined_subroutine_die): Check option variables.
11121 (dwarf2out_inline_entry): Likewise.
11122 (init_sections_and_labels): Likewise.
11123 (dwarf2out_early_finish): Likewise.
11124 (maybe_reset_location_view): New, from...
11125 (dwarf2out_var_location): ... here. Call it.
11126 * debug.h (dwarf2out_default_as_loc_support): Declare.
11127 (dwarf2out_default_as_locview_support): Declare.
11128 * hooks.c (hook_int_rtx_insn_0): New.
11129 * hooks.h (hook_int_rtx_insn_0): Declare.
11130 * toplev.c (process_options): Take -gas-loc-support and
11131 -gas-locview-support from dwarf2out. Enable
11132 -gvariable-location-views by default only with locview
11133 assembler support. Enable -ginternal-reset-location-views by
11134 default only if the target defines the corresponding hook.
11135 Enable -ginline-points by default if location views are
11136 enabled; force it disabled if statement frontiers are
11138 * tree-inline.c (expand_call_inline): Check option variables.
11139 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
11141 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
11143 PR tree-optimization/84321
11144 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
11145 handling. Also check whether the anti-range contains any values
11146 that satisfy the mask; switch to a VR_RANGE if not.
11148 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
11151 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
11153 2018-02-13 Martin Jambor <mjambor@suse.cz>
11156 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
11157 of call statements, also set location of a load to a temporary.
11159 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
11161 * config/rl78/rl78.c (add_vector_labels): New function.
11162 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
11163 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
11164 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
11165 which checks that no arguments are passed.
11166 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
11167 * doc/extend.texi: Documentation for the new attribute.
11169 2018-02-13 Andreas Schwab <schwab@suse.de>
11171 * config/riscv/linux.h (CPP_SPEC): Define.
11173 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11176 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
11177 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
11178 OPTION_MASK_ISA_AES as first argument to def_builtin_const
11179 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
11180 instead of OPTION_MASK_ISA_PCLMUL as first argument to
11181 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
11182 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
11183 temporarily for AES and PCLMUL builtins.
11185 PR tree-optimization/84339
11186 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
11187 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
11190 PR middle-end/84309
11191 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
11192 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
11193 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
11195 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
11197 * omp-simd-clone.h: New file.
11198 * omp-simd-clone.c: Include omp-simd-clone.h.
11199 (expand_simd_clones): No longer static.
11200 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
11201 cgraph.h and omp-simd-clone.h.
11202 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
11203 (vect_recog_widen_shift_pattern): Formatting fix.
11204 (vect_pattern_recog_1): Don't check optab for calls.
11207 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
11208 operands[2] into a REG before using gen_lowpart on it.
11210 2018-02-12 Jeff Law <law@redhat.com>
11213 * config/sh/sh.c (find_barrier): Consider a sibling call
11216 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
11217 successfully back substituting a reg.
11219 2018-02-12 Richard Biener <rguenther@suse.de>
11221 PR tree-optimization/84037
11222 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
11223 parameter, move visited init to caller.
11224 (vect_slp_analyze_operations): Separate cost from validity
11225 check, initialize visited once for all instances.
11226 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
11228 * tree-vect-stmts.c (vect_model_simple_cost): Make early
11230 (vect_model_promotion_demotion_cost): Likewise.
11231 (vectorizable_bswap): Guard cost modeling with !slp_node
11232 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
11234 (vectorizable_call): Likewise.
11235 (vectorizable_conversion): Likewise.
11236 (vectorizable_assignment): Likewise.
11237 (vectorizable_shift): Likewise.
11238 (vectorizable_operation): Likewise.
11239 (vectorizable_store): Likewise.
11240 (vectorizable_load): Likewise.
11241 (vectorizable_condition): Likewise.
11242 (vectorizable_comparison): Likewise.
11244 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
11247 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
11248 (ASAN_MARK): Fix fnspec to account for return value, change pointer
11249 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
11251 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11253 PR middle-end/83665
11254 * params.def (inline-min-speedup): Increase from 8 to 15.
11255 (max-inline-insns-auto): Decrease from 40 to 30.
11256 * ipa-split.c (consider_split): Add some buffer for function to
11257 be considered inlining candidate.
11258 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
11261 2018-02-12 Richard Biener <rguenther@suse.de>
11263 PR tree-optimization/84037
11264 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
11265 matched stmts if we cannot swap the non-matched ones.
11267 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11269 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
11270 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
11271 _mm_maskz_scalef_round_ss): New intrinsics.
11272 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
11273 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
11274 __builtin_ia32_scalefss_round): Remove.
11275 (__builtin_ia32_scalefsd_mask_round,
11276 __builtin_ia32_scalefss_mask_round): New intrinsics.
11277 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
11278 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
11279 ((match_operand:VF_128 2 "<round_nimm_predicate>"
11280 "<round_constraint>")): Changed to ...
11281 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
11282 "<round_scalar_constraint>")): ... this.
11283 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
11284 %0, %1, %2<round_op3>}"): Changed to ...
11285 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
11286 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
11287 %2<round_scalar_mask_op3>}"): ... this.
11288 * config/i386/subst.md (round_scalar_nimm_predicate): New.
11290 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11292 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
11293 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
11294 (_mm_maskz_sqrt_round_ss): New intrinsics.
11295 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
11296 (__builtin_ia32_sqrtsd_mask_round)
11297 (__builtin_ia32_sqrtss_mask_round): New builtins.
11298 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
11299 (__builtin_ia32_sqrtss_round): Remove.
11300 (__builtin_ia32_sqrtsd_mask_round)
11301 (__builtin_ia32_sqrtss_mask_round): New builtins.
11302 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
11303 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
11304 ((match_operand:VF_128 1 "vector_operand"
11305 "xBm,<round_constraint>")): Changed to ...
11306 ((match_operand:VF_128 1 "vector_operand"
11307 "xBm,<round_scalar_constraint>")): ... this.
11308 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
11309 %0, %2, %<iptr>1<round_op3>}): Changed to ...
11310 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
11311 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
11312 %<iptr>1<round_scalar_mask_op3>}): ... this.
11313 ((set_attr "prefix" "<round_prefix>")): Changed to ...
11314 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
11316 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
11319 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
11320 Cast vec_cmpeq result to correct type.
11321 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
11322 Cast vec_cmpgt result to correct type.
11324 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
11326 * final.c (final_scan_insn_1): Renamed from...
11327 (final_scan_insn): ... this. New wrapper, to recover
11328 seen from the outermost call in recursive ones.
11329 * config/sparc/sparc.c (output_return): Drop seen from call.
11330 (output_sibcall): Likewise.
11331 * config/visium/visium.c (output_branch): Likewise.
11333 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
11335 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
11338 2018-02-10 Alan Modra <amodra@gmail.com>
11341 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
11342 Specify LR as an input.
11344 2018-02-10 Jakub Jelinek <jakub@redhat.com>
11347 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
11348 remove_member_access_dummy_vars): New functions.
11349 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
11350 lower_omp_1, execute_lower_omp): Use them.
11352 PR rtl-optimization/84308
11353 * shrink-wrap.c (spread_components): Release todo vector.
11355 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
11357 PR rtl-optimization/57193
11358 * ira-color.c (struct allocno_color_data): Add member
11359 conflict_allocno_hard_prefs.
11360 (update_conflict_allocno_hard_prefs): New.
11361 (bucket_allocno_compare_func): Add a preference based on
11362 conflict_allocno_hard_prefs.
11363 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
11364 (color_allocnos): Remove a dead code. Initiate
11365 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
11367 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11370 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
11371 constraint from =wa to wa. Avoid a subreg on the output operand,
11372 instead use a pseudo and subreg it in a move.
11373 (p9_xxbrd_<mode>): Changed to ...
11374 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
11375 (p9_xxbrd_v2df): New expander.
11376 (p9_xxbrw_<mode>): Changed to ...
11377 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
11378 (p9_xxbrw_v4sf): New expander.
11380 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11382 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
11384 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
11387 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
11388 multiply in 32-bit mode.
11389 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
11390 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
11393 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11395 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
11396 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
11397 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
11398 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
11400 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
11403 * dwarf2out.c (is_trivial_indirect_ref): New function.
11404 (dwarf2out_late_global_decl): Do not generate a location
11405 attribute for variables that have a non-trivial DECL_VALUE_EXPR
11406 and that are not defined in the current unit.
11408 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11410 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
11411 instead of a libcall for UNORDERED.
11413 2018-02-09 Tamar Christina <tamar.christina@arm.com>
11416 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
11417 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
11419 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11422 * config/s390/s390.c (s390_set_current_function): Invoke
11423 s390_indirect_branch_settings also if fndecl didn't change.
11425 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11427 * config/rs6000/rs6000.md (blockage): Set length to zero.
11429 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11431 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
11433 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11436 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
11437 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
11441 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
11442 PARALLEL incoming that failed vt_get_decl_and_offset check.
11444 PR middle-end/84237
11445 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
11446 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
11448 (get_variable_section): For decls in named .bss* sections pass true as
11449 second argument to bss_initializer_p.
11451 2018-02-09 Marek Polacek <polacek@redhat.com>
11452 Jakub Jelinek <jakub@redhat.com>
11455 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
11456 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
11457 Sync some changes from cxx_fold_indirect_ref.
11459 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11461 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
11463 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
11464 (BLOCK_INLINE_ENTRY_LABEL): New.
11465 (dwarf2out_var_location): Disregard inline entry markers.
11466 (inline_entry_data): New struct.
11467 (inline_entry_data_hasher): New hashtable type.
11468 (inline_entry_data_hasher::hash): New.
11469 (inline_entry_data_hasher::equal): New.
11470 (inline_entry_data_table): New variable.
11471 (add_high_low_attributes): Add DW_AT_entry_pc and
11472 DW_AT_GNU_entry_view attributes if a pending entry is found
11473 in inline_entry_data_table. Add old entry_pc attribute only
11474 if debug nonbinding markers are disabled.
11475 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
11476 markers are enabled.
11477 (block_within_block_p, dwarf2out_inline_entry): New.
11478 (dwarf2out_finish): Check that no entries remained in
11479 inline_entry_data_table.
11480 * final.c (reemit_insn_block_notes): Handle inline entry notes.
11481 (final_scan_insn, notice_source_line): Likewise.
11482 (rest_of_clean_state): Skip inline entry markers.
11483 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
11485 * gimple.c (gimple_build_debug_inline_entry): New.
11486 * gimple.h (enum gimple_debug_subcode): Add
11487 GIMPLE_DEBUG_INLINE_ENTRY.
11488 (gimple_build_debug_inline_entry): Declare.
11489 (gimple_debug_inline_entry_p): New.
11490 (gimple_debug_nonbind_marker_p): Adjust.
11491 * insn-notes.def (INLINE_ENTRY): New.
11492 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
11493 inline entry marker notes.
11494 (print_insn): Likewise.
11495 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
11496 (INSN_DEBUG_MARKER_KIND): Likewise.
11497 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
11498 * tree-inline.c (expand_call_inline): Build and insert
11499 debug_inline_entry stmt.
11500 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
11501 inline entry blocks early, if nonbind markers are enabled.
11502 (dump_scope_block): Dump fragment info.
11503 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
11504 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
11505 (gimple_build_debug_inline_entry): New.
11506 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
11507 Enable/disable inline entry points too.
11508 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
11509 (DEBUG_INSN): Describe inline entry markers.
11511 * common.opt (gvariable-location-views): New.
11512 (gvariable-location-views=incompat5): New.
11513 * config.in: Rebuilt.
11514 * configure: Rebuilt.
11515 * configure.ac: Test assembler for view support.
11516 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
11517 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
11518 * dwarf2out.c (var_loc_view): New typedef.
11519 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11520 (dwarf2out_locviews_in_attribute): New.
11521 (dwarf2out_locviews_in_loclist): New.
11522 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11523 (enum dw_line_info_opcode): Add LI_adv_address.
11524 (struct dw_line_info_table): Add view.
11525 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11526 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11527 (zero_view_p): New variable.
11528 (ZERO_VIEW_P): New macro.
11529 (output_asm_line_debug_info): New.
11530 (struct var_loc_node): Add view.
11531 (add_AT_view_list, AT_loc_list): New.
11532 (add_var_loc_to_decl): Add view param. Test it against last.
11533 (new_loc_list): Add view params. Record them.
11534 (AT_loc_list_ptr): Handle loc and view lists.
11535 (view_list_to_loc_list_val_node): New.
11536 (print_dw_val): Handle dw_val_class_view_list.
11537 (size_of_die): Likewise.
11538 (value_format): Likewise.
11539 (loc_list_has_views): New.
11540 (gen_llsym): Set vl_symbol too.
11541 (maybe_gen_llsym, skip_loc_list_entry): New.
11542 (dwarf2out_maybe_output_loclist_view_pair): New.
11543 (output_loc_list): Output view list or entries too.
11544 (output_view_list_offset): New.
11545 (output_die): Handle dw_val_class_view_list.
11546 (output_dwarf_version): New.
11547 (output_compilation_unit_header): Use it.
11548 (output_skeleton_debug_sections): Likewise.
11549 (output_rnglists, output_line_info): Likewise.
11550 (output_pubnames, output_aranges): Update version comments.
11551 (output_one_line_info_table): Output view numbers in asm comments.
11552 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11553 Call maybe_gen_llsym.
11554 (loc_list_from_tree_1): Adjust.
11555 (add_AT_location_description): Create view list attribute if
11556 needed, check it's absent otherwise.
11557 (convert_cfa_to_fb_loc_list): Adjust.
11558 (maybe_emit_file): Call output_asm_line_debug_info for test.
11559 (dwarf2out_var_location): Reset views as needed. Precompute
11560 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11561 attribute. Set view.
11562 (new_line_info_table): Reset next view.
11563 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11564 (dwarf2out_source_line): Likewise. Output view resets and labels to
11565 the assembler, or select appropriate line info opcodes.
11566 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11567 (optimize_string_length): Catch it. Adjust.
11568 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11569 dw_val_class_view_list, and remove it if no longer needed.
11570 (hash_loc_list): Hash view numbers.
11571 (loc_list_hasher::equal): Compare them.
11572 (optimize_location_lists): Check whether a view list symbol is
11573 needed, and whether the locview attribute is present, and
11574 whether they match. Remove the locview attribute if no longer
11576 (index_location_lists): Call skip_loc_list_entry for test.
11577 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11578 Use output_dwarf_version.
11579 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11580 (struct dw_val_node): Add val_view_list.
11581 * final.c (SEEN_NEXT_VIEW): New.
11582 (set_next_view_needed): New.
11583 (clear_next_view_needed): New.
11584 (maybe_output_next_view): New.
11585 (final_start_function): Rename to...
11586 (final_start_function_1): ... this. Take pointer to FIRST,
11587 add SEEN parameter. Emit param bindings in the initial view.
11588 (final_start_function): Reintroduce SEEN-less interface.
11589 (final): Rename to...
11590 (final_1): ... this. Take SEEN parameter. Output final pending
11591 next view at the end.
11592 (final): Reintroduce seen-less interface.
11593 (final_scan_insn): Output pending next view before switching
11594 sections or ending a block. Mark the next view as needed when
11595 outputting variable locations. Notify debug backend of section
11596 changes, and of location view changes.
11597 (rest_of_handle_final): Adjust.
11598 * toplev.c (process_options): Autodetect value for debug variable
11599 location views option. Warn on incompat5 without -gdwarf-5.
11600 * doc/invoke.texi (gvariable-location-views): New.
11601 (gvariable-location-views=incompat5): New.
11602 (gno-variable-location-views): New.
11604 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11606 PR tree-optimization/84136
11607 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11608 that the result of find_edge is non-NULL.
11610 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11613 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11614 storing integer register in SImode. Fix cost of 256 and 512
11615 byte aligned SSE register store.
11617 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11619 * config/i386/i386.c (ix86_multiplication_cost): Fix
11620 multiplication cost for TARGET_AVX512DQ.
11622 2018-02-08 Marek Polacek <polacek@redhat.com>
11624 PR tree-optimization/84238
11625 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11628 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11630 PR tree-optimization/84265
11631 * tree-vect-stmts.c (vectorizable_store): Don't treat
11632 VMAT_CONTIGUOUS accesses as grouped.
11633 (vectorizable_load): Likewise.
11635 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11637 PR tree-optimization/81635
11638 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11639 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11640 (test_round_for_mask): New functions.
11641 (wide_int_cc_tests): Call test_round_for_mask.
11642 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11643 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11644 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11645 range returned by get_range_info.
11647 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11650 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11651 * symtab.c: Include builtins.h
11652 (symtab_node::output_to_lto_symbol_table_p): Move here
11653 from lto-streamer-out.c:output_symbol_p.
11654 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11655 (output_symbol_p): Move all logic to symtab.c
11656 (produce_symtab): Update.
11658 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11660 * config/s390/s390-opts.h (enum indirect_branch): Define.
11661 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11662 (s390_indirect_branch_via_thunk)
11663 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11664 (enum s390_indirect_branch_type): Define.
11665 * config/s390/s390.c (struct s390_frame_layout, struct
11666 machine_function): Remove.
11667 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11668 (indirect_branch_table_label_no, indirect_branch_table_name):
11670 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11671 (enum s390_indirect_branch_option): Define.
11672 (s390_return_addr_from_memory): New function.
11673 (s390_handle_string_attribute): New function.
11674 (s390_attribute_table): Add new attribute handler.
11675 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11676 (s390_indirect_branch_via_thunk): New function.
11677 (s390_indirect_branch_via_inline_thunk): New function.
11678 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11679 sibling call optimization for non z10 compiles.
11680 (s390_emit_call): Force indirect branch target to be a single
11681 register. Add r1 clobber for non-z10 compiles.
11682 (s390_emit_epilogue): Emit return jump via return_use expander.
11683 (s390_reorg): Handle JUMP_INSNs as execute targets.
11684 (s390_option_override_internal): Perform validity checks for the
11685 new command line options.
11686 (s390_indirect_branch_attrvalue): New function.
11687 (s390_indirect_branch_settings): New function.
11688 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11689 (s390_output_indirect_thunk_function): New function.
11690 (s390_code_end): Implement target hook.
11691 (s390_case_values_threshold): Implement target hook.
11692 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11694 * config/s390/s390.h (struct s390_frame_layout)
11695 (struct machine_function): Move here from s390.c.
11696 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11697 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11698 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11699 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11700 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11701 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11702 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11703 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11704 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11705 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11706 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11707 (mnemonic attribute): Add values which aren't recognized
11709 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11710 pattern for branch conversion. Fix mnemonic attribute.
11711 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11712 indirect branch via thunk if requested.
11713 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11714 ("*indirect_jump"): Disable for branch conversion using out of
11716 ("indirect_jump_via_thunk<mode>_z10")
11717 ("indirect_jump_via_thunk<mode>")
11718 ("indirect_jump_via_inlinethunk<mode>_z10")
11719 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11720 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11721 ("casesi_jump_via_inlinethunk<mode>_z10")
11722 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11723 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11724 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11725 ("*indirect2_jump"): Disable for branch conversion.
11726 ("casesi_jump"): Turn into expander and expand patterns for branch
11728 ("return_use"): New expander.
11729 ("*return"): Emit return via thunk and rename it to ...
11730 ("*return<mode>"): ... this one.
11731 * config/s390/s390.opt: Add new options and and enum for the
11734 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11736 * lra-constraints.c (match_reload): Unconditionally use
11737 gen_lowpart_SUBREG, rather than selecting between that
11738 and equivalent gen_rtx_SUBREG code.
11740 2018-02-08 Richard Biener <rguenther@suse.de>
11742 PR tree-optimization/84233
11743 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11744 changed flag instead of boguously re-using phi_inserted.
11746 2018-02-08 Martin Jambor <mjambor@suse.cz>
11748 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11749 static local variables.
11751 2018-02-08 Richard Biener <rguenther@suse.de>
11753 PR tree-optimization/84278
11754 * tree-vect-stmts.c (vectorizable_store): When looking for
11755 smaller vector types to perform grouped strided loads/stores
11756 make sure the mode is supported by the target.
11757 (vectorizable_load): Likewise.
11759 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11761 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11762 Increase LDP/STP opportunities by adding adjacent callee-saves.
11764 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11766 PR rtl-optimization/84068
11767 PR rtl-optimization/83459
11768 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11770 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11772 PR tree-optimization/84224
11773 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11774 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11775 non-zero arguments.
11777 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11780 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11781 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11782 count, remove one USE.
11784 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11786 * doc/install.texi (Configuration): Document the
11787 --with-long-double-format={ibm,ieee} PowerPC configuration
11791 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11792 Convert from define_expand to be define_insn_and_split. Rework
11793 float/double/_Float128 conversions to QI/HI/SImode to work with
11794 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11795 conversions to QI/HImode types did a store and then a load to
11796 truncate the value. For conversions to VSX registers, don't split
11797 the insn, instead emit the code directly. Use the code iterator
11798 any_fix to combine signed and unsigned conversions.
11799 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11800 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11801 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11802 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11803 (fix_<mode>di2_hw): Likewise.
11804 (fixuns_<mode>di2_hw): Likewise.
11805 (fix_<mode>si2_hw): Likewise.
11806 (fixuns_<mode>si2_hw): Likewise.
11807 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11808 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11809 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11810 fix<uns>_trunc<SFDF:mode>si2_p8.
11811 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11812 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11813 (fix<uns>_<mode>_mem): Likewise.
11814 (fctiw<u>z_<mode>_mem): Likewise.
11815 (fix<uns>_<mode>_mem): Likewise.
11816 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11817 the register allocator from doing a direct move to the GPRs to do
11818 a store, and instead use the ISA 3.0 store byte/half-word from
11819 vector register instruction. For IEEE 128-bit floating point,
11820 also optimize stores of 32-bit ints.
11821 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11823 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11825 * genextract.c (push_pathstr_operand): New function to support
11827 (walk_rtx): Call push_pathstr_operand.
11828 (print_path): Support [a-zA-Z].
11830 2018-02-07 Richard Biener <rguenther@suse.de>
11832 PR tree-optimization/84037
11833 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11834 (cse_and_gimplify_to_preheader): Declare.
11835 (vect_get_place_in_interleaving_chain): Likewise.
11836 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11838 (_loop_vec_info::~_loop_vec_info): Delete it.
11839 (cse_and_gimplify_to_preheader): New function.
11840 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11841 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11842 (vectorizable_load): Likewise. For grouped stores always base
11843 the IV on the first element.
11844 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11845 condition before gimplifying.
11847 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11849 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11850 *DIV_EXPR and *MOD_EXPR.
11852 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11855 * config/i386/i386.c (ix86_option_override_internal): Mask out
11856 the CF_SET bit when checking -fcf-protection.
11858 2018-02-07 Tom de Vries <tom@codesourcery.com>
11861 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11864 2018-02-07 Richard Biener <rguenther@suse.de>
11866 PR tree-optimization/84204
11867 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11870 PR tree-optimization/84205
11871 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11872 special-case isl_ast_op_zdiv_r.
11874 PR tree-optimization/84223
11875 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11876 Only add conditions from within the region.
11877 (gather_bbs::after_dom_children): Adjust.
11879 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11882 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11883 * config/avr/avr.md: Only post-reload split REG-REG moves if
11884 either register is GENERAL_REG_P.
11886 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11888 PR tree-optimization/84235
11889 * tree-ssa-scopedtables.c
11890 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11891 if the subtraction is performed in floating point type where NaNs are
11892 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11893 build 1. Formatting fix.
11895 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11898 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11899 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11900 and skip it regardless of bb boundaries. Use CALL_P macro,
11901 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11902 unnecessarily, formatting fix.
11904 2018-02-06 Michael Collison <michael.collison@arm.com>
11906 * config/arm/thumb2.md:
11907 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11908 (*thumb_mov_notscc): Ditto.
11910 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11913 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11914 unsigned_fix, not "s".
11916 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11918 * configure.ac (gcc_fn_eh_frame_ro): New function.
11919 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11920 correct .eh_frame permissions.
11921 * configure: Regenerate.
11923 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11925 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11926 irrelevant options.
11928 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11930 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11931 Display warning message for -mno-speculate-indirect-jumps.
11933 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11935 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11937 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11939 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11941 PR tree-optimization/84225
11942 * tree-eh.c (find_trapping_overflow): Only call
11943 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11945 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11948 * config/i386/i386.c: Reimplement the check of possible options
11949 -mibt/-mshstk conbination. Change error messages.
11950 * doc/invoke.texi: Fix a typo: remove extra '='.
11952 2018-02-06 Marek Polacek <polacek@redhat.com>
11954 PR tree-optimization/84228
11955 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11957 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11960 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11961 emitted arch directives.
11962 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11963 __ARM_FEATURE_COPROC before changing architectures.
11965 2018-02-06 Richard Biener <rguenther@suse.de>
11967 * config/i386/i386.c (print_reg): Fix typo.
11968 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11970 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11972 * configure: Regenerate.
11974 2018-02-05 Martin Sebor <msebor@redhat.com>
11976 PR tree-optimization/83369
11977 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11980 2018-02-05 Martin Liska <mliska@suse.cz>
11982 * doc/invoke.texi: Cherry-pick upstream r323995.
11984 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11986 * ira.c (ira_init_register_move_cost): Adjust comment.
11988 2018-02-05 Martin Liska <mliska@suse.cz>
11990 PR gcov-profile/84137
11991 * doc/gcov.texi: Fix typo in documentation.
11993 2018-02-05 Martin Liska <mliska@suse.cz>
11995 PR gcov-profile/83879
11996 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11997 using dlopen functionality.
11999 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
12001 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
12002 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
12003 _mm_maskz_range_ss, _mm_mask_range_round_ss,
12004 _mm_maskz_range_round_ss): New intrinsics.
12005 (__builtin_ia32_rangesd128_round)
12006 (__builtin_ia32_rangess128_round): Remove.
12007 (__builtin_ia32_rangesd128_mask_round,
12008 __builtin_ia32_rangess128_mask_round): New builtins.
12009 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
12010 __builtin_ia32_rangess128_round): Remove.
12011 (__builtin_ia32_rangesd128_mask_round,
12012 __builtin_ia32_rangess128_mask_round): New builtins.
12013 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
12014 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
12015 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
12016 "<round_saeonly_constraint>")): Changed to ...
12017 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
12018 "<round_saeonly_scalar_constraint>")): ... this.
12019 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
12020 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
12021 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
12022 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
12023 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
12025 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
12027 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
12029 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
12030 Remove all values except native, 8540 and 8548.
12032 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
12034 * config/i386/i386.c (ix86_output_function_return): Pass
12035 INVALID_REGNUM, instead of -1, as invalid register number to
12036 indirect_thunk_name and output_indirect_thunk.
12038 2018-02-02 Julia Koval <julia.koval@intel.com>
12040 * config.gcc: Add -march=icelake.
12041 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
12042 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
12043 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
12044 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
12045 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
12046 (processor_target_table): Add icelake.
12047 (ix86_option_override_internal): Handle new PTAs.
12048 (get_builtin_code_for_version): Handle icelake.
12049 (M_INTEL_COREI7_ICELAKE): New.
12050 (fold_builtin_cpu): Handle icelake.
12051 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
12052 * doc/invoke.texi: Add -march=icelake.
12054 2018-02-02 Julia Koval <julia.koval@intel.com>
12056 * config/i386/i386.c (ix86_option_override_internal): Change flags type
12057 to wide_int_bitmask.
12058 * wide-int-bitmask.h: New.
12060 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12063 * config/i386/i386.md: Replace Pmode with word_mode in
12064 builtin_setjmp_setup and builtin_longjmp to support x32.
12066 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
12070 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
12072 (rs6000_supported_cpu_names): New static variable.
12073 (linux_cpu_translation_table): Likewise.
12074 (elf_platform) <cpu>: Define new static variable and use it.
12075 Translate kernel AT_PLATFORM name to canonical name if needed.
12076 Error if platform name is unknown.
12078 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
12081 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
12083 2018-02-01 Jeff Law <law@redhat.com>
12086 * config/i386/i386.c (release_scratch_register_on_entry): Add new
12087 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
12088 the scratch if RELEASE_VIA_POP is false.
12089 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
12090 If we have to save a temporary register, decrement SIZE appropriately.
12091 Pass new arguments to release_scratch_register_on_entry.
12092 (ix86_adjust_stack_and_probe): Likewise.
12093 (ix86_emit_probe_stack_range): Pass new arguments to
12094 release_scratch_register_on_entry.
12096 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
12098 PR rtl-optimization/84157
12099 * combine.c (change_zero_ext): Use REG_P predicate in
12100 front of HARD_REGISTER_P predicate.
12102 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
12104 * config/avr/avr.c (avr_option_override): Move disabling of
12105 -fdelete-null-pointer-checks to...
12106 * common/config/avr/avr-common.c (avr_option_optimization_table):
12109 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12111 PR tree-optimization/81635
12112 * tree-data-ref.c (split_constant_offset_1): For types that
12113 wrap on overflow, try to use range info to prove that wrapping
12116 2018-02-01 Renlin Li <renlin.li@arm.com>
12119 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
12120 TAILCALL_ADDR_REGS.
12121 (aarch64_register_move_cost): Likewise.
12122 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
12123 TAILCALL_ADDR_REGS.
12124 (REG_CLASS_NAMES): Likewise.
12125 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
12126 TAILCALL_ADDR_REGS. Remove IP registers.
12127 * config/aarch64/aarch64.md (Ucs): Update register constraint.
12129 2018-02-01 Richard Biener <rguenther@suse.de>
12131 * domwalk.h (dom_walker::dom_walker): Add additional constructor
12132 for specifying RPO order and allow NULL for that.
12133 * domwalk.c (dom_walker::dom_walker): Likewise.
12134 (dom_walker::walk): Handle NULL RPO order.
12135 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
12137 (rewrite_update_dom_walker): Likewise.
12138 (mark_def_dom_walker): Likewise.
12140 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12142 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
12143 (aarch64_maybe_expand_sve_subreg_move): Declare.
12144 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
12145 * config/aarch64/predicates.md (aarch64_any_register_operand): New
12147 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
12148 that are semantically a reverse operation.
12149 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
12150 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
12151 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
12153 (aarch64_can_change_mode_class): For big-endian, forbid changes
12154 between two SVE modes if they have different element sizes.
12156 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12158 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
12159 the TImode handling for big-endian targets.
12161 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12163 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
12164 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
12166 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
12167 Remove BSWAP handing for big-endian targets and use the form of
12168 LD1RQ appropariate for the mode.
12170 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12172 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
12173 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
12174 duplicated element.
12176 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12179 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
12180 check for operands that need to go through aarch64_sve_reload_be.
12182 2018-02-01 Jakub Jelinek <jakub@redhat.com>
12184 PR tree-optimization/81661
12185 PR tree-optimization/84117
12186 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
12187 * tree-eh.c: Include gimplify.h.
12188 (find_trapping_overflow, replace_trapping_overflow,
12189 rewrite_to_non_trapping_overflow): New functions.
12190 * tree-vect-loop.c: Include tree-eh.h.
12191 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
12192 * tree-data-ref.c: Include tree-eh.h.
12193 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
12195 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
12197 PR rtl-optimization/84123
12198 * combine.c (change_zero_ext): Check if hard register satisfies
12199 can_change_dest_mode before calling gen_lowpart_SUBREG.
12201 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
12204 * ira.c (ira_init_register_move_cost): Remove assert.
12206 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12208 PR rtl-optimization/84071
12209 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
12210 * doc/tm.texi: Regenerate.
12212 2018-01-31 Richard Biener <rguenther@suse.de>
12214 PR tree-optimization/84132
12215 * tree-data-ref.c (analyze_miv_subscript): Properly
12216 check whether evolution_function_is_affine_multivariate_p
12217 before calling gcd_of_steps_may_divide_p.
12219 2018-01-31 Julia Koval <julia.koval@intel.com>
12222 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
12223 * config/i386/i386.md (rdpid_rex64) New.
12224 (rdpid): Make 32bit only.
12226 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
12229 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
12230 an IDENTIFIER_NODE for FUNCTION_TYPE's.
12232 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12235 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12237 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12239 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12241 PR rtl-optimization/84071
12242 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
12243 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
12245 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12247 * config/arc/arc.c (arc_handle_aux_attribute): New function.
12248 (arc_attribute_table): Add 'aux' attribute.
12249 (arc_in_small_data_p): Consider aux like variables.
12250 (arc_is_aux_reg_p): New function.
12251 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
12252 (arc_get_aux_arg): New function.
12253 (prepare_move_operands): Handle aux-register access.
12254 (arc_handle_aux_attribute): New function.
12255 * doc/extend.texi (ARC Variable attributes): Add subsection.
12257 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12259 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
12260 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
12261 (arc_attribute_table): Add 'uncached' attribute.
12262 (arc_print_operand): Print '.di' flag for uncached memory
12264 (arc_in_small_data_p): Do not consider for small data the uncached
12266 (arc_is_uncached_mem_p): New function.
12267 * config/arc/predicates.md (compact_store_memory_operand): Check
12268 for uncached memory accesses.
12269 (nonvol_nonimm_operand): Likewise.
12270 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
12272 2018-01-31 Jakub Jelinek <jakub@redhat.com>
12275 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
12276 falign-loops=): Add Optimization flag.
12278 2018-01-30 Jeff Law <law@redhat.com>
12281 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
12282 INT_REGISTERS_SAVED. Check it prior to calling
12283 get_scratch_register_on_entry.
12284 (ix86_adjust_stack_and_probe): Similarly.
12285 (ix86_emit_probe_stack_range): Similarly.
12286 (ix86_expand_prologue): Corresponding changes.
12288 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12291 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
12292 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
12294 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
12297 * lra-constraints.c (curr_insn_transform): Process AND in the
12300 2018-01-30 Jakub Jelinek <jakub@redhat.com>
12302 PR rtl-optimization/83986
12303 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
12304 dependence against last_pending_memory_flush in addition to
12305 pending_jump_insns.
12307 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
12309 PR tree-optimization/81611
12310 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
12313 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12316 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
12319 2018-01-30 Richard Biener <rguenther@suse.de>
12320 Jakub Jelinek <jakub@redhat.com>
12322 PR tree-optimization/84111
12323 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
12324 inner loops added during recursion, as they don't have up-to-date
12327 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12330 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
12331 (can_inline_edge_by_limits_p): ... here.
12332 (can_early_inline_edge_p, check_callers,
12333 update_caller_keys, update_callee_keys, recursive_inlining,
12334 add_new_edges_to_heap, speculation_useful_p,
12335 inline_small_functions,
12336 inline_small_functions, flatten_function,
12337 inline_to_all_callers_1): Update.
12339 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12341 * profile-count.c (profile_count::combine_with_ipa_count): Handle
12344 2018-01-30 Richard Biener <rguenther@suse.de>
12346 PR tree-optimization/83008
12347 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
12348 invariant and constant vector uses in stmts when they need
12349 more than one stmt.
12351 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12354 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
12355 * configure: Regenerate.
12357 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12359 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
12361 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
12362 Use gen_rtx_REG rather than gen_lowpart.
12364 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12366 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
12367 rather than 0 when creating partial subregs.
12369 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12371 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
12374 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
12377 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
12378 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
12379 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
12380 flags. This restores the settings used before the 2017-07-24.
12381 Turning off pre increment/decrement/modify allows IVOPTS to
12382 optimize DF/SF loops where the index is an int.
12384 2018-01-29 Richard Biener <rguenther@suse.de>
12385 Kelvin Nilsen <kelvin@gcc.gnu.org>
12388 * tree-vect-stmts.c (vectorizable_call): Don't call
12389 targetm.vectorize_builtin_md_vectorized_function if callee is
12392 2018-01-22 Carl Love <cel@us.ibm.com>
12394 * doc/extend.tex: Fix typo in second arg in
12395 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
12397 2018-01-29 Richard Biener <rguenther@suse.de>
12399 PR tree-optimization/84086
12400 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
12401 (flush_ssaname_freelist): When SSA names were released reset
12402 the SCEV hash table.
12404 2018-01-29 Richard Biener <rguenther@suse.de>
12406 PR tree-optimization/84057
12407 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
12408 removed paths when removing edges.
12410 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
12412 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
12413 -mfunction-return=@var{choice}.
12415 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
12417 PR diagnostic/84034
12418 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
12419 Handle CR like TAB.
12420 (layout::print_source_line): Likewise.
12421 (test_get_line_width_without_trailing_whitespace): Add test cases.
12423 2018-01-27 Jakub Jelinek <jakub@redhat.com>
12425 PR middle-end/84040
12426 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
12429 2018-01-26 Jim Wilson <jimw@sifive.com>
12431 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
12433 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
12436 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12438 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
12439 and CMP + SUB-immediate -> SUBS.
12441 2018-01-26 Martin Sebor <msebor@redhat.com>
12443 PR tree-optimization/83896
12444 * tree-ssa-strlen.c (get_string_len): Rename...
12445 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
12446 Avoid assuming length is constant.
12447 (handle_char_store): Use HOST_WIDE_INT for string length.
12449 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
12452 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
12453 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
12455 2018-01-26 Richard Biener <rguenther@suse.de>
12457 PR rtl-optimization/84003
12458 * dse.c (record_store): Only record redundant stores when
12459 the earlier store aliases at least all accesses the later one does.
12461 2018-01-26 Jakub Jelinek <jakub@redhat.com>
12463 PR rtl-optimization/83985
12464 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
12465 REG_CFA_RESTORE insns.
12466 (delete_unmarked_insns): Don't ignore separate shrink wrapping
12467 REG_CFA_RESTORE insns here.
12470 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
12471 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
12473 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12475 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
12476 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
12477 (arc_init): Likewise.
12478 (arc_override_options): Likewise.
12479 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
12481 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
12483 * config/arc/arc.h (TARGET_DBNZ): Define.
12484 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
12485 properly set the tune attribute.
12486 (dbnz): Use TARGET_DBNZ guard.
12487 * config/arc/arc.opt (mtune): Add core3 option.
12489 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12491 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
12492 recognize new pic like addresses.
12493 (arc_delegitimize_address): Clean up.
12495 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12497 * config/arc/arc-arches.def: Option mrf16 valid for all
12499 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
12500 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
12501 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
12502 * config/arc/arc-tables.opt: Regenerate.
12503 * config/arc/arc.c (arc_conditional_register_usage): Handle
12504 reduced register file case.
12505 (arc_file_start): Set must have build attributes.
12506 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
12507 mrf16 option value.
12508 * config/arc/arc.opt (mrf16): Add new option.
12509 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
12510 * config/arc/genmultilib.awk: Handle new mrf16 option.
12511 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
12512 * config/arc/t-multilib: Regenerate.
12513 * doc/invoke.texi (ARC Options): Document mrf16 option.
12515 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12517 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
12518 * config/arc/arc.c (arc_handle_secure_attribute): New function.
12519 (arc_attribute_table): Add 'secure_call' attribute.
12520 (arc_print_operand): Print secure call operand.
12521 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12523 (arc_is_secure_call_p): New function. * config/arc/arc.md
12524 (call_i): Add support for sjli instruction.
12525 (call_value_i): Likewise.
12526 * config/arc/constraints.md (Csc): New constraint.
12528 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12529 John Eric Martin <John.Martin@emmicro-us.com>
12531 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12532 * config/arc/arc.c (_arc_jli_section): New struct.
12533 (arc_jli_section): New type.
12534 (rc_jli_sections): New static variable.
12535 (arc_handle_jli_attribute): New function.
12536 (arc_attribute_table): Add jli_always and jli_fixed attribute.
12537 (arc_file_end): New function.
12538 (TARGET_ASM_FILE_END): Define.
12539 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
12540 (arc_add_jli_section): New function.
12541 (jli_call_scan): Likewise.
12542 (arc_reorg): Call jli_call_scan.
12543 (arc_output_addsi): Remove 'S' from printing asm operand.
12544 (arc_is_jli_call_p): New function.
12545 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
12547 (movhi_insn): Likewise.
12548 (movsi_insn): Likewise.
12549 (movsi_set_cc_insn): Likewise.
12550 (loadqi_update): Likewise.
12551 (load_zeroextendqisi_update): Likewise.
12552 (load_signextendqisi_update): Likewise.
12553 (loadhi_update): Likewise.
12554 (load_zeroextendhisi_update): Likewise.
12555 (load_signextendhisi_update): Likewise.
12556 (loadsi_update): Likewise.
12557 (loadsf_update): Likewise.
12558 (movsicc_insn): Likewise.
12559 (bset_insn): Likewise.
12560 (bxor_insn): Likewise.
12561 (bclr_insn): Likewise.
12562 (bmsk_insn): Likewise.
12563 (bicsi3_insn): Likewise.
12564 (cmpsi_cc_c_insn): Likewise.
12565 (movsi_ne): Likewise.
12566 (movsi_cond_exec): Likewise.
12567 (clrsbsi2): Likewise.
12568 (norm_f): Likewise.
12578 (call_i): Remove 'S' asm letter, add jli instruction.
12579 (call_value_i): Likewise.
12580 * config/arc/arc.op (mjli-always): New option.
12581 * config/arc/constraints.md (Cji): New constraint.
12582 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12584 (subsf3_fpx): Likewise.
12585 (mulsf3_fpx): Likewise.
12586 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12588 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12589 function attrbutes.
12590 * doc/invoke.texi (ARC): Document mjli-always option.
12592 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12594 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12595 avoid addition with 0 and use incw and decw where possible.
12597 2018-01-26 Richard Biener <rguenther@suse.de>
12599 PR tree-optimization/81082
12600 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12601 association if it requires casting to unsigned.
12602 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12603 from fold_plusminus_mult_expr to catch important cases late when
12604 range info is available.
12606 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12608 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12609 * configure.ac (hidden_linkonce): New test.
12610 * configure: Regenerate.
12611 * config.in: Regenerate.
12613 2018-01-26 Julia Koval <julia.koval@intel.com>
12615 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12616 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12617 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12618 _mm_mask_bitshuffle_epi64_mask): Fix type.
12619 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12620 USI_FTYPE_V4DI_V4DI_USI): Remove.
12621 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12622 __builtin_ia32_vpshufbitqmb256_mask,
12623 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12624 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12625 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12627 2018-01-26 Alan Modra <amodra@gmail.com>
12630 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12631 UNSPEC_VBPERMQ. Sort other unspecs.
12633 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12635 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12637 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12639 PR middle-end/83055
12640 * predict.c (drop_profile): Do not push/pop cfun; update also
12642 (handle_missing_profiles): Fix logic looking for zero profiles.
12644 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12646 PR middle-end/83977
12647 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12648 on functions with #pragma omp declare simd or functions with simd
12650 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12651 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12652 Remove trailing \n from warning_at calls.
12654 2018-01-25 Tom de Vries <tom@codesourcery.com>
12657 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12658 for neutered workers.
12660 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12663 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12664 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12666 2018-01-24 Jeff Law <law@redhat.com>
12669 * i386.c (get_probe_interval): Move to earlier point.
12670 (ix86_compute_frame_layout): If -fstack-clash-protection and
12671 the frame is larger than the probe interval, then use pushes
12672 to save registers rather than reg->mem moves.
12673 (ix86_expand_prologue): Remove conditional for int_registers_saved
12676 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12679 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12680 min/max for never referenced object.
12682 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12684 PR middle-end/83977
12685 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12687 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12688 attributes from DECL_ATTRIBUTES (decl) without affecting
12689 DECL_ATTRIBUTES (current_function_decl).
12690 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12691 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12693 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12695 PR tree-optimization/83979
12696 * fold-const.c (fold_comparison): Use constant_boolean_node
12697 instead of boolean_{true,false}_node.
12699 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12701 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12704 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12706 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12707 Simplify the clause that sets the length attribute.
12708 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12709 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12710 clause that sets the length attribute.
12711 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12713 2018-01-24 Tom de Vries <tom@codesourcery.com>
12716 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12717 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12718 Add strict parameter.
12719 (prevent_branch_around_nothing): Insert dummy insn between branch to
12720 label and label with no ptx insn inbetween.
12721 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12723 2018-01-24 Tom de Vries <tom@codesourcery.com>
12726 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12727 for neutered threads in warp.
12728 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12730 2018-01-24 Richard Biener <rguenther@suse.de>
12732 PR tree-optimization/83176
12733 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12736 2018-01-24 Richard Biener <rguenther@suse.de>
12738 PR tree-optimization/82819
12739 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12740 code generating pluses that are no-ops in the target precision.
12742 2018-01-24 Richard Biener <rguenther@suse.de>
12744 PR middle-end/84000
12745 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12747 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12749 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12750 to merge probabilities.
12751 * predict.c (probably_never_executed): Also mark as cold functions
12752 with global 0 profile and guessed local profile.
12753 * profile-count.c (profile_probability::combine_with_count): New
12755 * profile-count.h (profile_probability::operator*,
12756 profile_probability::operator*=, profile_probability::operator/,
12757 profile_probability::operator/=): Reduce precision to adjusted
12758 and set value to guessed on contradictory divisions.
12759 (profile_probability::combine_with_freq): Remove.
12760 (profile_probability::combine_wiht_count): Declare.
12761 (profile_count::force_nonzero):: Set to adjusted.
12762 (profile_count::probability_in):: Set quality to adjusted.
12763 * tree-ssa-tail-merge.c (replace_block_by): Use
12764 combine_with_count.
12766 2018-01-23 Andrew Waterman <andrew@sifive.com>
12767 Jim Wilson <jimw@sifive.com>
12769 * config/riscv/riscv.c (riscv_stack_boundary): New.
12770 (riscv_option_override): Set riscv_stack_boundary. Handle
12771 riscv_preferred_stack_boundary_arg.
12772 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12773 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12774 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12775 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12776 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12777 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12779 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12782 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12783 of struct ix86_frame.
12784 (ix86_expand_epilogue): Likewise. Add a local variable for
12785 the reg_save_offset field in struct ix86_frame.
12787 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12789 PR tree-optimization/82604
12790 * tree-loop-distribution.c (enum partition_kind): New enum item
12791 PKIND_PARTIAL_MEMSET.
12792 (partition_builtin_p): Support above new enum item.
12793 (generate_code_for_partition): Ditto.
12794 (compute_access_range): Differentiate cases that equality can be
12795 proven at all loops, the innermost loops or no loops.
12796 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12797 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12798 (finalize_partitions, distribute_loop): Don't fuse partition of
12799 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12800 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12801 parloop is enabled.
12803 2018-01-23 Martin Liska <mliska@suse.cz>
12805 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12806 order to ignore the predictor.
12807 (PRED_POLYMORPHIC_CALL): Likewise.
12808 (PRED_RECURSIVE_CALL): Likewise.
12810 2018-01-23 Martin Liska <mliska@suse.cz>
12812 * tree-profile.c (tree_profiling): Print function header to
12813 aware reader which function we are working on.
12814 * value-prof.c (gimple_find_values_to_profile): Do not print
12815 not interesting value histograms.
12817 2018-01-23 Martin Liska <mliska@suse.cz>
12819 * profile-count.h (enum profile_quality): Add
12820 profile_uninitialized as the first value. Do not number values
12821 as they are zero based.
12822 (profile_count::verify): Update sanity check.
12823 (profile_probability::verify): Likewise.
12825 2018-01-23 Nathan Sidwell <nathan@acm.org>
12827 * doc/invoke.texi (ffor-scope): Deprecate.
12829 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12831 PR tree-optimization/83510
12832 * domwalk.c (set_all_edges_as_executable): New function.
12833 (dom_walker::dom_walker): Convert bool param
12834 "skip_unreachable_blocks" to enum reachability. Move setup of
12835 edge flags to set_all_edges_as_executable and only do it when
12836 reachability is REACHABLE_BLOCKS.
12837 * domwalk.h (enum dom_walker::reachability): New enum.
12838 (dom_walker::dom_walker): Convert bool param
12839 "skip_unreachable_blocks" to enum reachability.
12840 (set_all_edges_as_executable): New decl.
12841 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12842 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12844 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12845 but converting true to REACHABLE_BLOCKS.
12846 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12848 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12849 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12850 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12852 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12853 if check_all_array_refs will be called.
12855 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12857 * tree.c (selftest::test_location_wrappers): Add more test
12860 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12862 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12863 (selftest::test_bit_in_range): Likewise.
12865 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12868 * doc/sourcebuild.texi (vect_float): Say that the selector
12869 only describes the situation when -funsafe-math-optimizations is on.
12870 (vect_float_strict): Document.
12872 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12874 PR tree-optimization/83965
12875 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12876 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12877 instead of checking only for a reduction.
12878 (vect_recog_widen_sum_pattern): Likewise.
12880 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12882 * predict.c (probably_never_executed): Only use precise profile info.
12883 (compute_function_frequency): Skip after inlining hack since we now
12884 have quality checking.
12886 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12888 * profile-count.h (profile_probability::very_unlikely,
12889 profile_probability::unlikely, profile_probability::even): Set
12890 precision to guessed.
12892 2018-01-23 Richard Biener <rguenther@suse.de>
12894 PR tree-optimization/83963
12895 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12896 Properly terminate dominator walk when crossing the exit edge not
12897 when visiting its source block.
12899 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12902 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12903 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12905 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12907 PR tree-optimization/83957
12908 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12909 semicolon after for body surrounded by braces.
12911 PR tree-optimization/83081
12912 * profile-count.h (profile_probability::split): New method.
12913 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12914 Use profile_probability::split.
12915 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12916 when splitting a single conditional jump into 2.
12918 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12920 PR tree-optimization/69452
12921 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12924 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12926 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12927 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12928 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12930 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12932 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12934 * config/rl78/rl78.md (movdi): New define_expand.
12935 * config/rl78/rl78.c (rl78_split_movdi): New function.
12937 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12940 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12942 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12943 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12944 128-bit to produce an UNSPEC move to get the double word with the
12945 signbit and then a shift directly to do signbit.
12946 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12947 implementation with a new version that just does either a direct
12948 move or a regular move. Move memory interface to separate insns.
12949 Move insns so they are next to the expander.
12950 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12951 with signbit move. Split big and little endian case.
12952 (signbit<mode>2_dm_mem_le): Likewise.
12953 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12954 (signbit<mode>2_dm2): Likewise.
12956 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12958 * config/rl78/rl78.md (anddi3): New define_expand.
12960 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12962 * config/rl78/rl78.md (umindi3): New define_expand.
12964 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12966 * config/rl78/rl78.md (smindi3): New define_expand.
12968 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12970 * config/rl78/rl78.md (smaxdi3): New define_expand.
12972 2018-01-22 Carl Love <cel@us.ibm.com>
12974 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12975 LVX_V1TI): Add macro expansion.
12976 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12977 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12978 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12979 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12980 Change check to determine if the instruction is a byte reversing
12981 entry. Fix typo in comment.
12982 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12983 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12984 Add def_builtin calls for new builtins.
12985 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12986 Add define_insn expansion.
12988 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12990 * config/rl78/rl78.md (umaxdi3): New define_expand.
12992 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12994 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12995 for non-QImode registers.
12997 2018-01-22 Richard Biener <rguenther@suse.de>
12999 PR tree-optimization/83963
13000 * graphite-scop-detection.c (scop_detection::get_sese): Delay
13001 including the loop exit block.
13002 (scop_detection::merge_sese): Likewise.
13003 (scop_detection::add_scop): Do it here instead.
13005 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13007 * doc/sourcebuild.texi (arm_softfloat): Document.
13009 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
13012 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
13013 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
13014 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
13016 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13017 David Edelsohn <dje.gcc@gmail.com>
13020 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13021 Change "crset eq" to "crset 2".
13022 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13023 (*call_indirect_aix<mode>_nospec): Likewise.
13024 (*call_value_indirect_aix<mode>_nospec): Likewise.
13025 (*call_indirect_elfv2<mode>_nospec): Likewise.
13026 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
13027 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
13028 change assembly output from . to $.
13029 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13030 (indirect_jump<mode>_nospec): Change assembly output from . to $.
13031 (*tablejump<mode>_internal1_nospec): Likewise.
13033 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
13036 * config/sh/sh_optimize_sett_clrt.cc:
13037 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
13039 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
13041 PR tree-optimization/83940
13042 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
13043 offset_dt to vect_constant_def rather than vect_unknown_def_type.
13044 (vect_check_load_store_mask): Add a mask_dt_out parameter and
13045 use it to pass back the definition type.
13046 (vect_check_store_rhs): Likewise rhs_dt_out.
13047 (vect_build_gather_load_calls): Add a mask_dt argument and use
13048 it instead of a call to vect_is_simple_use.
13049 (vectorizable_store): Update calls to vect_check_load_store_mask
13050 and vect_check_store_rhs. Use the dt returned by the latter instead
13051 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
13052 instead of calls to vect_is_simple_use. Pass the scalar rather
13053 than the vector operand to vect_is_simple_use when handling
13054 second and subsequent copies of an rhs value.
13055 (vectorizable_load): Update calls to vect_check_load_store_mask
13056 and vect_build_gather_load_calls. Use the cached mask_dt and
13057 gs_info.offset_dt instead of calls to vect_is_simple_use.
13059 2018-01-20 Jakub Jelinek <jakub@redhat.com>
13061 PR middle-end/83945
13062 * tree-emutls.c: Include gimplify.h.
13063 (lower_emutls_2): New function.
13064 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
13065 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
13066 it before further processing.
13069 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
13070 UINTVAL (trueop1) instead of INTVAL (op1).
13072 2018-01-19 Jakub Jelinek <jakub@redhat.com>
13076 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
13077 INCOMING_FRAME_SP_OFFSET if not defined.
13078 (scan_trace): Add ENTRY argument. If true and
13079 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
13080 emit a note to adjust the CFA offset.
13081 (create_cfi_notes): Adjust scan_trace callers.
13082 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
13083 INCOMING_FRAME_SP_OFFSET in the CIE.
13084 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
13085 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
13087 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
13088 * doc/tm.texi: Regenerated.
13090 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13092 PR rtl-optimization/83147
13093 * lra-constraints.c (remove_inheritance_pseudos): Use
13094 lra_substitute_pseudo_within_insn.
13096 2018-01-19 Tom de Vries <tom@codesourcery.com>
13097 Cesar Philippidis <cesar@codesourcery.com>
13100 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
13102 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
13105 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
13106 spaces for function labels.
13108 2018-01-19 Martin Liska <mliska@suse.cz>
13110 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
13111 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
13112 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
13113 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
13114 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
13115 (PRED_CONST_RETURN): Change from 69 to 65.
13116 (PRED_NULL_RETURN): Change from 91 to 71.
13117 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
13118 (PRED_LOOP_GUARD): Change from 66 to 73.
13120 2018-01-19 Martin Liska <mliska@suse.cz>
13122 * predict.c (predict_insn_def): Add new assert.
13123 (struct branch_predictor): Change type to signed integer.
13124 (test_prediction_value_range): Amend test to cover
13125 PROB_UNINITIALIZED.
13126 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
13127 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
13128 (PRED_LOOP_ITERATIONS_MAX): Likewise.
13129 (PRED_LOOP_IV_COMPARE): Likewise.
13130 * predict.h (PROB_UNINITIALIZED): Define new constant.
13132 2018-01-19 Martin Liska <mliska@suse.cz>
13134 * predict.c (dump_prediction): Add new format for
13135 analyze_brprob.py script which is enabled with -details
13137 * profile-count.h (precise_p): New function.
13139 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13141 PR tree-optimization/83922
13142 * tree-vect-loop.c (vect_verify_full_masking): Return false if
13143 there are no statements that need masking.
13144 (vect_active_double_reduction_p): New function.
13145 (vect_analyze_loop_operations): Use it when handling phis that
13146 are not in the loop header.
13148 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13150 PR tree-optimization/83914
13151 * tree-vect-loop.c (vectorizable_induction): Don't convert
13152 init_expr or apply the peeling adjustment for inductions
13153 that are nested within the vectorized loop.
13155 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13157 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
13160 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13164 * function.h (gimplify_parameters): Add gimple_seq * argument.
13165 * function.c: Include gimple.h and options.h.
13166 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
13167 for the added local temporaries if needed.
13168 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
13169 if there are any parameter cleanups, wrap whole body into a
13170 try/finally with the cleanups.
13172 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
13175 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
13176 Use GET_MODE_CLASS for scalar floating point.
13178 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13182 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
13183 Fix call of call_cgraph_insertion_hooks.
13185 2018-01-18 Martin Sebor <msebor@redhat.com>
13187 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
13189 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13192 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
13195 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
13198 * common.opt: (-ffile-prefix-map): New option.
13199 * opts.c (common_handle_option): Defer it.
13200 * opts-global.c (handle_common_deferred_options): Handle it.
13201 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
13202 * file-prefix-map.h: New file.
13203 (remap_debug_filename, add_debug_prefix_map): ...here.
13204 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
13205 * final.c (debug_prefix_map, add_debug_prefix_map
13206 remap_debug_filename): Move to...
13207 * file-prefix-map.c: New file.
13208 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
13209 generalize, get rid of alloca(), use strrchr() instead of strchr().
13210 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
13211 Implement in terms of add_prefix_map().
13212 (remap_macro_filename, remap_debug_filename): Implement in term of
13214 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
13215 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
13216 * dbxout.c: Include file-prefix-map.h.
13217 * varasm.c: Likewise.
13218 * vmsdbgout.c: Likewise.
13219 * xcoffout.c: Likewise.
13220 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
13221 * doc/cppopts.texi (-fmacro-prefix-map): Document.
13222 * doc/invoke.texi (-ffile-prefix-map): Document.
13223 (-fdebug-prefix-map): Update description.
13225 2018-01-18 Martin Liska <mliska@suse.cz>
13227 * config/i386/i386.c (indirect_thunk_name): Document that also
13229 (output_indirect_thunk): Document why both instructions
13230 (pause and lfence) are generated.
13232 2018-01-18 Richard Biener <rguenther@suse.de>
13234 PR tree-optimization/83887
13235 * graphite-scop-detection.c
13236 (scop_detection::get_nearest_dom_with_single_entry): Remove.
13237 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
13238 (scop_detection::merge_sese): Re-implement with a flood-fill
13239 algorithm that properly finds a SESE region if it exists.
13241 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13244 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
13245 pointer_diff optimizations use view_convert instead of convert.
13247 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13249 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13250 Generate different code for -mno-speculate-indirect-jumps.
13251 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13252 (*call_indirect_aix<mode>): Disable for
13253 -mno-speculate-indirect-jumps.
13254 (*call_indirect_aix<mode>_nospec): New define_insn.
13255 (*call_value_indirect_aix<mode>): Disable for
13256 -mno-speculate-indirect-jumps.
13257 (*call_value_indirect_aix<mode>_nospec): New define_insn.
13258 (*sibcall_nonlocal_sysv<mode>): Generate different code for
13259 -mno-speculate-indirect-jumps.
13260 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13262 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
13264 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
13265 long double type, set the flags for noting the default long double
13266 type, even if we don't pass or return a long double type.
13268 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
13271 * ipa-inline.c (flatten_function): Do not overwrite final inlining
13274 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
13276 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
13277 support for merge[hl].
13278 (fold_mergehl_helper): New helper function.
13279 (tree-vector-builder.h): New #include for tree_vector_builder usage.
13280 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
13281 (altivec_vmrglw_direct): Add xxmrglw insn.
13283 2018-01-17 Andrew Waterman <andrew@sifive.com>
13285 * config/riscv/riscv.c (riscv_conditional_register_usage): If
13286 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
13288 2018-01-17 David Malcolm <dmalcolm@redhat.com>
13291 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
13292 call the lto_location_cache before reading the
13293 DECL_SOURCE_LOCATION of the types.
13295 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
13296 Richard Sandiford <richard.sandiford@linaro.org>
13298 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
13299 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
13300 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
13301 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
13302 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
13304 * config/aarch64/constraints.md (aarch64_movti_operand):
13306 * config/aarch64/predicates.md (Uti): Add new constraint.
13308 2018-01-17 Carl Love <cel@us.ibm.com>
13310 * config/rs6000/vsx.md (define_expand xl_len_r,
13311 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
13312 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
13314 (define_expand, define_insn): Move the shift left from the
13315 define_insn to the define_expand for lxvl and stxvl instructions.
13316 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
13317 and XL_LEN_R definitions to PURE.
13319 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
13321 * config/i386/i386.c (indirect_thunk_name): Declare regno
13322 as unsigned int. Compare regno with INVALID_REGNUM.
13323 (output_indirect_thunk): Ditto.
13324 (output_indirect_thunk_function): Ditto.
13325 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
13326 in the call to output_indirect_thunk_function.
13328 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
13330 PR middle-end/83884
13331 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
13332 rather than the size of inner_type to determine the stack slot size
13333 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
13335 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
13338 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
13341 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
13343 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
13344 endian Linux systems to optionally enable multilibs for selecting
13345 the long double type if the user configured an explicit type.
13346 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
13347 have no long double multilibs if not defined.
13348 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
13349 warn if the user used -mabi={ieee,ibm}longdouble and we built
13350 multilibs for long double.
13351 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
13352 appropriate multilib option.
13353 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
13355 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
13356 for building long double multilibs.
13357 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
13359 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
13361 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
13364 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
13366 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
13369 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
13372 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
13375 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13377 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
13378 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
13380 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
13382 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
13383 different rtl trees depending on TARGET_64BIT.
13384 (rs6000_gen_lvx): Likewise.
13386 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13388 * config/visium/visium.md (nop): Tweak comment.
13389 (hazard_nop): Likewise.
13391 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13393 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
13394 -mspeculate-indirect-jumps.
13395 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
13396 for -mno-speculate-indirect-jumps.
13397 (*call_indirect_elfv2<mode>_nospec): New define_insn.
13398 (*call_value_indirect_elfv2<mode>): Disable for
13399 -mno-speculate-indirect-jumps.
13400 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
13401 (indirect_jump): Emit different RTL for
13402 -mno-speculate-indirect-jumps.
13403 (*indirect_jump<mode>): Disable for
13404 -mno-speculate-indirect-jumps.
13405 (*indirect_jump<mode>_nospec): New define_insn.
13406 (tablejump): Emit different RTL for
13407 -mno-speculate-indirect-jumps.
13408 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
13409 (tablejumpsi_nospec): New define_expand.
13410 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
13411 (tablejumpdi_nospec): New define_expand.
13412 (*tablejump<mode>_internal1): Disable for
13413 -mno-speculate-indirect-jumps.
13414 (*tablejump<mode>_internal1_nospec): New define_insn.
13415 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
13418 2018-01-16 Artyom Skrobov tyomitch@gmail.com
13420 * caller-save.c (insert_save): Drop unnecessary parameter. All
13423 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13424 Richard Biener <rguenth@suse.de>
13427 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
13428 return early, inline manually is_gimple_sizepos. Make sure if we
13429 call gimplify_expr we don't end up with a gimple constant.
13430 * tree.c (variably_modified_type_p): Don't return true for
13431 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
13432 * gimplify.h (is_gimple_sizepos): Remove.
13434 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13436 PR tree-optimization/83857
13437 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
13438 vectorizable_live_operation for pure SLP statements.
13439 (vectorizable_live_operation): Handle PHIs.
13441 2018-01-16 Richard Biener <rguenther@suse.de>
13443 PR tree-optimization/83867
13444 * tree-vect-stmts.c (vect_transform_stmt): Precompute
13445 nested_in_vect_loop_p since the scalar stmt may get invalidated.
13447 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13450 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
13451 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
13452 If off is not INTEGER_CST, issue a may not be aligned warning
13453 rather than isn't aligned. Use isn%'t rather than isn't.
13454 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
13456 <case MULT_EXPR>: Improve the case when bottom and one of the
13457 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
13458 operand, in that case check if the other operand is multiple of
13459 bottom divided by the INTEGER_CST operand.
13461 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13464 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
13465 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
13466 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
13467 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
13468 * config/pa/pa.c (pa_function_arg_advance): Likewise.
13469 (pa_function_arg, pa_arg_partial_bytes): Likewise.
13470 (pa_function_arg_size): New function.
13472 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13474 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
13475 in a separate statement.
13477 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13479 PR tree-optimization/83847
13480 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
13481 group gathers and scatters.
13483 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13485 PR rtl-optimization/86620
13486 * params.def (max-sched-ready-insns): Bump minimum value to 1.
13488 PR rtl-optimization/83213
13489 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
13490 to last if both are JUMP_INSNs.
13492 PR tree-optimization/83843
13493 * gimple-ssa-store-merging.c
13494 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
13495 store_immediate_info for bswap/nop orig_stores.
13497 2018-01-15 Andrew Waterman <andrew@sifive.com>
13499 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
13501 <UDIV>: Increase cost if !TARGET_DIV.
13503 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
13505 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
13506 (define_attr "cr_logical_3op"): New.
13507 (cceq_ior_compare): Adjust.
13508 (cceq_ior_compare_complement): Adjust.
13509 (*cceq_rev_compare): Adjust.
13510 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
13511 (is_cracked_insn): Adjust.
13512 (insn_must_be_first_in_group): Adjust.
13513 * config/rs6000/40x.md: Adjust.
13514 * config/rs6000/440.md: Adjust.
13515 * config/rs6000/476.md: Adjust.
13516 * config/rs6000/601.md: Adjust.
13517 * config/rs6000/603.md: Adjust.
13518 * config/rs6000/6xx.md: Adjust.
13519 * config/rs6000/7450.md: Adjust.
13520 * config/rs6000/7xx.md: Adjust.
13521 * config/rs6000/8540.md: Adjust.
13522 * config/rs6000/cell.md: Adjust.
13523 * config/rs6000/e300c2c3.md: Adjust.
13524 * config/rs6000/e500mc.md: Adjust.
13525 * config/rs6000/e500mc64.md: Adjust.
13526 * config/rs6000/e5500.md: Adjust.
13527 * config/rs6000/e6500.md: Adjust.
13528 * config/rs6000/mpc.md: Adjust.
13529 * config/rs6000/power4.md: Adjust.
13530 * config/rs6000/power5.md: Adjust.
13531 * config/rs6000/power6.md: Adjust.
13532 * config/rs6000/power7.md: Adjust.
13533 * config/rs6000/power8.md: Adjust.
13534 * config/rs6000/power9.md: Adjust.
13535 * config/rs6000/rs64.md: Adjust.
13536 * config/rs6000/titan.md: Adjust.
13538 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13540 * config/i386/predicates.md (indirect_branch_operand): Rewrite
13541 ix86_indirect_branch_register logic.
13543 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13545 * config/i386/constraints.md (Bs): Update
13546 ix86_indirect_branch_register check. Don't check
13547 ix86_indirect_branch_register with GOT_memory_operand.
13549 * config/i386/predicates.md (GOT_memory_operand): Don't check
13550 ix86_indirect_branch_register here.
13551 (GOT32_symbol_operand): Likewise.
13553 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13555 * config/i386/predicates.md (constant_call_address_operand):
13556 Rewrite ix86_indirect_branch_register logic.
13557 (sibcall_insn_operand): Likewise.
13559 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13561 * config/i386/constraints.md (Bs): Replace
13562 ix86_indirect_branch_thunk_register with
13563 ix86_indirect_branch_register.
13565 * config/i386/i386.md (indirect_jump): Likewise.
13566 (tablejump): Likewise.
13567 (*sibcall_memory): Likewise.
13568 (*sibcall_value_memory): Likewise.
13569 Peepholes of indirect call and jump via memory: Likewise.
13570 * config/i386/i386.opt: Likewise.
13571 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13572 (GOT_memory_operand): Likewise.
13573 (call_insn_operand): Likewise.
13574 (sibcall_insn_operand): Likewise.
13575 (GOT32_symbol_operand): Likewise.
13577 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13579 PR middle-end/83837
13580 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13581 type rather than type addr's type points to.
13582 (expand_omp_atomic_mutex): Likewise.
13583 (expand_omp_atomic): Likewise.
13585 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13588 * config/i386/i386.c (output_indirect_thunk_function): Use
13589 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13590 for __x86_return_thunk.
13592 2018-01-15 Richard Biener <rguenther@suse.de>
13594 PR middle-end/83850
13595 * expmed.c (extract_bit_field_1): Fix typo.
13597 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13600 * config/arm/iterators.md (VF): New mode iterator.
13601 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13602 Remove integer-related logic from pattern.
13603 (neon_vabd<mode>_3): Likewise.
13605 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13607 PR middle-end/82694
13608 * common.opt (fstrict-overflow): No longer an alias.
13609 (fwrapv-pointer): New option.
13610 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13611 also for pointer types based on flag_wrapv_pointer.
13612 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13613 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13614 opts->x_flag_wrapv got set.
13615 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13616 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13617 POINTER_TYPE_OVERFLOW_UNDEFINED.
13618 * match.pd: Likewise in address comparison pattern.
13619 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13621 2018-01-15 Richard Biener <rguenther@suse.de>
13624 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13625 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13626 Reset type names to their identifier if their TYPE_DECL doesn't
13627 have linkage (and thus is used for ODR and devirt).
13628 (save_debug_info_for_decl): Remove.
13629 (save_debug_info_for_type): Likewise.
13630 (add_tree_to_fld_list): Adjust.
13631 * tree-pretty-print.c (dump_generic_node): Make dumping of
13632 type names more robust.
13634 2018-01-15 Richard Biener <rguenther@suse.de>
13636 * BASE-VER: Bump to 8.0.1.
13638 2018-01-14 Martin Sebor <msebor@redhat.com>
13641 * builtins.c (check_access): Avoid warning when the no-warning bit
13644 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13646 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13647 * ira-color (allocno_hard_regs_compare): Likewise.
13649 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13652 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13653 Use .pushsection/.popsection.
13655 2018-01-14 Martin Sebor <msebor@redhat.com>
13658 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13660 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13662 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13663 entry from extra_headers.
13664 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13665 extra_headers, make the list bitwise identical to the i?86-*-* one.
13667 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13669 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13670 -mcmodel=large with -mindirect-branch=thunk,
13671 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13672 -mfunction-return=thunk-extern.
13673 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13674 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13675 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13677 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13679 * config/i386/i386.c (print_reg): Print the name of the full
13680 integer register without '%'.
13681 (ix86_print_operand): Handle 'V'.
13682 * doc/extend.texi: Document 'V' modifier.
13684 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13686 * config/i386/constraints.md (Bs): Disallow memory operand for
13687 -mindirect-branch-register.
13689 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13690 (GOT_memory_operand): Likewise.
13691 (call_insn_operand): Likewise.
13692 (sibcall_insn_operand): Likewise.
13693 (GOT32_symbol_operand): Likewise.
13694 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13695 for -mindirect-branch-register.
13696 (tablejump): Likewise.
13697 (*sibcall_memory): Likewise.
13698 (*sibcall_value_memory): Likewise.
13699 Disallow peepholes of indirect call and jump via memory for
13700 -mindirect-branch-register.
13701 (*call_pop): Replace m with Bw.
13702 (*call_value_pop): Likewise.
13703 (*sibcall_pop_memory): Replace m with Bs.
13704 * config/i386/i386.opt (mindirect-branch-register): New option.
13705 * doc/invoke.texi: Document -mindirect-branch-register option.
13707 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13709 * config/i386/i386-protos.h (ix86_output_function_return): New.
13710 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13711 set function_return_type.
13712 (indirect_thunk_name): Add ret_p to indicate thunk for function
13714 (output_indirect_thunk_function): Pass false to
13715 indirect_thunk_name.
13716 (ix86_output_indirect_branch_via_reg): Likewise.
13717 (ix86_output_indirect_branch_via_push): Likewise.
13718 (output_indirect_thunk_function): Create alias for function
13719 return thunk if regno < 0.
13720 (ix86_output_function_return): New function.
13721 (ix86_handle_fndecl_attribute): Handle function_return.
13722 (ix86_attribute_table): Add function_return.
13723 * config/i386/i386.h (machine_function): Add
13724 function_return_type.
13725 * config/i386/i386.md (simple_return_internal): Use
13726 ix86_output_function_return.
13727 (simple_return_internal_long): Likewise.
13728 * config/i386/i386.opt (mfunction-return=): New option.
13729 (indirect_branch): Mention -mfunction-return=.
13730 * doc/extend.texi: Document function_return function attribute.
13731 * doc/invoke.texi: Document -mfunction-return= option.
13733 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13735 * config/i386/i386-opts.h (indirect_branch): New.
13736 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13737 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13738 with local indirect jump when converting indirect call and jump.
13739 (ix86_set_indirect_branch_type): New.
13740 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13741 (indirectlabelno): New.
13742 (indirect_thunk_needed): Likewise.
13743 (indirect_thunk_bnd_needed): Likewise.
13744 (indirect_thunks_used): Likewise.
13745 (indirect_thunks_bnd_used): Likewise.
13746 (INDIRECT_LABEL): Likewise.
13747 (indirect_thunk_name): Likewise.
13748 (output_indirect_thunk): Likewise.
13749 (output_indirect_thunk_function): Likewise.
13750 (ix86_output_indirect_branch_via_reg): Likewise.
13751 (ix86_output_indirect_branch_via_push): Likewise.
13752 (ix86_output_indirect_branch): Likewise.
13753 (ix86_output_indirect_jmp): Likewise.
13754 (ix86_code_end): Call output_indirect_thunk_function if needed.
13755 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13757 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13758 (ix86_attribute_table): Add indirect_branch.
13759 * config/i386/i386.h (machine_function): Add indirect_branch_type
13760 and has_local_indirect_jump.
13761 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13763 (tablejump): Likewise.
13764 (*indirect_jump): Use ix86_output_indirect_jmp.
13765 (*tablejump_1): Likewise.
13766 (simple_return_indirect_internal): Likewise.
13767 * config/i386/i386.opt (mindirect-branch=): New option.
13768 (indirect_branch): New.
13771 (thunk-inline): Likewise.
13772 (thunk-extern): Likewise.
13773 * doc/extend.texi: Document indirect_branch function attribute.
13774 * doc/invoke.texi: Document -mindirect-branch= option.
13776 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13779 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13781 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13783 * ipa-inline.c (want_inline_small_function_p): Return false if
13784 inlining has already failed with CIF_FINAL_ERROR.
13785 (update_caller_keys): Call want_inline_small_function_p before
13787 (update_callee_keys): Likewise.
13789 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13791 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13793 (rs6000_quadword_masked_address_p): Likewise.
13794 (quad_aligned_load_p): Likewise.
13795 (quad_aligned_store_p): Likewise.
13796 (const_load_sequence_p): Add comment to describe the outer-most loop.
13797 (mimic_memory_attributes_and_flags): New function.
13798 (rs6000_gen_stvx): Likewise.
13799 (replace_swapped_aligned_store): Likewise.
13800 (rs6000_gen_lvx): Likewise.
13801 (replace_swapped_aligned_load): Likewise.
13802 (replace_swapped_load_constant): Capitalize argument name in
13803 comment describing this function.
13804 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13805 and stores that access quad-word aligned addresses and replace
13806 with stvx or lvx instructions when appropriate.
13807 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13808 New function prototype.
13809 (rs6000_quadword_masked_address_p): Likewise.
13810 (rs6000_gen_lvx): Likewise.
13811 (rs6000_gen_stvx): Likewise.
13812 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13813 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13814 when memory address is aligned.
13815 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13816 this split to select lvx instruction when memory address is aligned.
13817 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13818 instruction when memory address is aligned.
13819 (*vsx_le_perm_load_v16qi): Likewise.
13820 (four unnamed splitters): Modify to select the stvx instruction
13821 when memory is aligned.
13823 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13825 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13826 which appears in the queue multiple times.
13828 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13829 Alan Hayward <alan.hayward@arm.com>
13830 David Sherwood <david.sherwood@arm.com>
13832 * tree-vectorizer.h (vec_lower_bound): New structure.
13833 (_loop_vec_info): Add check_nonzero and lower_bounds.
13834 (LOOP_VINFO_CHECK_NONZERO): New macro.
13835 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13836 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13837 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13838 fields. Make seg_len the distance travelled, not including the
13840 (dr_direction_indicator): Declare.
13841 (dr_zero_step_indicator): Likewise.
13842 (dr_known_forward_stride_p): Likewise.
13843 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13845 (runtime_alias_check_p): Allow runtime alias checks with
13847 (operator ==): Compare access_size and align.
13848 (prune_runtime_alias_test_list): Rework for new distinction between
13849 the access_size and seg_len.
13850 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13852 (get_segment_min_max): New function.
13853 (create_intersect_range_checks): Use it.
13854 (dr_step_indicator): New function.
13855 (dr_direction_indicator): Likewise.
13856 (dr_zero_step_indicator): Likewise.
13857 (dr_known_forward_stride_p): Likewise.
13858 * tree-loop-distribution.c (data_ref_segment_size): Return
13859 DR_STEP * (niters - 1).
13860 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13862 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13863 (vect_preserves_scalar_order_p): New function, split out from...
13864 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13865 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13866 (vect_vfa_access_size): New function.
13867 (vect_vfa_align): Likewise.
13868 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13869 (dump_lower_bound): New function.
13870 (vect_check_lower_bound): Likewise.
13871 (vect_small_gap_p): Likewise.
13872 (vectorizable_with_step_bound_p): Likewise.
13873 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13874 depencies if the vectorization factor is 1. Convert the checks
13875 for nonzero steps into checks on the bounds of DR_STEP. Try using
13876 a bunds check for variable steps if the minimum required step is
13877 relatively small. Update calls to the dr_with_seg_len
13878 constructor and to vect_compile_time_alias.
13879 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13881 (vect_loop_versioning): Call it.
13882 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13884 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13886 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13887 Alan Hayward <alan.hayward@arm.com>
13888 David Sherwood <david.sherwood@arm.com>
13890 * doc/sourcebuild.texi (vect_scatter_store): Document.
13891 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13893 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13895 * genopinit.c (main): Add supports_vec_scatter_store and
13896 supports_vec_scatter_store_cached to target_optabs.
13897 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13898 IFN_MASK_SCATTER_STORE.
13899 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13901 * internal-fn.h (internal_store_fn_p): Declare.
13902 (internal_fn_stored_value_index): Likewise.
13903 * internal-fn.c (scatter_store_direct): New macro.
13904 (expand_scatter_store_optab_fn): New function.
13905 (direct_scatter_store_optab_supported_p): New macro.
13906 (internal_store_fn_p): New function.
13907 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13908 IFN_MASK_SCATTER_STORE.
13909 (internal_fn_mask_index): Likewise.
13910 (internal_fn_stored_value_index): New function.
13911 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13912 for scatter stores.
13913 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13914 * optabs-query.c (supports_vec_scatter_store_p): New function.
13915 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13916 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13917 true for scatter stores.
13918 (vect_gather_scatter_fn_p): Handle scatter stores too.
13919 (vect_check_gather_scatter): Consider using scatter stores if
13920 supports_vec_scatter_store_p.
13921 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13922 scatter stores too.
13923 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13924 internal_fn_stored_value_index.
13925 (check_load_store_masking): Handle scatter stores too.
13926 (vect_get_store_rhs): Make public.
13927 (vectorizable_call): Use internal_store_fn_p.
13928 (vectorizable_store): Handle scatter store internal functions.
13929 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13930 when deciding whether the end of the group has been reached.
13931 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13932 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13933 (mask_scatter_store<mode>): New insns.
13935 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13936 Alan Hayward <alan.hayward@arm.com>
13937 David Sherwood <david.sherwood@arm.com>
13939 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13940 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13941 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13943 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13944 Use vect_truncate_gather_scatter_offset if we can't treat the
13945 operation as a normal gather load or scatter store.
13946 (get_group_load_store_type): Take the gather_scatter_info
13947 as argument. Try using a gather load or scatter store for
13948 single-element groups.
13949 (get_load_store_type): Update calls to get_group_load_store_type
13950 and vect_use_strided_gather_scatters_p.
13952 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13953 Alan Hayward <alan.hayward@arm.com>
13954 David Sherwood <david.sherwood@arm.com>
13956 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13957 optional tree argument.
13958 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13960 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13961 but continue to use the current value as a fallback.
13962 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13963 to compare the updates.
13964 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13965 (get_load_store_type): Use it when handling a strided access.
13966 (vect_get_strided_load_store_ops): New function.
13967 (vect_get_data_ptr_increment): Likewise.
13968 (vectorizable_load): Handle strided gather loads. Always pass
13969 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13971 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13972 Alan Hayward <alan.hayward@arm.com>
13973 David Sherwood <david.sherwood@arm.com>
13975 * doc/md.texi (gather_load@var{m}): Document.
13976 (mask_gather_load@var{m}): Likewise.
13977 * genopinit.c (main): Add supports_vec_gather_load and
13978 supports_vec_gather_load_cached to target_optabs.
13979 * optabs-tree.c (init_tree_optimization_optabs): Use
13980 ggc_cleared_alloc to allocate target_optabs.
13981 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13982 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13984 * internal-fn.h (internal_load_fn_p): Declare.
13985 (internal_gather_scatter_fn_p): Likewise.
13986 (internal_fn_mask_index): Likewise.
13987 (internal_gather_scatter_fn_supported_p): Likewise.
13988 * internal-fn.c (gather_load_direct): New macro.
13989 (expand_gather_load_optab_fn): New function.
13990 (direct_gather_load_optab_supported_p): New macro.
13991 (direct_internal_fn_optab): New function.
13992 (internal_load_fn_p): Likewise.
13993 (internal_gather_scatter_fn_p): Likewise.
13994 (internal_fn_mask_index): Likewise.
13995 (internal_gather_scatter_fn_supported_p): Likewise.
13996 * optabs-query.c (supports_at_least_one_mode_p): New function.
13997 (supports_vec_gather_load_p): Likewise.
13998 * optabs-query.h (supports_vec_gather_load_p): Declare.
13999 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
14000 and memory_type field.
14001 (NUM_PATTERNS): Bump to 15.
14002 * tree-vect-data-refs.c: Include internal-fn.h.
14003 (vect_gather_scatter_fn_p): New function.
14004 (vect_describe_gather_scatter_call): Likewise.
14005 (vect_check_gather_scatter): Try using internal functions for
14006 gather loads. Recognize existing calls to a gather load function.
14007 (vect_analyze_data_refs): Consider using gather loads if
14008 supports_vec_gather_load_p.
14009 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
14010 (vect_get_gather_scatter_offset_type): Likewise.
14011 (vect_convert_mask_for_vectype): Likewise.
14012 (vect_add_conversion_to_patterm): Likewise.
14013 (vect_try_gather_scatter_pattern): Likewise.
14014 (vect_recog_gather_scatter_pattern): New pattern recognizer.
14015 (vect_vect_recog_func_ptrs): Add it.
14016 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14017 internal_fn_mask_index and internal_gather_scatter_fn_p.
14018 (check_load_store_masking): Take the gather_scatter_info as an
14019 argument and handle gather loads.
14020 (vect_get_gather_scatter_ops): New function.
14021 (vectorizable_call): Check internal_load_fn_p.
14022 (vectorizable_load): Likewise. Handle gather load internal
14024 (vectorizable_store): Update call to check_load_store_masking.
14025 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
14026 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
14027 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
14028 (aarch64_gather_scale_operand_d): New predicates.
14029 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
14030 (mask_gather_load<mode>): New insns.
14032 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14033 Alan Hayward <alan.hayward@arm.com>
14034 David Sherwood <david.sherwood@arm.com>
14036 * optabs.def (fold_left_plus_optab): New optab.
14037 * doc/md.texi (fold_left_plus_@var{m}): Document.
14038 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
14039 * internal-fn.c (fold_left_direct): Define.
14040 (expand_fold_left_optab_fn): Likewise.
14041 (direct_fold_left_optab_supported_p): Likewise.
14042 * fold-const-call.c (fold_const_fold_left): New function.
14043 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
14044 * tree-parloops.c (valid_reduction_p): New function.
14045 (gather_scalar_reductions): Use it.
14046 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
14047 (vect_finish_replace_stmt): Declare.
14048 * tree-vect-loop.c (fold_left_reduction_fn): New function.
14049 (needs_fold_left_reduction_p): New function, split out from...
14050 (vect_is_simple_reduction): ...here. Accept reductions that
14051 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
14052 (vect_force_simple_reduction): Also store the reduction type in
14053 the assignment's STMT_VINFO_REDUC_TYPE.
14054 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
14055 (merge_with_identity): New function.
14056 (vect_expand_fold_left): Likewise.
14057 (vectorize_fold_left_reduction): Likewise.
14058 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
14059 scalar phi in place for it. Check for target support and reject
14060 cases that would reassociate the operation. Defer the transform
14061 phase to vectorize_fold_left_reduction.
14062 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
14063 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
14064 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
14066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14068 * tree-if-conv.c (predicate_mem_writes): Remove redundant
14069 call to ifc_temp_var.
14071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14072 Alan Hayward <alan.hayward@arm.com>
14073 David Sherwood <david.sherwood@arm.com>
14075 * target.def (legitimize_address_displacement): Take the original
14076 offset as a poly_int.
14077 * targhooks.h (default_legitimize_address_displacement): Update
14079 * targhooks.c (default_legitimize_address_displacement): Likewise.
14080 * doc/tm.texi: Regenerate.
14081 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
14082 as an argument, moving assert of ad->disp == ad->disp_term to...
14083 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
14084 Try calling targetm.legitimize_address_displacement before expanding
14085 the address rather than afterwards, and adjust for the new interface.
14086 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
14087 Match the new hook interface. Handle SVE addresses.
14088 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
14089 new hook interface.
14091 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14093 * Makefile.in (OBJS): Add early-remat.o.
14094 * target.def (select_early_remat_modes): New hook.
14095 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
14096 * doc/tm.texi: Regenerate.
14097 * targhooks.h (default_select_early_remat_modes): Declare.
14098 * targhooks.c (default_select_early_remat_modes): New function.
14099 * timevar.def (TV_EARLY_REMAT): New timevar.
14100 * passes.def (pass_early_remat): New pass.
14101 * tree-pass.h (make_pass_early_remat): Declare.
14102 * early-remat.c: New file.
14103 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
14105 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
14107 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14108 Alan Hayward <alan.hayward@arm.com>
14109 David Sherwood <david.sherwood@arm.com>
14111 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
14112 vfm1 with a bound_epilog parameter.
14113 (vect_do_peeling): Update calls accordingly, and move the prologue
14114 call earlier in the function. Treat the base bound_epilog as 0 for
14115 fully-masked loops and retain vf - 1 for other loops. Add 1 to
14116 this base when peeling for gaps.
14117 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
14118 with fully-masked loops.
14119 (vect_estimate_min_profitable_iters): Handle the single peeled
14120 iteration in that case.
14122 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14123 Alan Hayward <alan.hayward@arm.com>
14124 David Sherwood <david.sherwood@arm.com>
14126 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
14127 single-element interleaving even if the size is not a power of 2.
14128 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
14129 accesses for single-element interleaving if the group size is
14132 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14133 Alan Hayward <alan.hayward@arm.com>
14134 David Sherwood <david.sherwood@arm.com>
14136 * doc/md.texi (fold_extract_last_@var{m}): Document.
14137 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
14138 * optabs.def (fold_extract_last_optab): New optab.
14139 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
14140 * internal-fn.c (fold_extract_direct): New macro.
14141 (expand_fold_extract_optab_fn): Likewise.
14142 (direct_fold_extract_optab_supported_p): Likewise.
14143 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
14144 * tree-vect-loop.c (vect_model_reduction_cost): Handle
14145 EXTRACT_LAST_REDUCTION.
14146 (get_initial_def_for_reduction): Do not create an initial vector
14147 for EXTRACT_LAST_REDUCTION reductions.
14148 (vectorizable_reduction): Leave the scalar phi in place for
14149 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
14150 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
14151 epilogue code for EXTRACT_LAST_REDUCTION and defer the
14152 transform phase to vectorizable_condition.
14153 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
14155 (vect_finish_stmt_generation): ...here.
14156 (vect_finish_replace_stmt): New function.
14157 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
14158 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
14160 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
14162 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14163 Alan Hayward <alan.hayward@arm.com>
14164 David Sherwood <david.sherwood@arm.com>
14166 * doc/md.texi (extract_last_@var{m}): Document.
14167 * optabs.def (extract_last_optab): New optab.
14168 * internal-fn.def (EXTRACT_LAST): New internal function.
14169 * internal-fn.c (cond_unary_direct): New macro.
14170 (expand_cond_unary_optab_fn): Likewise.
14171 (direct_cond_unary_optab_supported_p): Likewise.
14172 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
14173 loops using EXTRACT_LAST.
14174 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
14175 (extract_last_<mode>): ...this optab.
14176 (vec_extract<mode><Vel>): Update accordingly.
14178 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14179 Alan Hayward <alan.hayward@arm.com>
14180 David Sherwood <david.sherwood@arm.com>
14182 * target.def (empty_mask_is_expensive): New hook.
14183 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
14184 * doc/tm.texi: Regenerate.
14185 * targhooks.h (default_empty_mask_is_expensive): Declare.
14186 * targhooks.c (default_empty_mask_is_expensive): New function.
14187 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
14188 if the target says that empty masks are expensive.
14189 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
14191 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
14193 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14194 Alan Hayward <alan.hayward@arm.com>
14195 David Sherwood <david.sherwood@arm.com>
14197 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
14198 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
14199 (vect_use_loop_mask_for_alignment_p): New function.
14200 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
14201 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
14202 niters_skip argument. Make sure that the first niters_skip elements
14203 of the first iteration are inactive.
14204 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
14205 Update call to vect_set_loop_masks_directly.
14206 (get_misalign_in_elems): New function, split out from...
14207 (vect_gen_prolog_loop_niters): ...here.
14208 (vect_update_init_of_dr): Take a code argument that specifies whether
14209 the adjustment should be added or subtracted.
14210 (vect_update_init_of_drs): Likewise.
14211 (vect_prepare_for_masked_peels): New function.
14212 (vect_do_peeling): Skip prologue peeling if we're using a mask
14213 instead. Update call to vect_update_inits_of_drs.
14214 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14216 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
14217 alignment. Do not include the number of peeled iterations in
14218 the minimum threshold in that case.
14219 (vectorizable_induction): Adjust the start value down by
14220 LOOP_VINFO_MASK_SKIP_NITERS iterations.
14221 (vect_transform_loop): Call vect_prepare_for_masked_peels.
14222 Take the number of skipped iterations into account when calculating
14224 * tree-vect-stmts.c (vect_gen_while_not): New function.
14226 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14227 Alan Hayward <alan.hayward@arm.com>
14228 David Sherwood <david.sherwood@arm.com>
14230 * doc/sourcebuild.texi (vect_fully_masked): Document.
14231 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
14232 default value to 0.
14233 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
14235 (vect_analyze_loop_2): ...here. Don't check the vectorization
14236 factor against the number of loop iterations if the loop is
14239 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14240 Alan Hayward <alan.hayward@arm.com>
14241 David Sherwood <david.sherwood@arm.com>
14243 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
14244 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
14245 (dump_groups): Update accordingly.
14246 (iv_use::mem_type): New member variable.
14247 (address_p): New function.
14248 (record_use): Add a mem_type argument and initialize the new
14250 (record_group_use): Add a mem_type argument. Use address_p.
14251 Remove obsolete null checks of base_object. Update call to record_use.
14252 (find_interesting_uses_op): Update call to record_group_use.
14253 (find_interesting_uses_cond): Likewise.
14254 (find_interesting_uses_address): Likewise.
14255 (get_mem_type_for_internal_fn): New function.
14256 (find_address_like_use): Likewise.
14257 (find_interesting_uses_stmt): Try find_address_like_use before
14258 calling find_interesting_uses_op.
14259 (addr_offset_valid_p): Use the iv mem_type field as the type
14260 of the addressed memory.
14261 (add_autoinc_candidates): Likewise.
14262 (get_address_cost): Likewise.
14263 (split_small_address_groups_p): Use address_p.
14264 (split_address_groups): Likewise.
14265 (add_iv_candidate_for_use): Likewise.
14266 (autoinc_possible_for_pair): Likewise.
14267 (rewrite_groups): Likewise.
14268 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
14269 (determine_group_iv_cost): Update after split of USE_ADDRESS.
14270 (get_alias_ptr_type_for_ptr_address): New function.
14271 (rewrite_use_address): Rewrite address uses in calls that were
14272 identified by find_address_like_use.
14274 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14275 Alan Hayward <alan.hayward@arm.com>
14276 David Sherwood <david.sherwood@arm.com>
14278 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
14280 * gimple-expr.h (is_gimple_addressable: Likewise.
14281 * gimple-expr.c (is_gimple_address): Likewise.
14282 * internal-fn.c (expand_call_mem_ref): New function.
14283 (expand_mask_load_optab_fn): Use it.
14284 (expand_mask_store_optab_fn): Likewise.
14286 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14287 Alan Hayward <alan.hayward@arm.com>
14288 David Sherwood <david.sherwood@arm.com>
14290 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
14291 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
14292 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
14293 (cond_umax@var{mode}): Document.
14294 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
14295 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
14296 (cond_umin_optab, cond_umax_optab): New optabs.
14297 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
14298 (COND_IOR, COND_XOR): New internal functions.
14299 * internal-fn.h (get_conditional_internal_fn): Declare.
14300 * internal-fn.c (cond_binary_direct): New macro.
14301 (expand_cond_binary_optab_fn): Likewise.
14302 (direct_cond_binary_optab_supported_p): Likewise.
14303 (get_conditional_internal_fn): New function.
14304 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
14305 Cope with reduction statements that are vectorized as calls rather
14307 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
14308 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
14309 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
14310 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14311 (UNSPEC_COND_EOR): New unspecs.
14312 (optab): Add mappings for them.
14313 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
14314 (sve_int_op, sve_fp_op): New int attributes.
14316 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14317 Alan Hayward <alan.hayward@arm.com>
14318 David Sherwood <david.sherwood@arm.com>
14320 * optabs.def (while_ult_optab): New optab.
14321 * doc/md.texi (while_ult@var{m}@var{n}): Document.
14322 * internal-fn.def (WHILE_ULT): New internal function.
14323 * internal-fn.h (direct_internal_fn_supported_p): New override
14324 that takes two types as argument.
14325 * internal-fn.c (while_direct): New macro.
14326 (expand_while_optab_fn): New function.
14327 (convert_optab_supported_p): Likewise.
14328 (direct_while_optab_supported_p): New macro.
14329 * wide-int.h (wi::udiv_ceil): New function.
14330 * tree-vectorizer.h (rgroup_masks): New structure.
14331 (vec_loop_masks): New typedef.
14332 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
14333 and fully_masked_p.
14334 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
14335 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
14336 (vect_max_vf): New function.
14337 (slpeel_make_loop_iterate_ntimes): Delete.
14338 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
14339 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
14340 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
14341 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
14342 internal-fn.h, stor-layout.h and optabs-query.h.
14343 (vect_set_loop_mask): New function.
14344 (add_preheader_seq): Likewise.
14345 (add_header_seq): Likewise.
14346 (interleave_supported_p): Likewise.
14347 (vect_maybe_permute_loop_masks): Likewise.
14348 (vect_set_loop_masks_directly): Likewise.
14349 (vect_set_loop_condition_masked): Likewise.
14350 (vect_set_loop_condition_unmasked): New function, split out from
14351 slpeel_make_loop_iterate_ntimes.
14352 (slpeel_make_loop_iterate_ntimes): Rename to..
14353 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
14354 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
14355 (vect_do_peeling): Update call accordingly.
14356 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
14358 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14359 mask_compare_type, can_fully_mask_p and fully_masked_p.
14360 (release_vec_loop_masks): New function.
14361 (_loop_vec_info): Use it to free the loop masks.
14362 (can_produce_all_loop_masks_p): New function.
14363 (vect_get_max_nscalars_per_iter): Likewise.
14364 (vect_verify_full_masking): Likewise.
14365 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
14366 retries, and free the mask rgroups before retrying. Check loop-wide
14367 reasons for disallowing fully-masked loops. Make the final decision
14368 about whether use a fully-masked loop or not.
14369 (vect_estimate_min_profitable_iters): Do not assume that peeling
14370 for the number of iterations will be needed for fully-masked loops.
14371 (vectorizable_reduction): Disable fully-masked loops.
14372 (vectorizable_live_operation): Likewise.
14373 (vect_halve_mask_nunits): New function.
14374 (vect_double_mask_nunits): Likewise.
14375 (vect_record_loop_mask): Likewise.
14376 (vect_get_loop_mask): Likewise.
14377 (vect_transform_loop): Handle the case in which the final loop
14378 iteration might handle a partial vector. Call vect_set_loop_condition
14379 instead of slpeel_make_loop_iterate_ntimes.
14380 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
14381 (check_load_store_masking): New function.
14382 (prepare_load_store_mask): Likewise.
14383 (vectorizable_store): Handle fully-masked loops.
14384 (vectorizable_load): Likewise.
14385 (supportable_widening_operation): Use vect_halve_mask_nunits for
14387 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
14388 (vect_gen_while): New function.
14389 * config/aarch64/aarch64.md (umax<mode>3): New expander.
14390 (aarch64_uqdec<mode>): New insn.
14392 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14393 Alan Hayward <alan.hayward@arm.com>
14394 David Sherwood <david.sherwood@arm.com>
14396 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
14397 (reduc_xor_scal_optab): New optabs.
14398 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
14399 (reduc_xor_scal_@var{m}): Document.
14400 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
14401 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
14402 internal functions.
14403 * fold-const-call.c (fold_const_call): Handle them.
14404 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
14405 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
14406 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
14407 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
14408 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
14409 (UNSPEC_XORV): New unspecs.
14410 (optab): Add entries for them.
14411 (BITWISEV): New int iterator.
14412 (bit_reduc_op): New int attributes.
14414 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14415 Alan Hayward <alan.hayward@arm.com>
14416 David Sherwood <david.sherwood@arm.com>
14418 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
14419 * internal-fn.def (VEC_SHL_INSERT): New internal function.
14420 * optabs.def (vec_shl_insert_optab): New optab.
14421 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
14422 (duplicate_and_interleave): Likewise.
14423 * tree-vect-loop.c: Include internal-fn.h.
14424 (neutral_op_for_slp_reduction): New function, split out from
14425 get_initial_defs_for_reduction.
14426 (get_initial_def_for_reduction): Handle option 2 for variable-length
14427 vectors by loading the neutral value into a vector and then shifting
14428 the initial value into element 0.
14429 (get_initial_defs_for_reduction): Replace the code argument with
14430 the neutral value calculated by neutral_op_for_slp_reduction.
14431 Use gimple_build_vector for constant-length vectors.
14432 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
14433 but the first group_size elements have a neutral value.
14434 Use duplicate_and_interleave otherwise.
14435 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
14436 Update call to get_initial_defs_for_reduction. Handle SLP
14437 reductions for variable-length vectors by creating one vector
14438 result for each scalar result, with the elements associated
14439 with other scalar results stubbed out with the neutral value.
14440 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
14441 Require IFN_VEC_SHL_INSERT for double reductions on
14442 variable-length vectors, or SLP reductions that have
14443 a neutral value. Require can_duplicate_and_interleave_p
14444 support for variable-length unchained SLP reductions if there
14445 is no neutral value, such as for MIN/MAX reductions. Also require
14446 the number of vector elements to be a multiple of the number of
14447 SLP statements when doing variable-length unchained SLP reductions.
14448 Update call to vect_create_epilog_for_reduction.
14449 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
14450 and remove initial values.
14451 (duplicate_and_interleave): Make public.
14452 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
14453 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
14455 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14456 Alan Hayward <alan.hayward@arm.com>
14457 David Sherwood <david.sherwood@arm.com>
14459 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
14460 (can_duplicate_and_interleave_p): New function.
14461 (vect_get_and_check_slp_defs): Take the vector of statements
14462 rather than just the current one. Remove excess parentheses.
14463 Restriction rejectinon of vect_constant_def and vect_external_def
14464 for variable-length vectors to boolean types, or types for which
14465 can_duplicate_and_interleave_p is false.
14466 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
14467 (duplicate_and_interleave): New function.
14468 (vect_get_constant_vectors): Use gimple_build_vector for
14469 constant-length vectors and suitable variable-length constant
14470 vectors. Use duplicate_and_interleave for other variable-length
14471 vectors. Don't defer the update when inserting new statements.
14473 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14474 Alan Hayward <alan.hayward@arm.com>
14475 David Sherwood <david.sherwood@arm.com>
14477 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
14478 min_profitable_iters doesn't go negative.
14480 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14481 Alan Hayward <alan.hayward@arm.com>
14482 David Sherwood <david.sherwood@arm.com>
14484 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
14485 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
14486 * optabs.def (vec_mask_load_lanes_optab): New optab.
14487 (vec_mask_store_lanes_optab): Likewise.
14488 * internal-fn.def (MASK_LOAD_LANES): New internal function.
14489 (MASK_STORE_LANES): Likewise.
14490 * internal-fn.c (mask_load_lanes_direct): New macro.
14491 (mask_store_lanes_direct): Likewise.
14492 (expand_mask_load_optab_fn): Handle masked operations.
14493 (expand_mask_load_lanes_optab_fn): New macro.
14494 (expand_mask_store_optab_fn): Handle masked operations.
14495 (expand_mask_store_lanes_optab_fn): New macro.
14496 (direct_mask_load_lanes_optab_supported_p): Likewise.
14497 (direct_mask_store_lanes_optab_supported_p): Likewise.
14498 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
14500 (vect_load_lanes_supported): Likewise.
14501 * tree-vect-data-refs.c (strip_conversion): New function.
14502 (can_group_stmts_p): Likewise.
14503 (vect_analyze_data_ref_accesses): Use it instead of checking
14504 for a pair of assignments.
14505 (vect_store_lanes_supported): Take a masked_p parameter.
14506 (vect_load_lanes_supported): Likewise.
14507 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
14508 vect_store_lanes_supported and vect_load_lanes_supported.
14509 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14510 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
14511 parameter. Don't allow gaps for masked accesses.
14512 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
14513 and vect_load_lanes_supported.
14514 (get_load_store_type): Take a masked_p parameter and update
14515 call to get_group_load_store_type.
14516 (vectorizable_store): Update call to get_load_store_type.
14517 Handle IFN_MASK_STORE_LANES.
14518 (vectorizable_load): Update call to get_load_store_type.
14519 Handle IFN_MASK_LOAD_LANES.
14521 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14522 Alan Hayward <alan.hayward@arm.com>
14523 David Sherwood <david.sherwood@arm.com>
14525 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14527 * config/aarch64/aarch64-protos.h
14528 (aarch64_sve_struct_memory_operand_p): Declare.
14529 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14530 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14531 (VPRED, vpred): Handle SVE structure modes.
14532 * config/aarch64/constraints.md (Utx): New constraint.
14533 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14534 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14535 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
14536 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
14537 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
14538 structure modes. Split into pieces after RA.
14539 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
14540 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
14542 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
14543 SVE structure modes.
14544 (aarch64_classify_address): Likewise.
14545 (sizetochar): Move earlier in file.
14546 (aarch64_print_operand): Handle SVE register lists.
14547 (aarch64_array_mode): New function.
14548 (aarch64_sve_struct_memory_operand_p): Likewise.
14549 (TARGET_ARRAY_MODE): Redefine.
14551 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14552 Alan Hayward <alan.hayward@arm.com>
14553 David Sherwood <david.sherwood@arm.com>
14555 * target.def (array_mode): New target hook.
14556 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14557 * doc/tm.texi: Regenerate.
14558 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14559 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14560 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14561 targetm.array_mode.
14562 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14566 Alan Hayward <alan.hayward@arm.com>
14567 David Sherwood <david.sherwood@arm.com>
14569 * fold-const.c (fold_binary_loc): Check the argument types
14570 rather than the result type when testing for a vector operation.
14572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14574 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14575 * doc/tm.texi: Regenerate.
14577 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14578 Alan Hayward <alan.hayward@arm.com>
14579 David Sherwood <david.sherwood@arm.com>
14581 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14582 (sve): Document new AArch64 extension.
14583 * doc/md.texi (w): Extend the description of the AArch64
14584 constraint to include SVE vectors.
14585 (Upl, Upa): Document new AArch64 predicate constraints.
14586 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14588 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14589 (msve-vector-bits=): New option.
14590 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14591 SVE when these are disabled.
14592 (sve): New extension.
14593 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14594 modes. Adjust their number of units based on aarch64_sve_vg.
14595 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14596 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14597 aarch64_addr_query_type.
14598 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14599 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14600 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14601 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14602 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14603 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14604 (aarch64_simd_imm_zero_p): Delete.
14605 (aarch64_check_zero_based_sve_index_immediate): Declare.
14606 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14607 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14608 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14609 (aarch64_sve_float_mul_immediate_p): Likewise.
14610 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14611 rather than an rtx.
14612 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14613 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14614 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14615 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14616 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14617 (aarch64_regmode_natural_size): Likewise.
14618 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14619 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14621 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14622 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14623 for VG and the SVE predicate registers.
14624 (V_ALIASES): Add a "z"-prefixed alias.
14625 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14626 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14627 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14628 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14629 (REG_CLASS_NAMES): Add entries for them.
14630 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14631 and the predicate registers.
14632 (aarch64_sve_vg): Declare.
14633 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14634 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14635 (REGMODE_NATURAL_SIZE): Define.
14636 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14638 * config/aarch64/aarch64.c: Include cfgrtl.h.
14639 (simd_immediate_info): Add a constructor for series vectors,
14640 and an associated step field.
14641 (aarch64_sve_vg): New variable.
14642 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14643 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14644 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14645 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14646 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14647 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14648 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14649 (aarch64_get_mask_mode): New functions.
14650 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14651 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14652 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14653 predicate modes and predicate registers. Explicitly restrict
14654 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14655 to store a vector mode if it is recognized by
14656 aarch64_classify_vector_mode.
14657 (aarch64_regmode_natural_size): New function.
14658 (aarch64_hard_regno_caller_save_mode): Return the original mode
14660 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14661 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14662 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14663 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14665 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14666 does not overlap dest if the function is frame-related. Handle
14668 (aarch64_split_add_offset): New function.
14669 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14670 them aarch64_add_offset.
14671 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14672 and update call to aarch64_sub_sp.
14673 (aarch64_add_cfa_expression): New function.
14674 (aarch64_expand_prologue): Pass extra temporary registers to the
14675 functions above. Handle the case in which we need to emit new
14676 DW_CFA_expressions for registers that were originally saved
14677 relative to the stack pointer, but now have to be expressed
14678 relative to the frame pointer.
14679 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14681 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14682 IP0 and IP1 values for SVE frames.
14683 (aarch64_expand_vec_series): New function.
14684 (aarch64_expand_sve_widened_duplicate): Likewise.
14685 (aarch64_expand_sve_const_vector): Likewise.
14686 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14687 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14688 into the register, rather than emitting a SET directly.
14689 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14690 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14691 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14692 (offset_9bit_signed_scaled_p): New functions.
14693 (aarch64_replicate_bitmask_imm): New function.
14694 (aarch64_bitmask_imm): Use it.
14695 (aarch64_cannot_force_const_mem): Reject expressions involving
14696 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14697 (aarch64_classify_index): Handle SVE indices, by requiring
14698 a plain register index with a scale that matches the element size.
14699 (aarch64_classify_address): Handle SVE addresses. Assert that
14700 the mode of the address is VOIDmode or an integer mode.
14701 Update call to aarch64_classify_symbol.
14702 (aarch64_classify_symbolic_expression): Update call to
14703 aarch64_classify_symbol.
14704 (aarch64_const_vec_all_in_range_p): New function.
14705 (aarch64_print_vector_float_operand): Likewise.
14706 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14707 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14708 and the FP immediates 1.0 and 0.5.
14709 (aarch64_print_address_internal): Handle SVE addresses.
14710 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14711 (aarch64_regno_regclass): Handle predicate registers.
14712 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14714 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14715 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14716 (aarch64_convert_sve_vector_bits): New function.
14717 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14718 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14719 rather than an rtx.
14720 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14721 Handle SVE vector and predicate modes. Accept VL-based constants
14722 that need only one temporary register, and VL offsets that require
14723 no temporary registers.
14724 (aarch64_conditional_register_usage): Mark the predicate registers
14725 as fixed if SVE isn't available.
14726 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14727 Return true for SVE vector and predicate modes.
14728 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14729 rather than an unsigned int. Handle SVE modes.
14730 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14732 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14734 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14735 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14736 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14737 (aarch64_sve_float_mul_immediate_p): New functions.
14738 (aarch64_sve_valid_immediate): New function.
14739 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14740 Explicitly reject structure modes. Check for INDEX constants.
14741 Handle PTRUE and PFALSE constants.
14742 (aarch64_check_zero_based_sve_index_immediate): New function.
14743 (aarch64_simd_imm_zero_p): Delete.
14744 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14745 vector modes. Accept constants in the range of CNT[BHWD].
14746 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14747 ask for an Advanced SIMD mode.
14748 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14749 (aarch64_simd_vector_alignment): Handle SVE predicates.
14750 (aarch64_vectorize_preferred_vector_alignment): New function.
14751 (aarch64_simd_vector_alignment_reachable): Use it instead of
14753 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14754 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14756 (MAX_VECT_LEN): Delete.
14757 (expand_vec_perm_d): Add a vec_flags field.
14758 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14759 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14760 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14762 (aarch64_evpc_rev): Rename to...
14763 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14764 (aarch64_evpc_rev_global): New function.
14765 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14766 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14768 (aarch64_evpc_sve_tbl): New function.
14769 (aarch64_expand_vec_perm_const_1): Update after rename of
14770 aarch64_evpc_rev. Handle SVE permutes too, trying
14771 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14772 than aarch64_evpc_tbl.
14773 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14774 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14775 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14776 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14777 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14778 (aarch64_expand_sve_vcond): New functions.
14779 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14780 of aarch64_vector_mode_p.
14781 (aarch64_dwarf_poly_indeterminate_value): New function.
14782 (aarch64_compute_pressure_classes): Likewise.
14783 (aarch64_can_change_mode_class): Likewise.
14784 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14785 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14786 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14787 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14788 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14789 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14790 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14791 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14793 (Dn, Dl, Dr): Accept const as well as const_vector.
14794 (Dz): Likewise. Compare against CONST0_RTX.
14795 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14796 of "vector" where appropriate.
14797 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14798 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14799 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14800 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14801 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14802 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14803 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14804 (v_int_equiv): Extend to SVE modes.
14805 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14807 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14808 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14809 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14810 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14811 (SVE_COND_FP_CMP): New int iterators.
14812 (perm_hilo): Handle the new unpack unspecs.
14813 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14815 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14816 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14817 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14818 (aarch64_equality_operator, aarch64_constant_vector_operand)
14819 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14820 (aarch64_sve_nonimmediate_operand): Likewise.
14821 (aarch64_sve_general_operand): Likewise.
14822 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14823 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14824 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14825 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14826 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14827 (aarch64_sve_float_arith_immediate): Likewise.
14828 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14829 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14830 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14831 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14832 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14833 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14834 (aarch64_sve_float_arith_operand): Likewise.
14835 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14836 (aarch64_sve_float_mul_operand): Likewise.
14837 (aarch64_sve_vec_perm_operand): Likewise.
14838 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14839 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14840 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14841 as well as const_vector.
14842 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14843 in file. Use CONST0_RTX and CONSTM1_RTX.
14844 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14845 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14846 Use aarch64_simd_imm_zero.
14847 * config/aarch64/aarch64-sve.md: New file.
14848 * config/aarch64/aarch64.md: Include it.
14849 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14850 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14851 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14852 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14853 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14854 (sve): New attribute.
14855 (enabled): Disable instructions with the sve attribute unless
14857 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14858 aarch64_expand_mov_immediate.
14859 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14860 CNT[BHSD] immediates.
14861 (movti): Split CONST_POLY_INT moves into two halves.
14862 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14863 Split additions that need a temporary here if the destination
14864 is the stack pointer.
14865 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14866 (*add<mode>3_poly_1): New instruction.
14867 (set_clobber_cc): New expander.
14869 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14871 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14872 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14873 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14874 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14875 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14876 Change innermode from fixed_mode_size to machine_mode.
14877 (simplify_subreg): Update call accordingly. Handle a constant-sized
14878 subreg of a variable-length CONST_VECTOR.
14880 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14881 Alan Hayward <alan.hayward@arm.com>
14882 David Sherwood <david.sherwood@arm.com>
14884 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14885 (add_offset_to_base): New function, split out from...
14886 (create_mem_ref): ...here. When handling a scale other than 1,
14887 check first whether the address is valid without the offset.
14888 Add it into the base if so, leaving the index and scale as-is.
14890 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14893 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14894 fold_for_warn before checking if arg2 is INTEGER_CST.
14896 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14898 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14899 (store_multiple_operation): Delete.
14900 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14901 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14902 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14903 guarded by TARGET_STRING.
14904 (rs6000_output_load_multiple): Delete.
14905 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14906 OPTION_MASK_STRING / TARGET_STRING handling.
14907 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14908 (const rs6000_opt_masks) <"string">: Change mask to 0.
14909 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14910 (MASK_STRING): Delete.
14911 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14913 (load_multiple): Delete.
14920 (store_multiple): Delete.
14927 (movmemsi_8reg): Delete.
14928 (corresponding unnamed define_insn): Delete.
14929 (movmemsi_6reg): Delete.
14930 (corresponding unnamed define_insn): Delete.
14931 (movmemsi_4reg): Delete.
14932 (corresponding unnamed define_insn): Delete.
14933 (movmemsi_2reg): Delete.
14934 (corresponding unnamed define_insn): Delete.
14935 (movmemsi_1reg): Delete.
14936 (corresponding unnamed define_insn): Delete.
14937 * config/rs6000/rs6000.opt (mno-string): New.
14938 (mstring): Replace by deprecation warning stub.
14939 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14941 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14943 * regrename.c (regrename_do_replace): If replacing the same
14944 reg multiple times, try to reuse last created gen_raw_REG.
14947 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14948 main to workaround a bug in GDB.
14950 2018-01-12 Tom de Vries <tom@codesourcery.com>
14953 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14955 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14957 PR rtl-optimization/80481
14958 * ira-color.c (get_cap_member): New function.
14959 (allocnos_conflict_by_live_ranges_p): Use it.
14960 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14961 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14963 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14966 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14967 (*saddl_se_1): Ditto.
14968 (*ssubsi_1): Ditto.
14969 (*ssubl_se_1): Ditto.
14971 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14973 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14974 rather than wi::to_widest for DR_INITs.
14975 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14976 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14977 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14979 (vect_analyze_group_access_1): Note that here.
14981 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14983 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14984 polynomial type sizes.
14986 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14988 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14989 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14990 (gimple_add_tmp_var): Likewise.
14992 2018-01-12 Martin Liska <mliska@suse.cz>
14994 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14995 (gimple_alloc_sizes): Likewise.
14996 (dump_gimple_statistics): Use PRIu64 in printf format.
14997 * gimple.h: Change uint64_t to int.
14999 2018-01-12 Martin Liska <mliska@suse.cz>
15001 * tree-core.h: Use uint64_t instead of int.
15002 * tree.c (tree_node_counts): Likewise.
15003 (tree_node_sizes): Likewise.
15004 (dump_tree_statistics): Use PRIu64 in printf format.
15006 2018-01-12 Martin Liska <mliska@suse.cz>
15008 * Makefile.in: As qsort_chk is implemented in vec.c, add
15009 vec.o to linkage of gencfn-macros.
15010 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
15011 passing the info to record_node_allocation_statistics.
15012 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
15014 * ggc-common.c (struct ggc_usage): Add operator== and use
15015 it in operator< and compare function.
15016 * mem-stats.h (struct mem_usage): Likewise.
15017 * vec.c (struct vec_usage): Remove operator< and compare
15018 function. Can be simply inherited.
15020 2018-01-12 Martin Jambor <mjambor@suse.cz>
15023 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
15024 * tree-ssa-math-opts.c: Include domwalk.h.
15025 (convert_mult_to_fma_1): New function.
15026 (fma_transformation_info): New type.
15027 (fma_deferring_state): Likewise.
15028 (cancel_fma_deferring): New function.
15029 (result_of_phi): Likewise.
15030 (last_fma_candidate_feeds_initial_phi): Likewise.
15031 (convert_mult_to_fma): Added deferring logic, split actual
15032 transformation to convert_mult_to_fma_1.
15033 (math_opts_dom_walker): New type.
15034 (math_opts_dom_walker::after_dom_children): New method, body moved
15035 here from pass_optimize_widening_mul::execute, added deferring logic
15037 (pass_optimize_widening_mul::execute): Moved most of code to
15038 math_opts_dom_walker::after_dom_children.
15039 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
15040 * config/i386/i386.c (ix86_option_override_internal): Added
15041 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
15043 2018-01-12 Richard Biener <rguenther@suse.de>
15046 * dwarf2out.c (gen_variable_die): Do not reset old_die for
15047 inline instance vars.
15049 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
15052 * config/rx/rx.c (rx_is_restricted_memory_address):
15053 Handle SUBREG case.
15055 2018-01-12 Richard Biener <rguenther@suse.de>
15057 PR tree-optimization/80846
15058 * target.def (split_reduction): New target hook.
15059 * targhooks.c (default_split_reduction): New function.
15060 * targhooks.h (default_split_reduction): Declare.
15061 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
15062 target requests first reduce vectors by combining low and high
15064 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
15065 (get_vectype_for_scalar_type_and_size): Export.
15066 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
15067 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
15068 * doc/tm.texi: Regenerate.
15069 * config/i386/i386.c (ix86_split_reduction): Implement
15070 TARGET_VECTORIZE_SPLIT_REDUCTION.
15072 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15075 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
15076 in PIC mode except for TARGET_VXWORKS_RTP.
15077 * config/sparc/sparc.c: Include cfgrtl.h.
15078 (TARGET_INIT_PIC_REG): Define.
15079 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
15080 (sparc_pic_register_p): New predicate.
15081 (sparc_legitimate_address_p): Use it.
15082 (sparc_legitimize_pic_address): Likewise.
15083 (sparc_delegitimize_address): Likewise.
15084 (sparc_mode_dependent_address_p): Likewise.
15085 (gen_load_pcrel_sym): Remove 4th parameter.
15086 (load_got_register): Adjust call to above. Remove obsolete stuff.
15087 (sparc_expand_prologue): Do not call load_got_register here.
15088 (sparc_flat_expand_prologue): Likewise.
15089 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
15090 (sparc_use_pseudo_pic_reg): New function.
15091 (sparc_init_pic_reg): Likewise.
15092 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15093 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
15095 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
15097 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
15098 Add item for branch_cost.
15100 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15102 PR rtl-optimization/83565
15103 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
15104 not extend the result to a larger mode for rotate operations.
15105 (num_sign_bit_copies1): Likewise.
15107 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15110 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
15112 Use values-Xc.o for -pedantic.
15113 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
15115 2018-01-12 Martin Liska <mliska@suse.cz>
15118 * ipa-devirt.c (final_warning_record::grow_type_warnings):
15120 (possible_polymorphic_call_targets): Use it.
15121 (ipa_devirt): Likewise.
15123 2018-01-12 Martin Liska <mliska@suse.cz>
15125 * profile-count.h (enum profile_quality): Use 0 as invalid
15126 enum value of profile_quality.
15128 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
15130 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
15131 -mext-string options.
15133 2018-01-12 Richard Biener <rguenther@suse.de>
15135 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
15136 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
15137 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
15139 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
15141 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15143 * configure.ac (--with-long-double-format): Add support for the
15144 configuration option to change the default long double format on
15146 * config.gcc (powerpc*-linux*-*): Likewise.
15147 * configure: Regenerate.
15148 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
15149 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
15150 used without modification.
15152 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15154 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
15155 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
15156 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
15157 MISC_BUILTIN_SPEC_BARRIER.
15158 (rs6000_init_builtins): Likewise.
15159 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
15161 (speculation_barrier): New define_insn.
15162 * doc/extend.texi: Document __builtin_speculation_barrier.
15164 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15167 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
15168 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
15169 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
15171 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
15172 integral modes instead of "ss" and "sd".
15173 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
15174 vectors with 32-bit and 64-bit elements.
15175 (vecdupssescalarmodesuffix): New mode attribute.
15176 (vec_dup<mode>): Use it.
15178 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
15181 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
15182 frame if argument is passed on stack.
15184 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15187 * ree.c (combine_reaching_defs): Optimize also
15188 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
15189 reg2=any_extend(exp); reg1=reg2;, formatting fix.
15191 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15193 PR middle-end/83189
15194 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
15196 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15198 PR middle-end/83718
15199 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
15200 after they are computed.
15202 2018-01-11 Bin Cheng <bin.cheng@arm.com>
15204 PR tree-optimization/83695
15205 * gimple-loop-linterchange.cc
15206 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
15207 reset cached scev information after interchange.
15208 (pass_linterchange::execute): Remove call to scev_reset_htab.
15210 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15212 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
15213 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
15214 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
15215 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
15216 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
15217 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
15218 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
15219 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
15220 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
15221 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
15222 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
15223 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
15224 (V_lane_reg): Likewise.
15225 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
15227 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
15228 (vfmal_lane_low<mode>_intrinsic,
15229 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
15230 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
15231 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
15232 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
15233 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
15234 vfmsl_lane_high<mode>_intrinsic): New define_insns.
15236 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15238 * config/arm/arm-cpus.in (fp16fml): New feature.
15239 (ALL_SIMD): Add fp16fml.
15240 (armv8.2-a): Add fp16fml as an option.
15241 (armv8.3-a): Likewise.
15242 (armv8.4-a): Add fp16fml as part of fp16.
15243 * config/arm/arm.h (TARGET_FP16FML): Define.
15244 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
15246 * config/arm/arm-modes.def (V2HF): Define.
15247 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
15248 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
15249 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
15250 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
15251 vfmsl_low, vfmsl_high): New set of builtins.
15252 * config/arm/iterators.md (PLUSMINUS): New code iterator.
15253 (vfml_op): New code attribute.
15254 (VFMLHALVES): New int iterator.
15255 (VFML, VFMLSEL): New mode attributes.
15256 (V_reg): Define mapping for V2HF.
15257 (V_hi, V_lo): New mode attributes.
15258 (VF_constraint): Likewise.
15259 (vfml_half, vfml_half_selector): New int attributes.
15260 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
15262 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
15263 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
15265 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
15266 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
15267 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
15268 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
15270 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
15271 Document new effective target and option set.
15273 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15275 * config/arm/arm-cpus.in (armv8_4): New feature.
15276 (ARMv8_4a): New fgroup.
15277 (armv8.4-a): New arch.
15278 * config/arm/arm-tables.opt: Regenerate.
15279 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
15280 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
15281 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
15282 Add matching rules for -march=armv8.4-a and extensions.
15283 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
15285 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
15288 * config/rx/rx.md (BW): New mode attribute.
15289 (sync_lock_test_and_setsi): Add mode suffix to insn output.
15291 2018-01-11 Richard Biener <rguenther@suse.de>
15293 PR tree-optimization/83435
15294 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
15295 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
15296 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
15298 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15299 Alan Hayward <alan.hayward@arm.com>
15300 David Sherwood <david.sherwood@arm.com>
15302 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
15304 (aarch64_classify_address): Initialize it. Track polynomial offsets.
15305 (aarch64_print_address_internal): Use it to check for a zero offset.
15307 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15308 Alan Hayward <alan.hayward@arm.com>
15309 David Sherwood <david.sherwood@arm.com>
15311 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
15312 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
15313 Return a poly_int64 rather than a HOST_WIDE_INT.
15314 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
15315 rather than a HOST_WIDE_INT.
15316 * config/aarch64/aarch64.h (aarch64_frame): Protect with
15317 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
15318 hard_fp_offset, frame_size, initial_adjust, callee_offset and
15319 final_offset from HOST_WIDE_INT to poly_int64.
15320 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
15321 to_constant when getting the number of units in an Advanced SIMD
15323 (aarch64_builtin_vectorized_function): Check for a constant number
15325 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
15327 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
15328 attribute instead of GET_MODE_NUNITS.
15329 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
15330 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
15331 GET_MODE_SIZE for fixed-size registers.
15332 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
15333 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
15334 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
15335 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
15336 (aarch64_print_operand, aarch64_print_address_internal)
15337 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
15338 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
15339 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
15340 Handle polynomial GET_MODE_SIZE.
15341 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
15342 wider than SImode without modification.
15343 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
15344 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
15345 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
15346 passing and returning SVE modes.
15347 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
15348 rather than GEN_INT.
15349 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
15350 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
15351 (aarch64_allocate_and_probe_stack_space): Likewise.
15352 (aarch64_layout_frame): Cope with polynomial offsets.
15353 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
15354 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
15355 polynomial offsets.
15356 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
15357 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
15358 poly_int64 rather than a HOST_WIDE_INT.
15359 (aarch64_get_separate_components, aarch64_process_components)
15360 (aarch64_expand_prologue, aarch64_expand_epilogue)
15361 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
15362 (aarch64_anchor_offset): New function, split out from...
15363 (aarch64_legitimize_address): ...here.
15364 (aarch64_builtin_vectorization_cost): Handle polynomial
15365 TYPE_VECTOR_SUBPARTS.
15366 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
15368 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
15369 number of elements from the PARALLEL rather than the mode.
15370 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
15371 rather than GET_MODE_BITSIZE.
15372 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
15373 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
15374 (aarch64_expand_vec_perm_const_1): Handle polynomial
15375 d->perm.length () and d->perm elements.
15376 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
15377 Apply to_constant to d->perm elements.
15378 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
15379 polynomial CONST_VECTOR_NUNITS.
15380 (aarch64_move_pointer): Take amount as a poly_int64 rather
15382 (aarch64_progress_pointer): Avoid temporary variable.
15383 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
15384 the mode attribute instead of GET_MODE.
15386 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15387 Alan Hayward <alan.hayward@arm.com>
15388 David Sherwood <david.sherwood@arm.com>
15390 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
15391 x exists before using it.
15392 (aarch64_add_constant_internal): Rename to...
15393 (aarch64_add_offset_1): ...this. Replace regnum with separate
15394 src and dest rtxes. Handle the case in which they're different,
15395 including when the offset is zero. Replace scratchreg with an rtx.
15396 Use 2 additions if there is no spare register into which we can
15397 move a 16-bit constant.
15398 (aarch64_add_constant): Delete.
15399 (aarch64_add_offset): Replace reg with separate src and dest
15400 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
15401 Use aarch64_add_offset_1.
15402 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
15403 an rtx rather than an int. Take the delta as a poly_int64
15404 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
15405 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
15406 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
15407 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
15408 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
15409 and aarch64_add_sp.
15410 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
15411 aarch64_add_constant.
15413 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15415 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
15416 Use scalar_float_mode.
15418 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15420 * config/aarch64/aarch64-simd.md
15421 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
15422 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
15423 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
15424 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
15425 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
15426 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
15427 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
15428 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
15429 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
15430 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
15432 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15435 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
15436 targ_options->x_arm_arch_string is non NULL.
15438 2018-01-11 Tamar Christina <tamar.christina@arm.com>
15440 * config/aarch64/aarch64.h
15441 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
15443 2018-01-11 Sudakshina Das <sudi.das@arm.com>
15446 * expmed.c (emit_store_flag_force): Swap if const op0
15447 and change VOIDmode to mode of op0.
15449 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15451 PR rtl-optimization/83761
15452 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
15453 than bytes to mode_for_size.
15455 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15457 PR middle-end/83189
15458 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
15459 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
15462 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15464 PR middle-end/83575
15465 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
15466 when in layout mode.
15467 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
15468 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
15471 2018-01-10 Michael Collison <michael.collison@arm.com>
15473 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
15474 * config/aarch64/aarch64-option-extension.def: Add
15475 AARCH64_OPT_EXTENSION of 'fp16fml'.
15476 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15477 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
15478 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
15479 * config/aarch64/constraints.md (Ui7): New constraint.
15480 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
15481 (VFMLA_SEL_W): Ditto.
15484 (VFMLA16_LOW): New int iterator.
15485 (VFMLA16_HIGH): Ditto.
15486 (UNSPEC_FMLAL): New unspec.
15487 (UNSPEC_FMLSL): Ditto.
15488 (UNSPEC_FMLAL2): Ditto.
15489 (UNSPEC_FMLSL2): Ditto.
15490 (f16mac): New code attribute.
15491 * config/aarch64/aarch64-simd-builtins.def
15492 (aarch64_fmlal_lowv2sf): Ditto.
15493 (aarch64_fmlsl_lowv2sf): Ditto.
15494 (aarch64_fmlalq_lowv4sf): Ditto.
15495 (aarch64_fmlslq_lowv4sf): Ditto.
15496 (aarch64_fmlal_highv2sf): Ditto.
15497 (aarch64_fmlsl_highv2sf): Ditto.
15498 (aarch64_fmlalq_highv4sf): Ditto.
15499 (aarch64_fmlslq_highv4sf): Ditto.
15500 (aarch64_fmlal_lane_lowv2sf): Ditto.
15501 (aarch64_fmlsl_lane_lowv2sf): Ditto.
15502 (aarch64_fmlal_laneq_lowv2sf): Ditto.
15503 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
15504 (aarch64_fmlalq_lane_lowv4sf): Ditto.
15505 (aarch64_fmlsl_lane_lowv4sf): Ditto.
15506 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
15507 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
15508 (aarch64_fmlal_lane_highv2sf): Ditto.
15509 (aarch64_fmlsl_lane_highv2sf): Ditto.
15510 (aarch64_fmlal_laneq_highv2sf): Ditto.
15511 (aarch64_fmlsl_laneq_highv2sf): Ditto.
15512 (aarch64_fmlalq_lane_highv4sf): Ditto.
15513 (aarch64_fmlsl_lane_highv4sf): Ditto.
15514 (aarch64_fmlalq_laneq_highv4sf): Ditto.
15515 (aarch64_fmlsl_laneq_highv4sf): Ditto.
15516 * config/aarch64/aarch64-simd.md:
15517 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
15518 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15519 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15520 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15521 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15522 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15523 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15524 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15525 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15526 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15527 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15528 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15529 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15530 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15531 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15532 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15533 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15534 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15535 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
15536 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
15537 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
15538 (vfmlsl_low_u32): Ditto.
15539 (vfmlalq_low_u32): Ditto.
15540 (vfmlslq_low_u32): Ditto.
15541 (vfmlal_high_u32): Ditto.
15542 (vfmlsl_high_u32): Ditto.
15543 (vfmlalq_high_u32): Ditto.
15544 (vfmlslq_high_u32): Ditto.
15545 (vfmlal_lane_low_u32): Ditto.
15546 (vfmlsl_lane_low_u32): Ditto.
15547 (vfmlal_laneq_low_u32): Ditto.
15548 (vfmlsl_laneq_low_u32): Ditto.
15549 (vfmlalq_lane_low_u32): Ditto.
15550 (vfmlslq_lane_low_u32): Ditto.
15551 (vfmlalq_laneq_low_u32): Ditto.
15552 (vfmlslq_laneq_low_u32): Ditto.
15553 (vfmlal_lane_high_u32): Ditto.
15554 (vfmlsl_lane_high_u32): Ditto.
15555 (vfmlal_laneq_high_u32): Ditto.
15556 (vfmlsl_laneq_high_u32): Ditto.
15557 (vfmlalq_lane_high_u32): Ditto.
15558 (vfmlslq_lane_high_u32): Ditto.
15559 (vfmlalq_laneq_high_u32): Ditto.
15560 (vfmlslq_laneq_high_u32): Ditto.
15561 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15562 (AARCH64_FL_FOR_ARCH8_4): New.
15563 (AARCH64_ISA_F16FML): New ISA flag.
15564 (TARGET_F16FML): New feature flag for fp16fml.
15565 (doc/invoke.texi): Document new fp16fml option.
15567 2018-01-10 Michael Collison <michael.collison@arm.com>
15569 * config/aarch64/aarch64-builtins.c:
15570 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15571 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15572 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15573 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15574 (AARCH64_ISA_SHA3): New ISA flag.
15575 (TARGET_SHA3): New feature flag for sha3.
15576 * config/aarch64/iterators.md (sha512_op): New int attribute.
15577 (CRYPTO_SHA512): New int iterator.
15578 (UNSPEC_SHA512H): New unspec.
15579 (UNSPEC_SHA512H2): Ditto.
15580 (UNSPEC_SHA512SU0): Ditto.
15581 (UNSPEC_SHA512SU1): Ditto.
15582 * config/aarch64/aarch64-simd-builtins.def
15583 (aarch64_crypto_sha512hqv2di): New builtin.
15584 (aarch64_crypto_sha512h2qv2di): Ditto.
15585 (aarch64_crypto_sha512su0qv2di): Ditto.
15586 (aarch64_crypto_sha512su1qv2di): Ditto.
15587 (aarch64_eor3qv8hi): Ditto.
15588 (aarch64_rax1qv2di): Ditto.
15589 (aarch64_xarqv2di): Ditto.
15590 (aarch64_bcaxqv8hi): Ditto.
15591 * config/aarch64/aarch64-simd.md:
15592 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15593 (aarch64_crypto_sha512su0qv2di): Ditto.
15594 (aarch64_crypto_sha512su1qv2di): Ditto.
15595 (aarch64_eor3qv8hi): Ditto.
15596 (aarch64_rax1qv2di): Ditto.
15597 (aarch64_xarqv2di): Ditto.
15598 (aarch64_bcaxqv8hi): Ditto.
15599 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15600 (vsha512h2q_u64): Ditto.
15601 (vsha512su0q_u64): Ditto.
15602 (vsha512su1q_u64): Ditto.
15603 (veor3q_u16): Ditto.
15604 (vrax1q_u64): Ditto.
15605 (vxarq_u64): Ditto.
15606 (vbcaxq_u16): Ditto.
15607 * config/arm/types.md (crypto_sha512): New type attribute.
15608 (crypto_sha3): Ditto.
15609 (doc/invoke.texi): Document new sha3 option.
15611 2018-01-10 Michael Collison <michael.collison@arm.com>
15613 * config/aarch64/aarch64-builtins.c:
15614 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15615 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15616 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15617 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15618 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15619 (AARCH64_ISA_SM4): New ISA flag.
15620 (TARGET_SM4): New feature flag for sm4.
15621 * config/aarch64/aarch64-simd-builtins.def
15622 (aarch64_sm3ss1qv4si): Ditto.
15623 (aarch64_sm3tt1aq4si): Ditto.
15624 (aarch64_sm3tt1bq4si): Ditto.
15625 (aarch64_sm3tt2aq4si): Ditto.
15626 (aarch64_sm3tt2bq4si): Ditto.
15627 (aarch64_sm3partw1qv4si): Ditto.
15628 (aarch64_sm3partw2qv4si): Ditto.
15629 (aarch64_sm4eqv4si): Ditto.
15630 (aarch64_sm4ekeyqv4si): Ditto.
15631 * config/aarch64/aarch64-simd.md:
15632 (aarch64_sm3ss1qv4si): Ditto.
15633 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15634 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15635 (aarch64_sm4eqv4si): Ditto.
15636 (aarch64_sm4ekeyqv4si): Ditto.
15637 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15638 (sm3part_op): Ditto.
15639 (CRYPTO_SM3TT): Ditto.
15640 (CRYPTO_SM3PART): Ditto.
15641 (UNSPEC_SM3SS1): New unspec.
15642 (UNSPEC_SM3TT1A): Ditto.
15643 (UNSPEC_SM3TT1B): Ditto.
15644 (UNSPEC_SM3TT2A): Ditto.
15645 (UNSPEC_SM3TT2B): Ditto.
15646 (UNSPEC_SM3PARTW1): Ditto.
15647 (UNSPEC_SM3PARTW2): Ditto.
15648 (UNSPEC_SM4E): Ditto.
15649 (UNSPEC_SM4EKEY): Ditto.
15650 * config/aarch64/constraints.md (Ui2): New constraint.
15651 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15652 * config/arm/types.md (crypto_sm3): New type attribute.
15653 (crypto_sm4): Ditto.
15654 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15655 (vsm3tt1aq_u32): Ditto.
15656 (vsm3tt1bq_u32): Ditto.
15657 (vsm3tt2aq_u32): Ditto.
15658 (vsm3tt2bq_u32): Ditto.
15659 (vsm3partw1q_u32): Ditto.
15660 (vsm3partw2q_u32): Ditto.
15661 (vsm4eq_u32): Ditto.
15662 (vsm4ekeyq_u32): Ditto.
15663 (doc/invoke.texi): Document new sm4 option.
15665 2018-01-10 Michael Collison <michael.collison@arm.com>
15667 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15668 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15669 (AARCH64_FL_FOR_ARCH8_4): New.
15670 (AARCH64_FL_V8_4): New flag.
15671 (doc/invoke.texi): Document new armv8.4-a option.
15673 2018-01-10 Michael Collison <michael.collison@arm.com>
15675 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15676 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15677 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15678 * config/aarch64/aarch64-option-extension.def: Add
15679 AARCH64_OPT_EXTENSION of 'sha2'.
15680 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15681 (crypto): Disable sha2 and aes if crypto disabled.
15682 (crypto): Enable aes and sha2 if enabled.
15683 (simd): Disable sha2 and aes if simd disabled.
15684 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15686 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15687 (TARGET_SHA2): New feature flag for sha2.
15688 (TARGET_AES): New feature flag for aes.
15689 * config/aarch64/aarch64-simd.md:
15690 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15691 conditional on TARGET_AES.
15692 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15693 (aarch64_crypto_sha1hsi): Make pattern conditional
15695 (aarch64_crypto_sha1hv4si): Ditto.
15696 (aarch64_be_crypto_sha1hv4si): Ditto.
15697 (aarch64_crypto_sha1su1v4si): Ditto.
15698 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15699 (aarch64_crypto_sha1su0v4si): Ditto.
15700 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15701 (aarch64_crypto_sha256su0v4si): Ditto.
15702 (aarch64_crypto_sha256su1v4si): Ditto.
15703 (doc/invoke.texi): Document new aes and sha2 options.
15705 2018-01-10 Martin Sebor <msebor@redhat.com>
15707 PR tree-optimization/83781
15708 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15711 2018-01-11 Martin Sebor <msebor@gmail.com>
15712 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15714 PR tree-optimization/83501
15715 PR tree-optimization/81703
15717 * tree-ssa-strlen.c (get_string_cst): Rename...
15718 (get_string_len): ...to this. Handle global constants.
15719 (handle_char_store): Adjust.
15721 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15722 Jim Wilson <jimw@sifive.com>
15724 * config/riscv/riscv-protos.h (riscv_output_return): New.
15725 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15726 (riscv_attribute_table, riscv_output_return),
15727 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15728 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15729 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15731 (riscv_expand_prologue): Add early return for naked function.
15732 (riscv_expand_epilogue): Likewise.
15733 (riscv_function_ok_for_sibcall): Return false for naked function.
15734 (riscv_set_current_function): New.
15735 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15736 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15737 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15738 * doc/extend.texi (RISC-V Function Attributes): New.
15740 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15742 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15743 check for 128-bit long double before checking TCmode.
15744 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15745 128-bit long doubles before checking TFmode or TCmode.
15746 (FLOAT128_IBM_P): Likewise.
15748 2018-01-10 Martin Sebor <msebor@redhat.com>
15750 PR tree-optimization/83671
15751 * builtins.c (c_strlen): Unconditionally return zero for the empty
15753 Use -Warray-bounds for warnings.
15754 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15755 for non-constant array indices with COMPONENT_REF, arrays of
15756 arrays, and pointers to arrays.
15757 (gimple_fold_builtin_strlen): Determine and set length range for
15758 non-constant character arrays.
15760 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15762 PR middle-end/81897
15763 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15766 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15768 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15770 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15773 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15774 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15775 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15776 indexed_or_indirect_operand predicate.
15777 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15778 (*vsx_le_perm_load_v8hi): Likewise.
15779 (*vsx_le_perm_load_v16qi): Likewise.
15780 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15781 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15782 (*vsx_le_perm_store_v8hi): Likewise.
15783 (*vsx_le_perm_store_v16qi): Likewise.
15784 (eight unnamed splitters): Likewise.
15786 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15788 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15789 * config/rs6000/emmintrin.h: Likewise.
15790 * config/rs6000/mmintrin.h: Likewise.
15791 * config/rs6000/xmmintrin.h: Likewise.
15793 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15796 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15798 * tree.c (tree_nop_conversion): Return true for location wrapper
15800 (maybe_wrap_with_location): New function.
15801 (selftest::check_strip_nops): New function.
15802 (selftest::test_location_wrappers): New function.
15803 (selftest::tree_c_tests): Call it.
15804 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15805 (maybe_wrap_with_location): New decl.
15806 (EXPR_LOCATION_WRAPPER_P): New macro.
15807 (location_wrapper_p): New inline function.
15808 (tree_strip_any_location_wrapper): New inline function.
15810 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15813 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15814 stack_realign_offset for the largest alignment of stack slot
15816 (ix86_find_max_used_stack_alignment): New function.
15817 (ix86_finalize_stack_frame_flags): Use it. Set
15818 max_used_stack_alignment if we don't realign stack.
15819 * config/i386/i386.h (machine_function): Add
15820 max_used_stack_alignment.
15822 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15824 * config/arm/arm.opt (-mbranch-cost): New option.
15825 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15828 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15831 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15832 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15834 2018-01-10 Richard Biener <rguenther@suse.de>
15837 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15838 early out so it also covers the case where we have a non-NULL
15841 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15843 PR tree-optimization/83753
15844 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15845 for non-strided grouped accesses if the number of elements is 1.
15847 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15850 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15851 * i386.h (TARGET_USE_GATHER): Define.
15852 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15854 2018-01-10 Martin Liska <mliska@suse.cz>
15857 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15858 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15860 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15861 CLEANUP_NO_PARTITIONING is not set.
15863 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15865 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15866 for vectors, as a partial revert of r254296.
15867 * rtl.h (const_vec_p): Delete.
15868 (const_vec_duplicate_p): Don't test for vector CONSTs.
15869 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15870 * expmed.c (make_tree): Likewise.
15873 * common.md (E, F): Use CONSTANT_P instead of checking for
15875 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15876 checking for CONST_VECTOR.
15878 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15880 PR middle-end/83575
15881 * predict.c (force_edge_cold): Handle in more sane way edges
15882 with no prediction.
15884 2018-01-09 Carl Love <cel@us.ibm.com>
15886 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15888 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15889 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15890 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15891 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15892 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15893 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15894 * config/rs6000/rs6000-protos.h: Add extern defition for
15895 rs6000_generate_float2_double_code.
15896 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15898 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15899 (float2_v2df): Add define_expand.
15901 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15904 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15905 op_mode in the force_to_mode call.
15907 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15909 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15910 instead of checking each element individually.
15911 (aarch64_evpc_uzp): Likewise.
15912 (aarch64_evpc_zip): Likewise.
15913 (aarch64_evpc_ext): Likewise.
15914 (aarch64_evpc_rev): Likewise.
15915 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15916 instead of checking each element individually. Return true without
15918 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15919 whether all selected elements come from the same input, instead of
15920 checking each element individually. Remove calls to gen_rtx_REG,
15921 start_sequence and end_sequence and instead assert that no rtl is
15924 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15926 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15927 order of HIGH and CONST checks.
15929 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15931 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15932 if the destination isn't an SSA_NAME.
15934 2018-01-09 Richard Biener <rguenther@suse.de>
15936 PR tree-optimization/83668
15937 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15939 (canonicalize_loop_form): ... here, renamed from ...
15940 (canonicalize_loop_closed_ssa_form): ... this and amended to
15941 swap successor edges for loop exit blocks to make us use
15942 the RPO order we need for initial schedule generation.
15944 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15946 PR tree-optimization/64811
15947 * match.pd: When optimizing comparisons with Inf, avoid
15948 introducing or losing exceptions from comparisons with NaN.
15950 2018-01-09 Martin Liska <mliska@suse.cz>
15953 * asan.c (shadow_mem_size): Add gcc_assert.
15955 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15957 Don't save registers in main().
15960 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15961 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15962 * config/avr/avr.c (avr_set_current_function): Don't error if
15963 naked, OS_task or OS_main are specified at the same time.
15964 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15966 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15968 * common/config/avr/avr-common.c (avr_option_optimization_table):
15969 Switch on -mmain-is-OS_task for optimizing compilations.
15971 2018-01-09 Richard Biener <rguenther@suse.de>
15973 PR tree-optimization/83572
15974 * graphite.c: Include cfganal.h.
15975 (graphite_transform_loops): Connect infinite loops to exit
15976 and remove fake edges at the end.
15978 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15980 * ipa-inline.c (edge_badness): Revert accidental checkin.
15982 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15985 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15986 symbols; not inline clones.
15988 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15991 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15992 hard registers. Formatting fixes.
15994 PR preprocessor/83722
15995 * gcc.c (try_generate_repro): Pass
15996 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15997 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
16000 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
16001 Kito Cheng <kito.cheng@gmail.com>
16003 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
16004 (riscv_leaf_function_p): Delete.
16005 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
16007 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16009 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
16011 (do_ifelse): New function.
16012 (do_isel): New function.
16013 (do_sub3): New function.
16014 (do_add3): New function.
16015 (do_load_mask_compare): New function.
16016 (do_overlap_load_compare): New function.
16017 (expand_compare_loop): New function.
16018 (expand_block_compare): Call expand_compare_loop() when appropriate.
16019 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
16020 option description.
16021 (-mblock-compare-inline-loop-limit): New option.
16023 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16026 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
16027 Reverse order of second and third operands in first alternative.
16028 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
16029 of first and second elements in UNSPEC_VPERMR vector.
16030 (altivec_expand_vec_perm_le): Likewise.
16032 2018-01-08 Jeff Law <law@redhat.com>
16034 PR rtl-optimizatin/81308
16035 * tree-switch-conversion.c (cfg_altered): New file scoped static.
16036 (process_switch): If group_case_labels makes a change, then set
16038 (pass_convert_switch::execute): If a switch is converted, then
16039 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
16041 PR rtl-optimization/81308
16042 * recog.c (split_all_insns): Conditionally cleanup the CFG after
16045 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
16047 PR target/83663 - Revert r255946
16048 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
16049 generation for cases where splatting a value is not useful.
16050 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
16051 across a vec_duplicate and a paradoxical subreg forming a vector
16052 mode to a vec_concat.
16054 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16056 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
16057 -march=armv8.3-a variants.
16058 * config/arm/t-multilib: Likewise.
16059 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
16061 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16063 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
16065 (cceq_ior_compare_complement): Give it a name so I can use it, and
16066 change boolean_or_operator predicate to boolean_operator so it can
16067 be used to generate a crand.
16068 (eqne): New code iterator.
16069 (bd/bd_neg): New code_attrs.
16070 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
16071 a single define_insn.
16072 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
16073 decrement (bdnzt/bdnzf/bdzt/bdzf).
16074 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
16075 with the new names of the branch decrement patterns, and added the
16076 names of the branch decrement conditional patterns.
16078 2018-01-08 Richard Biener <rguenther@suse.de>
16080 PR tree-optimization/83563
16081 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
16084 2018-01-08 Richard Biener <rguenther@suse.de>
16086 PR middle-end/83713
16087 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
16089 2018-01-08 Richard Biener <rguenther@suse.de>
16091 PR tree-optimization/83685
16092 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
16093 references to abnormals.
16095 2018-01-08 Richard Biener <rguenther@suse.de>
16098 * dwarf2out.c (output_indirect_strings): Handle empty
16099 skeleton_debug_str_hash.
16100 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
16102 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16104 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
16105 (emit_store_direct): Likewise.
16106 (arc_trampoline_adjust_address): Likewise.
16107 (arc_asm_trampoline_template): New function.
16108 (arc_initialize_trampoline): Use asm_trampoline_template.
16109 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
16110 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
16111 * config/arc/arc.md (flush_icache): Delete pattern.
16113 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16115 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
16116 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
16119 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16122 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
16123 by not USED_FOR_TARGET.
16124 (make_pass_resolve_sw_modes): Likewise.
16126 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16128 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
16131 2018-01-08 Richard Biener <rguenther@suse.de>
16133 PR middle-end/83580
16134 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
16136 2018-01-08 Richard Biener <rguenther@suse.de>
16138 PR middle-end/83517
16139 * match.pd ((t * 2) / 2) -> t): Add missing :c.
16141 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
16143 PR middle-end/81897
16144 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
16145 basic blocks with a small number of successors.
16146 (convert_control_dep_chain_into_preds): Improve handling of
16148 (dump_predicates): Split apart into...
16149 (dump_pred_chain): ...here...
16150 (dump_pred_info): ...and here.
16151 (can_one_predicate_be_invalidated_p): Add debugging printfs.
16152 (can_chain_union_be_invalidated_p): Improve check for invalidation
16154 (uninit_uses_cannot_happen): Avoid unnecessary if
16155 convert_control_dep_chain_into_preds yielded nothing.
16157 2018-01-06 Martin Sebor <msebor@redhat.com>
16159 PR tree-optimization/83640
16160 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
16161 subtracting negative offset from size.
16162 (builtin_access::overlap): Adjust offset bounds of the access to fall
16163 within the size of the object if possible.
16165 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
16167 PR rtl-optimization/83699
16168 * expmed.c (extract_bit_field_1): Restrict the vector usage of
16169 extract_bit_field_as_subreg to cases in which the extracted
16170 value is also a vector.
16172 * lra-constraints.c (process_alt_operands): Test for the equivalence
16173 substitutions when detecting a possible reload cycle.
16175 2018-01-06 Jakub Jelinek <jakub@redhat.com>
16178 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
16179 by default if flag_selective_schedling{,2}. Formatting fixes.
16181 PR rtl-optimization/83682
16182 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
16183 if it has non-VECTOR_MODE element mode.
16184 (vec_duplicate_p): Likewise.
16186 PR middle-end/83694
16187 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
16188 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
16190 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16193 * config/i386/i386-builtin.def
16194 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
16195 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
16196 Require also OPTION_MASK_ISA_AVX512F in addition to
16197 OPTION_MASK_ISA_GFNI.
16198 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
16199 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
16200 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
16201 to OPTION_MASK_ISA_GFNI.
16202 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
16203 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
16204 OPTION_MASK_ISA_AVX512BW.
16205 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
16206 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
16207 addition to OPTION_MASK_ISA_GFNI.
16208 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
16209 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
16210 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
16211 to OPTION_MASK_ISA_GFNI.
16212 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
16213 a requirement for all ISAs rather than any of them with a few
16215 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
16217 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
16218 bitmasks to be enabled with 3 exceptions, instead of requiring any
16219 enabled ISA with lots of exceptions.
16220 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
16221 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
16222 Change avx512bw in isa attribute to avx512f.
16223 * config/i386/sgxintrin.h: Add license boilerplate.
16224 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
16225 to __AVX512F__ and __AVX512VL to __AVX512VL__.
16226 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
16227 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
16229 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
16230 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
16231 temporarily sse2 rather than sse if not enabled already.
16234 * config/i386/sse.md (VI248_VLBW): Rename to ...
16235 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
16236 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
16237 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
16238 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
16239 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
16240 mode iterator instead of VI248_VLBW.
16242 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
16244 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
16245 (record_modified): Skip clobbers; add debug output.
16246 (param_change_prob): Use sreal frequencies.
16248 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16250 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
16251 punt for user-aligned variables.
16253 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16255 * tree-chrec.c (chrec_contains_symbols): Return true for
16258 2018-01-05 Sudakshina Das <sudi.das@arm.com>
16261 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
16262 of (x|y) == x for BICS pattern.
16264 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16266 PR tree-optimization/83605
16267 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
16268 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
16271 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
16273 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
16274 * config/epiphany/rtems.h: New file.
16276 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16277 Uros Bizjak <ubizjak@gmail.com>
16280 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
16281 QIreg_operand instead of register_operand predicate.
16282 * config/i386/i386.c (ix86_rop_should_change_byte_p,
16283 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
16284 comments instead of -fmitigate[-_]rop.
16286 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16289 * cgraphunit.c (symbol_table::compile): Switch to text_section
16290 before calling assembly_start debug hook.
16291 * run-rtl-passes.c (run_rtl_passes): Likewise.
16294 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16296 * tree-vrp.c (extract_range_from_binary_expr_1): Check
16297 range_int_cst_p rather than !symbolic_range_p before calling
16298 extract_range_from_multiplicative_op_1.
16300 2018-01-04 Jeff Law <law@redhat.com>
16302 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
16303 redundant test in assertion.
16305 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16307 * doc/rtl.texi: Document machine_mode wrapper classes.
16309 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16311 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
16312 using tree_to_uhwi.
16314 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16316 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
16317 the VEC_PERM_EXPR fold to fail.
16319 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16322 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
16323 to switched_sections.
16325 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16328 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
16329 test for d.testing.
16331 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
16334 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
16335 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
16337 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16340 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
16341 is BLKmode and bitpos not zero or mode change is needed.
16343 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16346 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
16349 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
16352 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
16353 instead of MULT rtx. Update all corresponding splitters.
16354 (*saddl_se): Ditto.
16355 (*ssub<modesuffix>): Ditto.
16356 (*ssubl_se): Ditto.
16357 (*cmp_sadd_di): Update split patterns.
16358 (*cmp_sadd_si): Ditto.
16359 (*cmp_sadd_sidi): Ditto.
16360 (*cmp_ssub_di): Ditto.
16361 (*cmp_ssub_si): Ditto.
16362 (*cmp_ssub_sidi): Ditto.
16363 * config/alpha/predicates.md (const23_operand): New predicate.
16364 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
16365 Look for ASHIFT, not MULT inner operand.
16366 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
16368 2018-01-04 Martin Liska <mliska@suse.cz>
16370 PR gcov-profile/83669
16371 * gcov.c (output_intermediate_file): Add version to intermediate
16373 * doc/gcov.texi: Document new field 'version' in intermediate
16374 file format. Fix location of '-k' option of gcov command.
16376 2018-01-04 Martin Liska <mliska@suse.cz>
16379 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
16381 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16383 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
16385 2018-01-03 Martin Sebor <msebor@redhat.com>
16387 PR tree-optimization/83655
16388 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
16389 checking calls with invalid arguments.
16391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16393 * tree-vect-stmts.c (vect_get_store_rhs): New function.
16394 (vectorizable_mask_load_store): Delete.
16395 (vectorizable_call): Return false for masked loads and stores.
16396 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
16397 instead of gimple_assign_rhs1.
16398 (vectorizable_load): Handle IFN_MASK_LOAD.
16399 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
16401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16403 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
16405 (vectorizable_mask_load_store): ...here.
16406 (vectorizable_load): ...and here.
16408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16410 * tree-vect-stmts.c (vect_build_all_ones_mask)
16411 (vect_build_zero_merge_argument): New functions, split out from...
16412 (vectorizable_load): ...here.
16414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16416 * tree-vect-stmts.c (vect_check_store_rhs): New function,
16418 (vectorizable_mask_load_store): ...here.
16419 (vectorizable_store): ...and here.
16421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16423 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
16425 (vectorizable_mask_load_store): ...here.
16427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16429 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
16430 (vect_model_store_cost): Take a vec_load_store_type instead of a
16432 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
16433 (vect_model_store_cost): Take a vec_load_store_type instead of a
16435 (vectorizable_mask_load_store): Update accordingly.
16436 (vectorizable_store): Likewise.
16437 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
16439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16441 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
16442 IFN_MASK_LOAD calls here rather than...
16443 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
16445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16446 Alan Hayward <alan.hayward@arm.com>
16447 David Sherwood <david.sherwood@arm.com>
16449 * expmed.c (extract_bit_field_1): For vector extracts,
16450 fall back to extract_bit_field_as_subreg if vec_extract
16453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16454 Alan Hayward <alan.hayward@arm.com>
16455 David Sherwood <david.sherwood@arm.com>
16457 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
16458 they are variable or constant sized.
16459 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
16460 slots for constant-sized data.
16462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16463 Alan Hayward <alan.hayward@arm.com>
16464 David Sherwood <david.sherwood@arm.com>
16466 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
16467 handling COND_EXPRs with boolean comparisons, try to find a better
16468 basis for the mask type than the boolean itself.
16470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16472 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
16473 is calculated and how it can be overridden.
16474 * genmodes.c (max_bitsize_mode_any_mode): New variable.
16475 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
16477 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
16480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16481 Alan Hayward <alan.hayward@arm.com>
16482 David Sherwood <david.sherwood@arm.com>
16484 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
16485 Remove the mode argument.
16486 (aarch64_simd_valid_immediate): Remove the mode and inverse
16488 * config/aarch64/iterators.md (bitsize): New iterator.
16489 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
16490 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
16491 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
16492 aarch64_simd_valid_immediate.
16493 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
16494 (aarch64_reg_or_bic_imm): Likewise.
16495 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
16496 with an insn_type enum and msl with a modifier_type enum.
16497 Replace element_width with a scalar_mode. Change the shift
16498 to unsigned int. Add constructors for scalar_float_mode and
16499 scalar_int_mode elements.
16500 (aarch64_vect_float_const_representable_p): Delete.
16501 (aarch64_can_const_movi_rtx_p)
16502 (aarch64_simd_scalar_immediate_valid_for_move)
16503 (aarch64_simd_make_constant): Update call to
16504 aarch64_simd_valid_immediate.
16505 (aarch64_advsimd_valid_immediate_hs): New function.
16506 (aarch64_advsimd_valid_immediate): Likewise.
16507 (aarch64_simd_valid_immediate): Remove mode and inverse
16508 arguments. Rewrite to use the above. Use const_vec_duplicate_p
16509 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
16510 and aarch64_float_const_representable_p on the result.
16511 (aarch64_output_simd_mov_immediate): Remove mode argument.
16512 Update call to aarch64_simd_valid_immediate and use of
16513 simd_immediate_info.
16514 (aarch64_output_scalar_simd_mov_immediate): Update call
16517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16518 Alan Hayward <alan.hayward@arm.com>
16519 David Sherwood <david.sherwood@arm.com>
16521 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16522 (mode_nunits): Likewise CONST_MODE_NUNITS.
16523 * machmode.def (ADJUST_NUNITS): Document.
16524 * genmodes.c (mode_data::need_nunits_adj): New field.
16525 (blank_mode): Update accordingly.
16526 (adj_nunits): New variable.
16527 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16529 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16530 listed in adj_nunits.
16531 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16532 listed in adj_nunits. Don't emit case statements for such modes.
16533 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16534 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16535 nothing if adj_nunits is nonnull.
16536 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
16537 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
16538 (emit_mode_fbit): Update use of print_maybe_const_decl.
16539 (emit_move_size): Likewise. Treat the array as non-const
16541 (emit_mode_adjustments): Handle adj_nunits.
16543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16545 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
16546 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
16547 (VECTOR_MODES): Use it.
16548 (make_vector_modes): Take the prefix as an argument.
16550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16551 Alan Hayward <alan.hayward@arm.com>
16552 David Sherwood <david.sherwood@arm.com>
16554 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16555 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16556 for MODE_VECTOR_BOOL.
16557 * machmode.def (VECTOR_BOOL_MODE): Document.
16558 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16559 (make_vector_bool_mode): New function.
16560 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16562 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16563 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16565 * stor-layout.c (int_mode_for_mode): Likewise.
16566 * tree.c (build_vector_type_for_mode): Likewise.
16567 * varasm.c (output_constant_pool_2): Likewise.
16568 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16569 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16570 for MODE_VECTOR_BOOL.
16571 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16572 of mode class checks.
16573 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16574 instead of a list of mode class checks.
16575 (expand_vector_scalar_condition): Likewise.
16576 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16579 Alan Hayward <alan.hayward@arm.com>
16580 David Sherwood <david.sherwood@arm.com>
16582 * machmode.h (mode_size): Change from unsigned short to
16584 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16585 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16586 or if measurement_type is not polynomial.
16587 (fixed_size_mode::includes_p): Check for constant-sized modes.
16588 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16589 return a poly_uint16 rather than an unsigned short.
16590 (emit_mode_size): Change the type of mode_size from unsigned short
16591 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16592 (emit_mode_adjustments): Cope with polynomial vector sizes.
16593 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16595 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16597 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16598 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16599 * caller-save.c (setup_save_areas): Likewise.
16600 (replace_reg_with_saved_mem): Likewise.
16601 * calls.c (emit_library_call_value_1): Likewise.
16602 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16603 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16604 (gen_lowpart_for_combine): Likewise.
16605 * convert.c (convert_to_integer_1): Likewise.
16606 * cse.c (equiv_constant, cse_insn): Likewise.
16607 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16608 (cselib_subst_to_values): Likewise.
16609 * dce.c (word_dce_process_block): Likewise.
16610 * df-problems.c (df_word_lr_mark_ref): Likewise.
16611 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16612 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16613 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16614 (rtl_for_decl_location): Likewise.
16615 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16616 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16617 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16618 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16619 (expand_expr_real_1): Likewise.
16620 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16621 (pad_below): Likewise.
16622 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16623 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16624 * ira.c (get_subreg_tracking_sizes): Likewise.
16625 * ira-build.c (ira_create_allocno_objects): Likewise.
16626 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16627 (ira_sort_regnos_for_alter_reg): Likewise.
16628 * ira-costs.c (record_operand_costs): Likewise.
16629 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16630 (resolve_simple_move): Likewise.
16631 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16632 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16633 (lra_constraints): Likewise.
16634 (CONST_POOL_OK_P): Reject variable-sized modes.
16635 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16636 (add_pseudo_to_slot, lra_spill): Likewise.
16637 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16638 * optabs-query.c (get_best_extraction_insn): Likewise.
16639 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16640 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16641 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16642 * recog.c (offsettable_address_addr_space_p): Likewise.
16643 * regcprop.c (maybe_mode_change): Likewise.
16644 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16645 * regrename.c (build_def_use): Likewise.
16646 * regstat.c (dump_reg_info): Likewise.
16647 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16648 (find_reloads, find_reloads_subreg_address): Likewise.
16649 * reload1.c (eliminate_regs_1): Likewise.
16650 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16651 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16652 (simplify_binary_operation_1, simplify_subreg): Likewise.
16653 * targhooks.c (default_function_arg_padding): Likewise.
16654 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16655 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16656 (verify_gimple_assign_ternary): Likewise.
16657 * tree-inline.c (estimate_move_cost): Likewise.
16658 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16659 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16660 (get_address_cost_ainc): Likewise.
16661 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16662 (vect_supportable_dr_alignment): Likewise.
16663 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16664 (vectorizable_reduction): Likewise.
16665 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16666 (vectorizable_operation, vectorizable_load): Likewise.
16667 * tree.c (build_same_sized_truth_vector_type): Likewise.
16668 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16669 * var-tracking.c (emit_note_insn_var_location): Likewise.
16670 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16671 (ADDR_VEC_ALIGN): Likewise.
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 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16679 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16680 or if measurement_type is polynomial.
16681 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16682 * combine.c (make_extraction): Likewise.
16683 * dse.c (find_shift_sequence): Likewise.
16684 * dwarf2out.c (mem_loc_descriptor): Likewise.
16685 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16686 (extract_bit_field, extract_low_bits): Likewise.
16687 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16688 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16689 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16690 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16691 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16692 * reload.c (find_reloads): Likewise.
16693 * reload1.c (alter_reg): Likewise.
16694 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16695 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16696 * tree-if-conv.c (predicate_mem_writes): Likewise.
16697 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16698 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16699 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16700 * valtrack.c (dead_debug_insert_temp): Likewise.
16701 * varasm.c (mergeable_constant_section): Likewise.
16702 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16705 Alan Hayward <alan.hayward@arm.com>
16706 David Sherwood <david.sherwood@arm.com>
16708 * expr.c (expand_assignment): Cope with polynomial mode sizes
16709 when assigning to a CONCAT.
16711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16712 Alan Hayward <alan.hayward@arm.com>
16713 David Sherwood <david.sherwood@arm.com>
16715 * machmode.h (mode_precision): Change from unsigned short to
16717 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16719 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16720 or if measurement_type is not polynomial.
16721 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16722 in which the mode is already known to be a scalar_int_mode.
16723 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16724 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16726 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16727 for GET_MODE_PRECISION.
16728 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16729 for GET_MODE_PRECISION.
16730 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16732 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16733 (expand_field_assignment, make_extraction): Likewise.
16734 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16735 (get_last_value): Likewise.
16736 * convert.c (convert_to_integer_1): Likewise.
16737 * cse.c (cse_insn): Likewise.
16738 * expr.c (expand_expr_real_1): Likewise.
16739 * lra-constraints.c (simplify_operand_subreg): Likewise.
16740 * optabs-query.c (can_atomic_load_p): Likewise.
16741 * optabs.c (expand_atomic_load): Likewise.
16742 (expand_atomic_store): Likewise.
16743 * ree.c (combine_reaching_defs): Likewise.
16744 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16745 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16746 * tree.h (type_has_mode_precision_p): Likewise.
16747 * ubsan.c (instrument_si_overflow): Likewise.
16749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16750 Alan Hayward <alan.hayward@arm.com>
16751 David Sherwood <david.sherwood@arm.com>
16753 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16754 polynomial numbers of units.
16755 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16756 (valid_vector_subparts_p): New function.
16757 (build_vector_type): Remove temporary shim and take the number
16758 of units as a poly_uint64 rather than an int.
16759 (build_opaque_vector_type): Take the number of units as a
16760 poly_uint64 rather than an int.
16761 * tree.c (build_vector_from_ctor): Handle polynomial
16762 TYPE_VECTOR_SUBPARTS.
16763 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16764 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16765 (build_vector_from_val): If the number of units is variable,
16766 use build_vec_duplicate_cst for constant operands and
16767 VEC_DUPLICATE_EXPR otherwise.
16768 (make_vector_type): Remove temporary is_constant ().
16769 (build_vector_type, build_opaque_vector_type): Take the number of
16770 units as a poly_uint64 rather than an int.
16771 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16773 * cfgexpand.c (expand_debug_expr): Likewise.
16774 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16775 (store_constructor, expand_expr_real_1): Likewise.
16776 (const_scalar_mask_from_tree): Likewise.
16777 * fold-const-call.c (fold_const_reduction): Likewise.
16778 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16779 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16780 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16781 (fold_relational_const): Likewise.
16782 (native_interpret_vector): Likewise. Change the size from an
16783 int to an unsigned int.
16784 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16785 TYPE_VECTOR_SUBPARTS.
16786 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16787 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16788 duplicating a non-constant operand into a variable-length vector.
16789 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16790 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16791 * ipa-icf.c (sem_variable::equals): Likewise.
16792 * match.pd: Likewise.
16793 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16794 * print-tree.c (print_node): Likewise.
16795 * stor-layout.c (layout_type): Likewise.
16796 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16797 * tree-cfg.c (verify_gimple_comparison): Likewise.
16798 (verify_gimple_assign_binary): Likewise.
16799 (verify_gimple_assign_ternary): Likewise.
16800 (verify_gimple_assign_single): Likewise.
16801 * tree-pretty-print.c (dump_generic_node): Likewise.
16802 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16803 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16804 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16805 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16806 (vect_shift_permute_load_chain): Likewise.
16807 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16808 (expand_vector_condition, optimize_vector_constructor): Likewise.
16809 (lower_vec_perm, get_compute_type): Likewise.
16810 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16811 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16812 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16813 (vect_recog_mask_conversion_pattern): Likewise.
16814 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16815 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16816 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16817 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16818 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16819 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16820 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16821 (supportable_widening_operation): Likewise.
16822 (supportable_narrowing_operation): Likewise.
16823 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16825 * varasm.c (output_constant): Likewise.
16827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16828 Alan Hayward <alan.hayward@arm.com>
16829 David Sherwood <david.sherwood@arm.com>
16831 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16832 so that both the length == 3 and length != 3 cases set up their
16833 own permute vectors. Add comments explaining why we know the
16834 number of elements is constant.
16835 (vect_permute_load_chain): Likewise.
16837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16838 Alan Hayward <alan.hayward@arm.com>
16839 David Sherwood <david.sherwood@arm.com>
16841 * machmode.h (mode_nunits): Change from unsigned char to
16843 (ONLY_FIXED_SIZE_MODES): New macro.
16844 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16845 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16846 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16848 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16849 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16850 or if measurement_type is not polynomial.
16851 * genmodes.c (ZERO_COEFFS): New macro.
16852 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16854 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16855 Use ZERO_COEFFS when emitting initializers.
16856 * data-streamer.h (bp_pack_poly_value): New function.
16857 (bp_unpack_poly_value): Likewise.
16858 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16859 for GET_MODE_NUNITS.
16860 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16861 for GET_MODE_NUNITS.
16862 * tree.c (make_vector_type): Remove temporary shim and make
16863 the real function take the number of units as a poly_uint64
16864 rather than an int.
16865 (build_vector_type_for_mode): Handle polynomial nunits.
16866 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16867 * emit-rtl.c (const_vec_series_p_1): Likewise.
16868 (gen_rtx_CONST_VECTOR): Likewise.
16869 * fold-const.c (test_vec_duplicate_folding): Likewise.
16870 * genrecog.c (validate_pattern): Likewise.
16871 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16872 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16873 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16874 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16875 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16876 * rtlanal.c (subreg_get_info): Likewise.
16877 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16878 (vect_grouped_load_supported): Likewise.
16879 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16880 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16881 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16882 (simplify_const_unary_operation, simplify_binary_operation_1)
16883 (simplify_const_binary_operation, simplify_ternary_operation)
16884 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16885 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16886 instead of CONST_VECTOR_NUNITS.
16887 * varasm.c (output_constant_pool_2): Likewise.
16888 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16889 explicit-encoded elements in the XVEC for variable-length vectors.
16891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16893 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16896 Alan Hayward <alan.hayward@arm.com>
16897 David Sherwood <david.sherwood@arm.com>
16899 * coretypes.h (fixed_size_mode): Declare.
16900 (fixed_size_mode_pod): New typedef.
16901 * builtins.h (target_builtins::x_apply_args_mode)
16902 (target_builtins::x_apply_result_mode): Change type to
16903 fixed_size_mode_pod.
16904 * builtins.c (apply_args_size, apply_result_size, result_vector)
16905 (expand_builtin_apply_args_1, expand_builtin_apply)
16906 (expand_builtin_return): Update accordingly.
16908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16910 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16911 * cselib.c (cselib_hash_rtx): Likewise.
16912 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16913 CONST_VECTOR encoding.
16915 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16916 Jeff Law <law@redhat.com>
16919 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16920 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16921 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16922 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16925 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16926 explicitly probe *sp in a noreturn function if there were any callee
16927 register saves or frame pointer is needed.
16929 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16932 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16933 BLKmode for ternary, binary or unary expressions.
16936 * var-tracking.c (delete_vta_debug_insn): New inline function.
16937 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16938 insns from get_insns () to NULL instead of each bb separately.
16939 Use delete_vta_debug_insn. No longer static.
16940 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16941 delete_vta_debug_insns callers.
16942 * rtl.h (delete_vta_debug_insns): Declare.
16943 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16944 instead of variable_tracking_main.
16946 2018-01-03 Martin Sebor <msebor@redhat.com>
16948 PR tree-optimization/83603
16949 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16950 arguments past the endof the argument list in functions declared
16951 without a prototype.
16952 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16953 Avoid checking when arguments are null.
16955 2018-01-03 Martin Sebor <msebor@redhat.com>
16958 * doc/extend.texi (attribute const): Fix a typo.
16959 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16960 issuing -Wsuggest-attribute for void functions.
16962 2018-01-03 Martin Sebor <msebor@redhat.com>
16964 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16965 offset_int::from instead of wide_int::to_shwi.
16966 (maybe_diag_overlap): Remove assertion.
16967 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16968 * gimple-ssa-sprintf.c (format_directive): Same.
16969 (parse_directive): Same.
16970 (sprintf_dom_walker::compute_format_length): Same.
16971 (try_substitute_return_value): Same.
16973 2018-01-03 Jeff Law <law@redhat.com>
16975 PR middle-end/83654
16976 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16977 non-constant residual for zero at runtime and avoid probing in
16978 that case. Reorganize code for trailing problem to mirror handling
16981 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16983 PR tree-optimization/83501
16984 * tree-ssa-strlen.c (get_string_cst): New.
16985 (handle_char_store): Call get_string_cst.
16987 2018-01-03 Martin Liska <mliska@suse.cz>
16989 PR tree-optimization/83593
16990 * tree-ssa-strlen.c: Include tree-cfg.h.
16991 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16992 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16993 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16995 (strlen_dom_walker::before_dom_children): Call
16996 gimple_purge_dead_eh_edges. Dump tranformation with details
16998 (strlen_dom_walker::before_dom_children): Update call by adding
16999 new argument cleanup_eh.
17000 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
17002 2018-01-03 Martin Liska <mliska@suse.cz>
17005 * cif-code.def (VARIADIC_THUNK): New enum value.
17006 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
17009 2018-01-03 Jan Beulich <jbeulich@suse.com>
17011 * sse.md (mov<mode>_internal): Tighten condition for when to use
17012 vmovdqu<ssescalarsize> for TI and OI modes.
17014 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17016 Update copyright years.
17018 2018-01-03 Martin Liska <mliska@suse.cz>
17021 * ipa-visibility.c (function_and_variable_visibility): Skip
17022 functions with noipa attribure.
17024 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17026 * gcc.c (process_command): Update copyright notice dates.
17027 * gcov-dump.c (print_version): Ditto.
17028 * gcov.c (print_version): Ditto.
17029 * gcov-tool.c (print_version): Ditto.
17030 * gengtype.c (create_file): Ditto.
17031 * doc/cpp.texi: Bump @copying's copyright year.
17032 * doc/cppinternals.texi: Ditto.
17033 * doc/gcc.texi: Ditto.
17034 * doc/gccint.texi: Ditto.
17035 * doc/gcov.texi: Ditto.
17036 * doc/install.texi: Ditto.
17037 * doc/invoke.texi: Ditto.
17039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17041 * vector-builder.h (vector_builder::m_full_nelts): Change from
17042 unsigned int to poly_uint64.
17043 (vector_builder::full_nelts): Update prototype accordingly.
17044 (vector_builder::new_vector): Likewise.
17045 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
17046 (vector_builder::operator ==): Likewise.
17047 (vector_builder::finalize): Likewise.
17048 * int-vector-builder.h (int_vector_builder::int_vector_builder):
17049 Take the number of elements as a poly_uint64 rather than an
17051 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
17052 from unsigned int to poly_uint64.
17053 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
17054 (vec_perm_indices::new_vector): Likewise.
17055 (vec_perm_indices::length): Likewise.
17056 (vec_perm_indices::nelts_per_input): Likewise.
17057 (vec_perm_indices::input_nelts): Likewise.
17058 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
17059 number of elements per input as a poly_uint64 rather than an
17060 unsigned int. Use the original encoding for variable-length
17061 vectors, rather than clamping each individual element.
17062 For the second and subsequent elements in each pattern,
17063 clamp the step and base before clamping their sum.
17064 (vec_perm_indices::series_p): Handle polynomial element counts.
17065 (vec_perm_indices::all_in_range_p): Likewise.
17066 (vec_perm_indices_to_tree): Likewise.
17067 (vec_perm_indices_to_rtx): Likewise.
17068 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
17069 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
17070 (tree_vector_builder::new_binary_operation): Handle polynomial
17071 element counts. Return false if we need to know the number
17072 of elements at compile time.
17073 * fold-const.c (fold_vec_perm): Punt if the number of elements
17074 isn't known at compile time.
17076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17078 * vec-perm-indices.h (vec_perm_builder): Change element type
17079 from HOST_WIDE_INT to poly_int64.
17080 (vec_perm_indices::element_type): Update accordingly.
17081 (vec_perm_indices::clamp): Handle polynomial element_types.
17082 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17083 (vec_perm_indices::all_in_range_p): Likewise.
17084 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
17086 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
17087 polynomial vec_perm_indices element types.
17088 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
17089 * fold-const.c (fold_vec_perm): Likewise.
17090 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
17091 * tree-vect-generic.c (lower_vec_perm): Likewise.
17092 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17093 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
17094 element type to HOST_WIDE_INT.
17096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17097 Alan Hayward <alan.hayward@arm.com>
17098 David Sherwood <david.sherwood@arm.com>
17100 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
17101 rather than an int. Use plus_constant.
17102 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
17103 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
17105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17106 Alan Hayward <alan.hayward@arm.com>
17107 David Sherwood <david.sherwood@arm.com>
17109 * calls.c (emit_call_1, expand_call): Change struct_value_size from
17110 a HOST_WIDE_INT to a poly_int64.
17112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17113 Alan Hayward <alan.hayward@arm.com>
17114 David Sherwood <david.sherwood@arm.com>
17116 * calls.c (load_register_parameters): Cope with polynomial
17117 mode sizes. Require a constant size for BLKmode parameters
17118 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
17119 forces a parameter to be padded at the lsb end in order to
17120 fill a complete number of words, require the parameter size
17121 to be ordered wrt UNITS_PER_WORD.
17123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17124 Alan Hayward <alan.hayward@arm.com>
17125 David Sherwood <david.sherwood@arm.com>
17127 * reload1.c (spill_stack_slot_width): Change element type
17128 from unsigned int to poly_uint64_pod.
17129 (alter_reg): Treat mode sizes as polynomial.
17131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17132 Alan Hayward <alan.hayward@arm.com>
17133 David Sherwood <david.sherwood@arm.com>
17135 * reload.c (complex_word_subreg_p): New function.
17136 (reload_inner_reg_of_subreg, push_reload): Use it.
17138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17139 Alan Hayward <alan.hayward@arm.com>
17140 David Sherwood <david.sherwood@arm.com>
17142 * lra-constraints.c (process_alt_operands): Reject matched
17143 operands whose sizes aren't ordered.
17144 (match_reload): Refer to this check here.
17146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17147 Alan Hayward <alan.hayward@arm.com>
17148 David Sherwood <david.sherwood@arm.com>
17150 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
17151 that the mode size is in the set {1, 2, 4, 8, 16}.
17153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17154 Alan Hayward <alan.hayward@arm.com>
17155 David Sherwood <david.sherwood@arm.com>
17157 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
17158 Use plus_constant instead of gen_rtx_PLUS.
17160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17161 Alan Hayward <alan.hayward@arm.com>
17162 David Sherwood <david.sherwood@arm.com>
17164 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
17165 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
17166 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
17167 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
17168 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
17169 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
17170 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
17171 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
17172 * config/i386/i386.c (ix86_push_rounding): ...this new function.
17173 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
17175 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
17176 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
17177 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
17178 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
17179 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
17180 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
17181 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
17182 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
17183 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
17184 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
17186 * expr.c (emit_move_resolve_push): Treat the input and result
17187 of PUSH_ROUNDING as a poly_int64.
17188 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
17189 (emit_push_insn): Likewise.
17190 * lra-eliminations.c (mark_not_eliminable): Likewise.
17191 * recog.c (push_operand): Likewise.
17192 * reload1.c (elimination_effects): Likewise.
17193 * rtlanal.c (nonzero_bits1): Likewise.
17194 * calls.c (store_one_arg): Likewise. Require the padding to be
17195 known at compile time.
17197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17198 Alan Hayward <alan.hayward@arm.com>
17199 David Sherwood <david.sherwood@arm.com>
17201 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
17202 Use plus_constant instead of gen_rtx_PLUS.
17204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17205 Alan Hayward <alan.hayward@arm.com>
17206 David Sherwood <david.sherwood@arm.com>
17208 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
17209 rather than an int.
17211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17212 Alan Hayward <alan.hayward@arm.com>
17213 David Sherwood <david.sherwood@arm.com>
17215 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
17216 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
17217 via stack temporaries. Treat the mode size as polynomial too.
17219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17220 Alan Hayward <alan.hayward@arm.com>
17221 David Sherwood <david.sherwood@arm.com>
17223 * expr.c (expand_expr_real_2): When handling conversions involving
17224 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
17225 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
17226 as a poly_uint64 too.
17228 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17229 Alan Hayward <alan.hayward@arm.com>
17230 David Sherwood <david.sherwood@arm.com>
17232 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
17234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17235 Alan Hayward <alan.hayward@arm.com>
17236 David Sherwood <david.sherwood@arm.com>
17238 * combine.c (can_change_dest_mode): Handle polynomial
17239 REGMODE_NATURAL_SIZE.
17240 * expmed.c (store_bit_field_1): Likewise.
17241 * expr.c (store_constructor): Likewise.
17242 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
17243 and polynomial REGMODE_NATURAL_SIZE.
17244 (gen_lowpart_common): Likewise.
17245 * reginfo.c (record_subregs_of_mode): Likewise.
17246 * rtlanal.c (read_modify_subreg_p): Likewise.
17248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17249 Alan Hayward <alan.hayward@arm.com>
17250 David Sherwood <david.sherwood@arm.com>
17252 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
17253 numbers of elements.
17255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17256 Alan Hayward <alan.hayward@arm.com>
17257 David Sherwood <david.sherwood@arm.com>
17259 * match.pd: Cope with polynomial numbers of vector elements.
17261 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17262 Alan Hayward <alan.hayward@arm.com>
17263 David Sherwood <david.sherwood@arm.com>
17265 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
17266 in a POINTER_PLUS_EXPR.
17268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17269 Alan Hayward <alan.hayward@arm.com>
17270 David Sherwood <david.sherwood@arm.com>
17272 * omp-simd-clone.c (simd_clone_subparts): New function.
17273 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
17274 (ipa_simd_modify_function_body): Likewise.
17276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17277 Alan Hayward <alan.hayward@arm.com>
17278 David Sherwood <david.sherwood@arm.com>
17280 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
17281 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
17282 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
17283 (expand_vector_condition, vector_element): Likewise.
17284 (subparts_gt): New function.
17285 (get_compute_type): Use subparts_gt.
17286 (count_type_subparts): Delete.
17287 (expand_vector_operations_1): Use subparts_gt instead of
17288 count_type_subparts.
17290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17291 Alan Hayward <alan.hayward@arm.com>
17292 David Sherwood <david.sherwood@arm.com>
17294 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
17295 (vect_compile_time_alias): ...this new function. Do the calculation
17296 on poly_ints rather than trees.
17297 (vect_prune_runtime_alias_test_list): Update call accordingly.
17299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17300 Alan Hayward <alan.hayward@arm.com>
17301 David Sherwood <david.sherwood@arm.com>
17303 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
17305 (vect_schedule_slp_instance): Likewise.
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 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
17312 constant and extern definitions for variable-length vectors.
17313 (vect_get_constant_vectors): Note that the number of units
17314 is known to be constant.
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 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
17321 of units as polynomial. Choose between WIDE and NARROW based
17324 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17325 Alan Hayward <alan.hayward@arm.com>
17326 David Sherwood <david.sherwood@arm.com>
17328 * tree-vect-stmts.c (simd_clone_subparts): New function.
17329 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
17331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17332 Alan Hayward <alan.hayward@arm.com>
17333 David Sherwood <david.sherwood@arm.com>
17335 * tree-vect-stmts.c (vectorizable_call): Treat the number of
17336 vectors as polynomial. Use build_index_vector for
17337 IFN_GOMP_SIMD_LANE.
17339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17340 Alan Hayward <alan.hayward@arm.com>
17341 David Sherwood <david.sherwood@arm.com>
17343 * tree-vect-stmts.c (get_load_store_type): Treat the number of
17344 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
17345 for variable-length vectors.
17346 (vectorizable_mask_load_store): Treat the number of units as
17347 polynomial, asserting that it is constant if the condition has
17348 already been enforced.
17349 (vectorizable_store, vectorizable_load): Likewise.
17351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17352 Alan Hayward <alan.hayward@arm.com>
17353 David Sherwood <david.sherwood@arm.com>
17355 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
17356 of units as polynomial. Punt if we can't tell at compile time
17357 which vector contains the final result.
17359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17360 Alan Hayward <alan.hayward@arm.com>
17361 David Sherwood <david.sherwood@arm.com>
17363 * tree-vect-loop.c (vectorizable_induction): Treat the number
17364 of units as polynomial. Punt on SLP inductions. Use an integer
17365 VEC_SERIES_EXPR for variable-length integer reductions. Use a
17366 cast of such a series for variable-length floating-point
17369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17370 Alan Hayward <alan.hayward@arm.com>
17371 David Sherwood <david.sherwood@arm.com>
17373 * tree.h (build_index_vector): Declare.
17374 * tree.c (build_index_vector): New function.
17375 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
17376 of units as polynomial, forcibly converting it to a constant if
17377 vectorizable_reduction has already enforced the condition.
17378 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
17379 to create a {1,2,3,...} vector.
17380 (vectorizable_reduction): Treat the number of units as polynomial.
17381 Choose vectype_in based on the largest scalar element size rather
17382 than the smallest number of units. Enforce the restrictions
17385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17386 Alan Hayward <alan.hayward@arm.com>
17387 David Sherwood <david.sherwood@arm.com>
17389 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
17390 number of units as polynomial.
17392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17393 Alan Hayward <alan.hayward@arm.com>
17394 David Sherwood <david.sherwood@arm.com>
17396 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
17397 * target.def (autovectorize_vector_sizes): Return the vector sizes
17398 by pointer, using vector_sizes rather than a bitmask.
17399 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
17400 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
17401 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
17403 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
17404 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
17405 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
17406 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
17407 * omp-general.c (omp_max_vf): Likewise.
17408 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17409 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
17410 * tree-vect-loop.c (vect_analyze_loop): Likewise.
17411 * tree-vect-slp.c (vect_slp_bb): Likewise.
17412 * doc/tm.texi: Regenerate.
17413 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
17415 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
17416 the vector size as a poly_uint64 rather than an unsigned int.
17417 (current_vector_size): Change from an unsigned int to a poly_uint64.
17418 (get_vectype_for_scalar_type): Update accordingly.
17419 * tree.h (build_truth_vector_type): Take the size and number of
17420 units as a poly_uint64 rather than an unsigned int.
17421 (build_vector_type): Add a temporary overload that takes
17422 the number of units as a poly_uint64 rather than an unsigned int.
17423 * tree.c (make_vector_type): Likewise.
17424 (build_truth_vector_type): Take the number of units as a poly_uint64
17425 rather than an unsigned int.
17427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17428 Alan Hayward <alan.hayward@arm.com>
17429 David Sherwood <david.sherwood@arm.com>
17431 * target.def (get_mask_mode): Take the number of units and length
17432 as poly_uint64s rather than unsigned ints.
17433 * targhooks.h (default_get_mask_mode): Update accordingly.
17434 * targhooks.c (default_get_mask_mode): Likewise.
17435 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17436 * doc/tm.texi: Regenerate.
17438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17439 Alan Hayward <alan.hayward@arm.com>
17440 David Sherwood <david.sherwood@arm.com>
17442 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
17443 * omp-general.c (omp_max_vf): Likewise.
17444 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
17445 (expand_omp_simd): Handle polynomial safelen.
17446 * omp-low.c (omplow_simd_context): Add a default constructor.
17447 (omplow_simd_context::max_vf): Change from int to poly_uint64.
17448 (lower_rec_simd_input_clauses): Update accordingly.
17449 (lower_rec_input_clauses): Likewise.
17451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17452 Alan Hayward <alan.hayward@arm.com>
17453 David Sherwood <david.sherwood@arm.com>
17455 * tree-vectorizer.h (vect_nunits_for_cost): New function.
17456 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
17457 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
17458 (vect_analyze_slp_cost): Likewise.
17459 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
17460 (vect_model_load_cost): Likewise.
17462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17463 Alan Hayward <alan.hayward@arm.com>
17464 David Sherwood <david.sherwood@arm.com>
17466 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
17467 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
17468 from an unsigned int * to a poly_uint64_pod *.
17469 (calculate_unrolling_factor): New function.
17470 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
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 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
17477 from an unsigned int to a poly_uint64.
17478 (_loop_vec_info::slp_unrolling_factor): Likewise.
17479 (_loop_vec_info::vectorization_factor): Change from an int
17481 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
17482 (vect_get_num_vectors): New function.
17483 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
17484 (vect_get_num_copies): Use vect_get_num_vectors.
17485 (vect_analyze_data_ref_dependences): Change max_vf from an int *
17486 to an unsigned int *.
17487 (vect_analyze_data_refs): Change min_vf from an int * to a
17489 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17490 than an unsigned HOST_WIDE_INT.
17491 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
17492 (vect_analyze_data_ref_dependence): Change max_vf from an int *
17493 to an unsigned int *.
17494 (vect_analyze_data_ref_dependences): Likewise.
17495 (vect_compute_data_ref_alignment): Handle polynomial vf.
17496 (vect_enhance_data_refs_alignment): Likewise.
17497 (vect_prune_runtime_alias_test_list): Likewise.
17498 (vect_shift_permute_load_chain): Likewise.
17499 (vect_supportable_dr_alignment): Likewise.
17500 (dependence_distance_ge_vf): Take the vectorization factor as a
17501 poly_uint64 rather than an unsigned HOST_WIDE_INT.
17502 (vect_analyze_data_refs): Change min_vf from an int * to a
17504 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
17505 vfm1 as a poly_uint64 rather than an int. Make the same change
17506 for the returned bound_scalar.
17507 (vect_gen_vector_loop_niters): Handle polynomial vf.
17508 (vect_do_peeling): Likewise. Update call to
17509 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
17510 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
17512 * tree-vect-loop.c (vect_determine_vectorization_factor)
17513 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
17514 (vect_get_known_peeling_cost): Likewise.
17515 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
17516 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
17517 (vect_transform_loop): Likewise. Use the lowest possible VF when
17518 updating the upper bounds of the loop.
17519 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17520 rather than an int.
17521 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17522 polynomial unroll factors.
17523 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17524 (vect_make_slp_decision): Likewise.
17525 (vect_supported_load_permutation_p): Likewise, and polynomial
17527 (vect_analyze_slp_cost): Handle polynomial vf.
17528 (vect_slp_analyze_node_operations): Likewise.
17529 (vect_slp_analyze_bb_1): Likewise.
17530 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17531 than an unsigned HOST_WIDE_INT.
17532 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17533 (vectorizable_load): Handle polynomial vf.
17534 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
17536 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
17538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17539 Alan Hayward <alan.hayward@arm.com>
17540 David Sherwood <david.sherwood@arm.com>
17542 * match.pd: Handle bit operations involving three constants
17543 and try to fold one pair.
17545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17547 * tree-vect-loop-manip.c: Include gimple-fold.h.
17548 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17549 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17550 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17551 Add a path that uses a step of VF instead of 1, but disable it
17553 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17554 and niters_no_overflow parameters. Update calls to
17555 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17556 Create a new SSA name if the latter choses to use a ste other
17557 than zero, and return it via niters_vector_mult_vf_var.
17558 * tree-vect-loop.c (vect_transform_loop): Update calls to
17559 vect_do_peeling, vect_gen_vector_loop_niters and
17560 slpeel_make_loop_iterate_ntimes.
17561 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17562 (vect_gen_vector_loop_niters): Update declarations after above changes.
17564 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17566 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17567 128-bit round to integer instructions.
17568 (ceil<mode>2): Likewise.
17569 (btrunc<mode>2): Likewise.
17570 (round<mode>2): Likewise.
17572 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17574 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17575 unaligned VSX load/store on P8/P9.
17576 (expand_block_clear): Allow the use of unaligned VSX
17577 load/store on P8/P9.
17579 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17581 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17583 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17584 swap associated with both a load and a store.
17586 2018-01-02 Andrew Waterman <andrew@sifive.com>
17588 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17589 * config/riscv/riscv.md (clear_cache): Use it.
17591 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17593 * web.c: Remove out-of-date comment.
17595 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17597 * expr.c (fixup_args_size_notes): Check that any existing
17598 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17599 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17600 (emit_single_push_insn): ...here.
17602 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17604 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17605 (const_vector_encoded_nelts): New function.
17606 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17607 (const_vector_int_elt, const_vector_elt): Declare.
17608 * emit-rtl.c (const_vector_int_elt_1): New function.
17609 (const_vector_elt): Likewise.
17610 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17611 of CONST_VECTOR_ELT.
17613 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17615 * expr.c: Include rtx-vector-builder.h.
17616 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17617 directly on the tree encoding.
17618 (const_vector_from_tree): Likewise.
17619 * optabs.c: Include rtx-vector-builder.h.
17620 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17621 sequence of "u" values.
17622 * vec-perm-indices.c: Include rtx-vector-builder.h.
17623 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17624 directly on the vec_perm_indices encoding.
17626 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17628 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17629 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17630 * rtx-vector-builder.h: New file.
17631 * rtx-vector-builder.c: Likewise.
17632 * rtl.h (rtx_def::u2): Add a const_vector field.
17633 (CONST_VECTOR_NPATTERNS): New macro.
17634 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17635 (CONST_VECTOR_DUPLICATE_P): Likewise.
17636 (CONST_VECTOR_STEPPED_P): Likewise.
17637 (CONST_VECTOR_ENCODED_ELT): Likewise.
17638 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17639 (unwrap_const_vec_duplicate): Likewise.
17640 (const_vec_series_p): Check for a non-duplicated vector encoding.
17641 Say that the function only returns true for integer vectors.
17642 * emit-rtl.c: Include rtx-vector-builder.h.
17643 (gen_const_vec_duplicate_1): Delete.
17644 (gen_const_vector): Call gen_const_vec_duplicate instead of
17645 gen_const_vec_duplicate_1.
17646 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17647 (gen_const_vec_duplicate): Use rtx_vector_builder.
17648 (gen_const_vec_series): Likewise.
17649 (gen_rtx_CONST_VECTOR): Likewise.
17650 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17651 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17652 Build a new vector rather than modifying a CONST_VECTOR in-place.
17653 (handle_special_swappables): Update call accordingly.
17654 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17655 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17656 Build a new vector rather than modifying a CONST_VECTOR in-place.
17657 (handle_special_swappables): Update call accordingly.
17659 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17661 * simplify-rtx.c (simplify_const_binary_operation): Use
17662 CONST_VECTOR_ELT instead of XVECEXP.
17664 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17666 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17667 the selector elements to be different from the data elements
17668 if the selector is a VECTOR_CST.
17669 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17670 ssizetype for the selector.
17672 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17674 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17675 before testing each element individually.
17676 * tree-vect-generic.c (lower_vec_perm): Likewise.
17678 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17680 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17681 * selftest-run-tests.c (selftest::run_tests): Call it.
17682 * vector-builder.h (vector_builder::operator ==): New function.
17683 (vector_builder::operator !=): Likewise.
17684 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17685 (vec_perm_indices::all_from_input_p): New function.
17686 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17687 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17688 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17689 instead of reading the VECTOR_CST directly. Detect whether both
17690 vector inputs are the same before constructing the vec_perm_indices,
17691 and update the number of inputs argument accordingly. Use the
17692 utility functions added above. Only construct sel2 if we need to.
17694 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17696 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17697 the broadcast of the low byte.
17698 (expand_mult_highpart): Use an explicit encoding for the permutes.
17699 * optabs-query.c (can_mult_highpart_p): Likewise.
17700 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17701 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17702 (vectorizable_bswap): Likewise.
17703 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17704 explicit encoding for the power-of-2 permutes.
17705 (vect_permute_store_chain): Likewise.
17706 (vect_grouped_load_supported): Likewise.
17707 (vect_permute_load_chain): Likewise.
17709 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17711 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17712 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17713 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17714 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17715 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17716 (vect_gen_perm_mask_any): Likewise.
17718 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17720 * int-vector-builder.h: New file.
17721 * vec-perm-indices.h: Include int-vector-builder.h.
17722 (vec_perm_indices): Redefine as an int_vector_builder.
17723 (auto_vec_perm_indices): Delete.
17724 (vec_perm_builder): Redefine as a stand-alone class.
17725 (vec_perm_indices::vec_perm_indices): New function.
17726 (vec_perm_indices::clamp): Likewise.
17727 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17728 (vec_perm_indices::new_vector): New function.
17729 (vec_perm_indices::new_expanded_vector): Update for new
17730 vec_perm_indices class.
17731 (vec_perm_indices::rotate_inputs): New function.
17732 (vec_perm_indices::all_in_range_p): Operate directly on the
17733 encoded form, without computing elided elements.
17734 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17735 encoding. Update for new vec_perm_indices class.
17736 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17737 the given vec_perm_builder.
17738 (expand_vec_perm_var): Update vec_perm_builder constructor.
17739 (expand_mult_highpart): Use vec_perm_builder instead of
17740 auto_vec_perm_indices.
17741 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17742 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17743 or double series encoding as appropriate.
17744 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17745 vec_perm_indices instead of auto_vec_perm_indices.
17746 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17747 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17748 (vect_permute_store_chain): Likewise.
17749 (vect_grouped_load_supported): Likewise.
17750 (vect_permute_load_chain): Likewise.
17751 (vect_shift_permute_load_chain): Likewise.
17752 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17753 (vect_transform_slp_perm_load): Likewise.
17754 (vect_schedule_slp_instance): Likewise.
17755 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17756 (vectorizable_mask_load_store): Likewise.
17757 (vectorizable_bswap): Likewise.
17758 (vectorizable_store): Likewise.
17759 (vectorizable_load): Likewise.
17760 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17761 vec_perm_indices instead of auto_vec_perm_indices. Use
17762 tree_to_vec_perm_builder to read the vector from a tree.
17763 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17764 vec_perm_builder instead of a vec_perm_indices.
17765 (have_whole_vector_shift): Use vec_perm_builder and
17766 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17767 truncation to calc_vec_perm_mask_for_shift.
17768 (vect_create_epilog_for_reduction): Likewise.
17769 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17770 from auto_vec_perm_indices to vec_perm_indices.
17771 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17772 instead of changing individual elements.
17773 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17774 the vector in d.perm.
17775 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17776 from auto_vec_perm_indices to vec_perm_indices.
17777 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17778 instead of changing individual elements.
17779 (arm_vectorize_vec_perm_const): Use new_vector to install
17780 the vector in d.perm.
17781 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17782 Update vec_perm_builder constructor.
17783 (rs6000_expand_interleave): Likewise.
17784 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17785 (rs6000_expand_interleave): Likewise.
17787 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17789 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17790 to qimode could truncate the indices.
17791 * optabs.c (expand_vec_perm_var): Likewise.
17793 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17795 * Makefile.in (OBJS): Add vec-perm-indices.o.
17796 * vec-perm-indices.h: New file.
17797 * vec-perm-indices.c: Likewise.
17798 * target.h (vec_perm_indices): Replace with a forward class
17800 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17801 * optabs.h: Include vec-perm-indices.h.
17802 (expand_vec_perm): Delete.
17803 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17804 (expand_vec_perm_const): Declare.
17805 * target.def (vec_perm_const_ok): Replace with...
17806 (vec_perm_const): ...this new hook.
17807 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17808 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17809 * doc/tm.texi: Regenerate.
17810 * optabs.def (vec_perm_const): Delete.
17811 * doc/md.texi (vec_perm_const): Likewise.
17812 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17813 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17814 expand_vec_perm for constant permutation vectors. Assert that
17815 the mode of variable permutation vectors is the integer equivalent
17816 of the mode that is being permuted.
17817 * optabs-query.h (selector_fits_mode_p): Declare.
17818 * optabs-query.c: Include vec-perm-indices.h.
17819 (selector_fits_mode_p): New function.
17820 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17821 is defined, instead of checking whether the vec_perm_const_optab
17822 exists. Use targetm.vectorize.vec_perm_const instead of
17823 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17824 fit in the vector mode before using a variable permute.
17825 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17826 vec_perm_indices instead of an rtx.
17827 (expand_vec_perm): Replace with...
17828 (expand_vec_perm_const): ...this new function. Take the selector
17829 as a vec_perm_indices rather than an rtx. Also take the mode of
17830 the selector. Update call to shift_amt_for_vec_perm_mask.
17831 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17832 Use vec_perm_indices::new_expanded_vector to expand the original
17833 selector into bytes. Check whether the indices fit in the vector
17834 mode before using a variable permute.
17835 (expand_vec_perm_var): Make global.
17836 (expand_mult_highpart): Use expand_vec_perm_const.
17837 * fold-const.c: Includes vec-perm-indices.h.
17838 * tree-ssa-forwprop.c: Likewise.
17839 * tree-vect-data-refs.c: Likewise.
17840 * tree-vect-generic.c: Likewise.
17841 * tree-vect-loop.c: Likewise.
17842 * tree-vect-slp.c: Likewise.
17843 * tree-vect-stmts.c: Likewise.
17844 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17846 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17847 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17848 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17849 (aarch64_vectorize_vec_perm_const): ...this new function.
17850 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17851 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17852 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17853 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17854 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17855 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17856 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17858 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17859 check for NEON modes.
17860 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17861 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17862 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17863 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17865 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17866 the old VEC_PERM_CONST conditions.
17867 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17868 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17869 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17870 (ia64_vectorize_vec_perm_const_ok): Merge into...
17871 (ia64_vectorize_vec_perm_const): ...this new function.
17872 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17873 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17874 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17875 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17876 * config/mips/mips.c (mips_expand_vec_perm_const)
17877 (mips_vectorize_vec_perm_const_ok): Merge into...
17878 (mips_vectorize_vec_perm_const): ...this new function.
17879 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17880 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17881 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17882 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17883 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17884 (rs6000_expand_vec_perm_const): Delete.
17885 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17887 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17888 (altivec_expand_vec_perm_const_le): Take each operand individually.
17889 Operate on constant selectors rather than rtxes.
17890 (altivec_expand_vec_perm_const): Likewise. Update call to
17891 altivec_expand_vec_perm_const_le.
17892 (rs6000_expand_vec_perm_const): Delete.
17893 (rs6000_vectorize_vec_perm_const_ok): Delete.
17894 (rs6000_vectorize_vec_perm_const): New function.
17895 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17896 an element count and rtx array.
17897 (rs6000_expand_extract_even): Update call accordingly.
17898 (rs6000_expand_interleave): Likewise.
17899 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17900 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17901 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17902 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17903 (rs6000_expand_vec_perm_const): Delete.
17904 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17905 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17906 (altivec_expand_vec_perm_const_le): Take each operand individually.
17907 Operate on constant selectors rather than rtxes.
17908 (altivec_expand_vec_perm_const): Likewise. Update call to
17909 altivec_expand_vec_perm_const_le.
17910 (rs6000_expand_vec_perm_const): Delete.
17911 (rs6000_vectorize_vec_perm_const_ok): Delete.
17912 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17913 reference to the SPE evmerge intructions.
17914 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17915 an element count and rtx array.
17916 (rs6000_expand_extract_even): Update call accordingly.
17917 (rs6000_expand_interleave): Likewise.
17918 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17919 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17921 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17923 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17925 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17926 vector mode and that that mode matches the mode of the data
17928 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17929 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17930 directly using expand_vec_perm_1 when forcing selectors into
17932 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17934 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17936 * optabs-query.h (can_vec_perm_p): Delete.
17937 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17938 * optabs-query.c (can_vec_perm_p): Split into...
17939 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17940 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17941 particular selector is valid.
17942 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17943 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17944 (vect_grouped_load_supported): Likewise.
17945 (vect_shift_permute_load_chain): Likewise.
17946 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17947 (vect_transform_slp_perm_load): Likewise.
17948 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17949 (vectorizable_bswap): Likewise.
17950 (vect_gen_perm_mask_checked): Likewise.
17951 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17952 implementations of variable permutation vectors into account
17953 when deciding which selector to use.
17954 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17955 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17956 with a false third argument.
17957 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17958 to test whether the constant selector is valid and can_vec_perm_var_p
17959 to test whether a variable selector is valid.
17961 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17963 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17964 * optabs-query.c (can_vec_perm_p): Likewise.
17965 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17966 instead of vec_perm_indices.
17967 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17968 (vect_gen_perm_mask_checked): Likewise,
17969 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17970 (vect_gen_perm_mask_checked): Likewise,
17972 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17974 * optabs-query.h (qimode_for_vec_perm): Declare.
17975 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17976 (qimode_for_vec_perm): ...this new function.
17977 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17979 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17981 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17982 does not have a conditional at the top.
17984 2018-01-02 Richard Biener <rguenther@suse.de>
17986 * ipa-inline.c (big_speedup_p): Fix expression.
17988 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17991 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17994 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17998 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17999 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
18000 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
18001 cond_taken_branch_cost 3->4.
18003 2018-01-01 Jakub Jelinek <jakub@redhat.com>
18005 PR tree-optimization/83581
18006 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
18007 TODO_cleanup_cfg if any changes have been made.
18009 PR middle-end/83608
18010 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
18011 convert_modes if target mode has the right side, but different mode
18014 PR middle-end/83609
18015 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
18016 last argument when extracting from CONCAT. If either from_real or
18017 from_imag is NULL, use expansion through memory. If result is not
18018 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
18019 the parts directly to inner mode, if even that fails, use expansion
18022 PR middle-end/83623
18023 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
18024 check for bswap in mode rather than HImode and use that in expand_unop
18027 Copyright (C) 2018 Free Software Foundation, Inc.
18029 Copying and distribution of this file, with or without modification,
18030 are permitted in any medium without royalty provided the copyright
18031 notice and this notice are preserved.