1 2018-06-27 Jeff Law <law@redhat.com>
3 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
4 (MULTILIB_DIRNAMES): Similarly.
6 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
8 * gimple.h (gimple_return_retbnd): Delete.
9 (gimple_return_set_retbnd): Likewise.
10 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
11 gimple_return_set_retbnd.
12 * gimple-pretty-print.c (dump_gimple_return): Remove call to
13 gimple_return_retbnd and adjust.
14 * tree-inline.h (struct copy_body_data): Remove retbnd field.
15 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
16 Explicitly return NULL in a couple more cases. Move assertion
17 on debug statements and remove unreachable code.
18 (reset_debug_binding): Do not test id->retbnd.
19 (expand_call_inline): Do not set it.
21 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
23 * configure.ac: Add --disable-gcov option.
24 * configure: Regenerate.
25 * Makefile.in: Honour @enable_gcov@.
26 * doc/install.texi: Document --disable-gcov.
28 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30 * config/arm/arm-cpus.in (cortex-a76): New entry.
31 (cortex-a76.cortex-a55): Likewise.
32 * config/arm/arm-tables.opt: Regenerate.
33 * config/arm/arm-tune.md: Likewise.
34 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
35 * doc/invoke.texi (ARM Options): Document cortex-a76 and
36 cortex-a76.cortex-a55.
38 2018-06-27 Tamar Christina <tamar.christina@arm.com>
41 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
43 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
45 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
47 (EPILOGUE_USES): Likewise.
49 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
51 * tree-inline.c (remap_location): New function extracted from...
52 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
53 (copy_phis_for_bb): ...here. Call remap_location.
54 (copy_cfg_body): Adjust call to copy_edges_for_bb.
56 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
58 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
59 unaligned vsx for 16B memset.
61 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
64 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
65 ieee128_float_type_node to long_double_type_node unless
66 TARGET_LONG_DOUBLE_128 is set.
68 2018-06-26 David Malcolm <dmalcolm@redhat.com>
70 * cfgloop.c (get_loop_location): Convert return type from
71 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
72 by implicit construction from rtx_insn *, and using
73 dump_user_location_t::from_function_decl for the fallback case.
74 * cfgloop.h (get_loop_location): Convert return type from
75 location_t to dump_user_location_t.
76 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
77 dump_printf_loc to pass in a dump_location_t rather than a
78 location_t, via the gimple stmt.
79 * coverage.c (get_coverage_counts): Update calls to
80 dump_printf_loc to pass in dump_location_t rather than a
82 * doc/optinfo.texi (Dump types): Convert example of
83 dump_printf_loc from taking "locus" to taking "insn". Update
84 description of the "_loc" calls to cover dump_location_t.
85 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
87 (dump_user_location_t::dump_user_location_t): New constructors,
88 from gimple *stmt and rtx_insn *.
89 (dump_user_location_t::from_function_decl): New function.
90 (dump_loc): Make static.
91 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
92 const dump_location_t &.
93 (dump_generic_expr_loc): Delete.
94 (dump_printf_loc): Convert param "loc" from location_t to
95 const dump_location_t &.
96 (selftest::test_impl_location): New function.
97 (selftest::dumpfile_c_tests): New function.
98 * dumpfile.h: Include "profile-count.h".
99 (class dump_user_location_t): New class.
100 (struct dump_impl_location_t): New struct.
101 (class dump_location_t): New class.
102 (dump_printf_loc): Convert 2nd param from source_location to
103 const dump_location_t &.
104 (dump_generic_expr_loc): Delete.
105 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
106 const dump_location_t &.
107 * gimple-fold.c (fold_gimple_assign): Update call to
108 dump_printf_loc to pass in a dump_location_t rather than a
109 location_t, via the gimple stmt.
110 (gimple_fold_call): Likewise.
111 * gimple-loop-interchange.cc
112 (loop_cand::analyze_iloop_reduction_var): Update for change to
113 check_reduction_path.
114 (tree_loop_interchange::interchange): Update for change to
116 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
117 change in return-type of find_loop_location.
118 (graphite_regenerate_ast_isl): Likewise.
119 * graphite-optimize-isl.c (optimize_isl): Likewise.
120 * graphite.c (graphite_transform_loops): Likewise.
121 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
122 pass in a dump_location_t rather than a location_t, via the
124 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
125 * ipa.c (walk_polymorphic_call_targets): Likewise.
126 * loop-unroll.c (report_unroll): Convert "locus" param from
127 location_t to dump_location_t.
128 (decide_unrolling): Update for change to get_loop_location's
130 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
131 location_t to dump_user_location_t.
132 (grid_find_single_omp_among_assignments_1): Updates calls to
133 dump_printf_loc to pass in a dump_location_t rather than a
134 location_t, via the gimple stmt.
135 (grid_parallel_clauses_gridifiable): Convert "tloc" from
136 location_t to dump_location_t. Updates calls to dump_printf_loc
137 to pass in a dump_location_t rather than a location_t, via the
139 (grid_inner_loop_gridifiable_p): Likewise.
140 (grid_dist_follows_simple_pattern): Likewise.
141 (grid_gfor_follows_tiling_pattern): Likewise.
142 (grid_target_follows_gridifiable_pattern): Likewise.
143 (grid_attempt_target_gridification): Convert initialization
144 of local "grid" from memset to zero-initialization; FIXME: does
145 this require C++11? Update call to dump_printf_loc to pass in a
146 optinfo_location rather than a location_t, via the gimple stmt.
147 * profile.c (read_profile_edge_counts): Updates call to
148 dump_printf_loc to pass in a dump_location_t rather than a
150 (compute_branch_probabilities): Likewise.
151 * selftest-run-tests.c (selftest::run_tests): Call
153 * selftest.h (dumpfile_c_tests): New decl.
154 * tree-loop-distribution.c (pass_loop_distribution::execute):
155 Update for change in return type of find_loop_location.
156 * tree-parloops.c (parallelize_loops): Likewise.
157 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
158 "locus" from location_t to dump_user_location_t.
159 (canonicalize_loop_induction_variables): Likewise.
160 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
161 for change in return type of find_loop_location.
162 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
163 to dump_printf_loc to pass in a dump_location_t rather than a
164 location_t, via the stmt.
165 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
167 * tree-vect-loop-manip.c (find_loop_location): Convert return
168 type from source_location to dump_user_location_t.
169 (vect_do_peeling): Update for above change.
170 (vect_loop_versioning): Update for change in type of
172 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
173 from location_t to dump_user_location_t.
174 (vect_estimate_min_profitable_iters): Update for change in type
176 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
177 location_t to dump_location_t.
178 (vect_slp_bb): Update for change in type of vect_location.
179 * tree-vectorizer.c (vect_location): Convert from source_location
180 to dump_user_location_t.
181 (try_vectorize_loop_1): Update for change in vect_location's type.
182 (vectorize_loops): Likewise.
183 (increase_alignment): Likewise.
184 * tree-vectorizer.h (vect_location): Convert from source_location
185 to dump_user_location_t.
186 (find_loop_location): Convert return type from source_location to
187 dump_user_location_t.
188 (check_reduction_path): Convert 1st param from location_t to
189 dump_user_location_t.
190 * value-prof.c (check_counter): Update call to dump_printf_loc to
191 pass in a dump_user_location_t rather than a location_t; update
192 call to error_at for change in type of "locus".
193 (check_ic_target): Update call to dump_printf_loc to
194 pass in a dump_user_location_t rather than a location_t, via the
197 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
199 * config/s390/s390.h (enum processor_flags): Do not use
200 default tune parameter when -march was specified.
202 2018-06-26 Jakub Jelinek <jakub@redhat.com>
205 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
206 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
209 2018-06-26 Richard Biener <rguenther@suse.de>
211 PR tree-optimization/86287
213 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
214 (vect_analyze_loop): Initialize n_stmts.
216 2018-06-26 Richard Biener <rguenther@suse.de>
219 * fold-const.c (fold_convertible_p): Pointer extension
222 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
225 * dwarf2out.c (loc_list_has_views): Adjust comments.
226 (dw_loc_list): Split single cross-partition range with
229 2018-06-25 Jeff Law <law@redhat.com>
231 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
232 on -mbig-switch by default.
234 * config/v850/predicates.md (const_float_1_operand): Fix match_code
236 (const_float_0_operand): Remove unused predicate.
237 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
238 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
239 (recipsf2): New expander. Original pattern now called
241 (recipdf2, recipdf2_insn): Similarly.
242 (rsqrtsf2, rsqrtsf2_insn): Similarly
243 (rsqrtdf2, rsqrtdf2_insn): Similarly
245 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
247 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
248 Simplify logic for FreeBSD (twice).
250 2018-06-25 Martin Sebor <msebor@redhat.com>
252 PR tree-optimization/86204
253 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
254 a strnlen result if it's less than the length of the string.
256 2018-06-25 Martin Sebor <msebor@redhat.com>
258 PR tree-optimization/85700
259 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
260 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
261 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
263 2018-06-25 Martin Sebor <msebor@redhat.com>
265 * doc/extend.texi (Zero-length arrays): Update and clarify.
267 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
269 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
270 added IEEE/IBM long double multilib support on PowerPC little
271 endian Linux systems.
272 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
273 (MULTILIB_DEFAULTS): Likewise.
274 * config/rs6000/rs6000.c (rs6000_option_override_internal):
276 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
277 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
278 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
280 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
283 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
284 (REORDER_45): Likewise.
286 2018-06-25 Jeff Law <law@redhat.com>
288 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
289 dividend to 32 bits. Adjust length.
290 (udivmodhi4): Cleanup output template. Fix length.
292 2018-06-25 Carl Love <cel@us.ibm.com>
294 * config/rs6000/vsx.md: Change word selector to prefered location.
296 2018-06-25 Richard Biener <rguenther@suse.de>
298 PR tree-optimization/86304
299 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
300 epilogue-if-converted loops as well.
302 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
304 * lto-section-out.c (lto_begin_section): Do not print section
305 name for noaddr and unnumbered dumps.
307 2018-06-25 Richard Biener <rguenther@suse.de>
309 * tree-vectorizer.h (struct vec_info_shared): New structure
310 with parts split out from struct vec_info and loop_nest from
311 struct _loop_vec_info.
312 (struct vec_info): Adjust accordingly.
313 (struct _loop_vec_info): Likewise.
314 (LOOP_VINFO_LOOP_NEST): Adjust.
315 (LOOP_VINFO_DATAREFS): Likewise.
316 (LOOP_VINFO_DDRS): Likewise.
317 (struct _bb_vec_info): Likewise.
318 (BB_VINFO_DATAREFS): Likewise.
319 (BB_VINFO_DDRS): Likewise.
320 (struct _stmt_vec_info): Add dr_aux member.
321 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
322 (DR_MISALIGNMENT_UNINITIALIZED): New.
323 (set_dr_misalignment): Adjust.
324 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
325 (vect_analyze_loop): Adjust prototype.
326 (vect_analyze_loop_form): Likewise.
327 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
328 Compute dependences lazily.
329 (vect_record_base_alignments): Use shared datarefs/ddrs.
330 (vect_verify_datarefs_alignment): Likewise.
331 (vect_analyze_data_refs_alignment): Likewise.
332 (vect_analyze_data_ref_accesses): Likewise.
333 (vect_analyze_data_refs): Likewise.
334 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
335 constructor parameter for shared part.
336 (vect_analyze_loop_form): Pass in shared part and adjust.
337 (vect_analyze_loop_2): Pass in storage for the number of
338 stmts. Move loop nest finding to the caller. Compute
340 (vect_analyze_loop): Pass in shared part.
341 (vect_transform_loop): Verify shared datarefs are unchanged.
342 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
343 constructor parameter for shared part.
344 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
345 (vect_slp_bb): Verify shared datarefs are unchanged before
347 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
349 (new_stmt_vec_info): Initialize DR_AUX misalignment to
350 DR_MISALIGNMENT_UNINITIALIZED.
351 * tree-vectorizer.c (vec_info::vec_info): Add constructor
352 parameter for shared part.
353 (vec_info::~vec_info): Adjust.
354 (vec_info_shared::vec_info_shared): New.
355 (vec_info_shared::~vec_info_shared): Likewise.
356 (vec_info_shared::save_datarefs): Likewise.
357 (vec_info_shared::check_datarefs): Likewise.
358 (try_vectorize_loop_1): Construct shared part live for analyses
359 of a single loop for multiple vector sizes.
360 * tree-parloops.c (gather_scalar_reductions): Adjust.
362 2018-06-25 Richard Biener <rguenther@suse.de>
364 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
365 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
366 (vect_analyze_data_refs): Remove similar code from here and
367 simplify accordingly.
369 2018-06-25 Richard Biener <rguenther@suse.de>
371 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
372 for reverse storage order accesses rather than asserting
373 they cannot happen here.
375 2018-06-25 Tom de Vries <tdevries@suse.de>
378 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
379 Use data16 instead of .byte for insn prefix.
381 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
384 * parser.c (make_char_string_pack): Pass this literal chars
385 through cpp_interpret_string.
386 (cp_parser_userdef_numeric_literal): Check the result of
387 make_char_string_pack.
389 2018-06-24 Maya Rashish <coypu@sdf.org>
391 * ginclude/stddef.h: Simplify conditions around avoiding
392 re-definition of __size_t.
394 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
396 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
397 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
399 2018-06-22 Maya Rashish <coypu@sdf.org>
401 * doc/invoke.texi (mno-fancy-math-387): Update for changes
402 made to OpenBSD and NetBSD through the years.
404 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
406 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
407 behavior of vec_pack (vector double, vector double) to match
408 behavior of vec_float2 (vector double, vector double).
410 2018-06-22 Olivier Hainque <hainque@adacore.com>
412 * gimplify.c (gimplify_function_tree): Prevent creation
413 of a trampoline for the address of the current function
414 passed to entry/exit instrumentation hooks.
416 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
419 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
422 2018-06-22 Martin Liska <mliska@suse.cz>
424 PR tree-optimization/86263
425 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
426 Bail out if is_enabled is false.
427 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
429 (jump_table_cluster::is_enabled): New function.
431 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
433 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
434 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
435 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
436 (lto_input_ts_binfo_tree_pointers): Likewise.
437 * tree-streamer-out.c (streamer_write_tree_bitfields,
438 write_ts_binfo_tree_pointers): Likewise.
439 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
441 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
443 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
445 2018-06-22 Martin Liska <mliska@suse.cz>
447 * symbol-summary.h (get): Make it pure and inline move
448 functionality from ::get function.
449 (get): Remove and inline into ::get and ::get_create.
450 (get_create): Move code from ::get function.
452 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
455 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
456 -x assembler-with-cpp.
458 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
460 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
461 _FILE_OFFSET_BITS=64 for C++.
463 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
465 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
466 conversion insn that shows up when pr85657-3.c is compiled using
467 IEEE 128-bit long double.
468 (neg<mode>2_internal): Use the correct mode to check whether the
469 mode is IBM extended.
470 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
471 multiply and divide external functions from being created more
474 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
476 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
478 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
479 the edge can be forwarded.
480 (cfg_layout_merge_blocks): Likewise.
482 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
484 * except.c (finish_eh_generation): Commit edge insertions only after
485 the EH edges have been redirected from post-landing to landing pads.
487 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
489 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
490 create_tmp_var_for to create the FRAME decl.
491 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
493 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
495 * tree-inline.c (copy_edges_for_bb): Minor tweak.
496 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
497 debug statement when resetting its value.
498 (expand_call_inline): Copy the locus of the call onto the assignment
499 of the return value, if any. Use local variable in more cases.
501 2018-06-21 Martin Liska <mliska@suse.cz>
503 * ipa-pure-const.c (propagate_nothrow): Use
504 funct_state_summaries->get.
505 (dump_malloc_lattice): Likewise.
506 (propagate_malloc): Likewise.
508 2018-06-21 Richard Biener <rguenther@suse.de>
510 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
511 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
512 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
514 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
515 BLOCK_ABSTRACT_ORIGIN unconditionally.
517 2018-06-21 David Malcolm <dmalcolm@redhat.com>
519 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
521 * ipa-reference.c (ipa_reference_c_finalize): Delete
522 ipa_ref_opt_sum_summaries and set it to NULL.
524 2018-06-21 Tom de Vries <tdevries@suse.de>
526 PR tree-optimization/85859
527 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
528 test with comment from bb_no_side_effects_p.
530 2018-06-21 Richard Biener <rguenther@suse.de>
532 PR tree-optimization/86232
533 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
534 max for constant niter.
536 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
538 * config/aarch64/aarch64-simd.md
539 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
541 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
543 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
544 Make opernads of the unspec commutative.
546 2018-06-21 Richard Biener <rguenther@suse.de>
548 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
549 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
550 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
551 (vect_analyze_data_ref_dependence): Re-order checks to deal with
553 (vect_record_base_alignments): Do not record base alignment
554 for gathers or scatters.
555 (vect_compute_data_ref_alignment): Drop return value that is always
556 true. Bail out early for gathers or scatters.
557 (vect_enhance_data_refs_alignment): Bail out early for gathers
559 (vect_find_same_alignment_drs): Likewise.
560 (vect_analyze_data_refs_alignment): Remove dead code.
561 (vect_slp_analyze_and_verify_node_alignment): Likewise.
562 (vect_analyze_data_refs): For possible gathers or scatters do
563 not create an alternate DR, just check their possible validity
564 and mark them. Adjust DECL_NONALIASED handling to not rely
566 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
567 update inits of gathers or scatters.
568 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
569 Also copy gather/scatter flag to pattern vinfo.
571 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
573 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
574 behavior of vec_packsu (vector unsigned long long, vector unsigned
575 long long) to match behavior of vec_packs with same signature.
577 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
578 Thomas Schwinge <thomas@codesourcery.com>
579 Cesar Philippidis <cesar@codesourcery.com>
581 * gimplify.c (gimplify_scan_omp_clauses): Add support for
582 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
583 (gimplify_adjust_omp_clauses): Likewise.
584 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
585 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
586 (gimplify_omp_target_update): Update handling of acc update and
588 * omp-low.c (install_var_field): Remove unused parameter
589 base_pointers_restrict.
590 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
591 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
593 (omp_target_base_pointers_restrict_p): Delete.
594 (scan_omp_target): Update call to scan_sharing_clauses.
595 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
597 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
598 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
599 (convert_local_omp_clauses): Likewise.
600 * tree-pretty-print.c (dump_omp_clause): Likewise.
601 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
603 (omp_clause_code_name): Likewise.
605 2018-06-20 Jakub Jelinek <jakub@redhat.com>
608 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
611 PR tree-optimization/86231
612 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
613 anti-range don't overwrite *vr0min before using it to compute *vr0max.
615 2018-06-20 Tom de Vries <tdevries@suse.de>
617 PR tree-optimization/86097
618 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
619 iv type if signedness of iv type is not the same as that of *nit.
621 2018-06-20 Jakub Jelinek <jakub@redhat.com>
623 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
624 EDGE_EH edges, verify they are all EDGE_EH.
626 2018-06-20 Maya Rashish <coypu@sdf.org>
628 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
630 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
632 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
633 * config/aarch64/aarch64.c (xgene1_tunings): Add
634 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
635 (aarch64_mode_valid_for_sched_fusion_p):
637 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
638 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
640 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
642 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
643 * config/aarch64/iterators.md (VQ2): New mode iterator.
645 2018-06-20 Martin Liska <mliska@suse.cz>
647 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
648 Change default ratio from 10 to 8.
650 2018-06-20 Martin Liska <mliska@suse.cz>
652 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
654 (bit_test_cluster::find_bit_tests): Likewise.
655 (switch_decision_tree::analyze_switch_statement): Find clusters.
656 * tree-switch-conversion.h (struct jump_table_cluster): Document
659 2018-06-20 Martin Liska <mliska@suse.cz>
661 * tree-switch-conversion.c (switch_conversion::collect):
662 Record m_uniq property.
663 (switch_conversion::expand): Bail out for special conditions.
664 (group_cluster::~group_cluster): New.
665 (group_cluster::group_cluster): Likewise.
666 (group_cluster::dump): Likewise.
667 (jump_table_cluster::emit): New.
668 (switch_decision_tree::fix_phi_operands_for_edges): New.
669 (struct case_node): Remove struct.
670 (jump_table_cluster::can_be_handled): New.
671 (case_values_threshold): Moved to header.
672 (reset_out_edges_aux): Likewise.
673 (jump_table_cluster::is_beneficial): New.
674 (bit_test_cluster::can_be_handled): Likewise.
675 (add_case_node): Remove.
676 (bit_test_cluster::is_beneficial): New.
677 (case_bit_test::cmp): New.
678 (bit_test_cluster::emit): New.
679 (expand_switch_as_decision_tree_p): Remove.
680 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
681 (fix_phi_operands_for_edge): Likewise.
682 (switch_decision_tree::analyze_switch_statement): New.
683 (compute_cases_per_edge): Move ...
684 (switch_decision_tree::compute_cases_per_edge): ... here.
685 (try_switch_expansion): Likewise.
686 (switch_decision_tree::try_switch_expansion): Likewise.
687 (record_phi_operand_mapping): Likewise.
688 (switch_decision_tree::record_phi_operand_mapping): Likewise.
689 (emit_case_decision_tree): Likewise.
690 (switch_decision_tree::emit): Likewise.
691 (balance_case_nodes): Likewise.
692 (switch_decision_tree::balance_case_nodes): Likewise.
693 (dump_case_nodes): Likewise.
694 (switch_decision_tree::dump_case_nodes): Likewise.
695 (emit_jump): Likewise.
696 (switch_decision_tree::emit_jump): Likewise.
697 (emit_cmp_and_jump_insns): Likewise.
698 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
699 (emit_case_nodes): Likewise.
700 (switch_decision_tree::emit_case_nodes): Likewise.
701 (conditional_probability): Remove.
702 * tree-switch-conversion.h (enum cluster_type): New.
704 (struct cluster): Likewise.
705 (cluster::cluster): Likewise.
706 (struct simple_cluster): Likewise.
707 (simple_cluster::simple_cluster): Likewise.
708 (struct group_cluster): Likewise.
709 (struct jump_table_cluster): Likewise.
710 (struct bit_test_cluster): Likewise.
711 (struct min_cluster_item): Likewise.
712 (struct case_tree_node): Likewise.
713 (case_tree_node::case_tree_node): Likewise.
714 (jump_table_cluster::case_values_threshold): Likewise.
715 (struct case_bit_test): Likewise.
716 (struct switch_decision_tree): Likewise.
717 (struct switch_conversion): Likewise.
718 (switch_decision_tree::reset_out_edges_aux): Likewise.
720 2018-06-20 Martin Liska <mliska@suse.cz>
722 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
723 (hoist_edge_and_branch_if_true): Likewise.
724 (expand_switch_using_bit_tests_p): Likewise.
725 (struct case_bit_test): Likewise.
726 (case_bit_test_cmp): Likewise.
727 (emit_case_bit_tests): Likewise.
728 (switch_conversion::switch_conversion): New class.
729 (struct switch_conv_info): Remove old struct.
730 (collect_switch_conv_info): More to ...
731 (switch_conversion::collect): ... this.
732 (check_range): Likewise.
733 (switch_conversion::check_range): Likewise.
734 (check_all_empty_except_final): Likewise.
735 (switch_conversion::check_all_empty_except_final): Likewise.
736 (check_final_bb): Likewise.
737 (switch_conversion::check_final_bb): Likewise.
738 (create_temp_arrays): Likewise.
739 (switch_conversion::create_temp_arrays): Likewise.
740 (free_temp_arrays): Likewise.
741 (gather_default_values): Likewise.
742 (switch_conversion::gather_default_values): Likewise.
743 (build_constructors): Likewise.
744 (switch_conversion::build_constructors): Likewise.
745 (constructor_contains_same_values_p): Likewise.
746 (switch_conversion::contains_same_values_p): Likewise.
747 (array_value_type): Likewise.
748 (switch_conversion::array_value_type): Likewise.
749 (build_one_array): Likewise.
750 (switch_conversion::build_one_array): Likewise.
751 (build_arrays): Likewise.
752 (switch_conversion::build_arrays): Likewise.
753 (gen_def_assigns): Likewise.
754 (switch_conversion::gen_def_assigns): Likewise.
755 (prune_bbs): Likewise.
756 (switch_conversion::prune_bbs): Likewise.
757 (fix_phi_nodes): Likewise.
758 (switch_conversion::fix_phi_nodes): Likewise.
759 (gen_inbound_check): Likewise.
760 (switch_conversion::gen_inbound_check): Likewise.
761 (process_switch): Use the newly created class.
762 (switch_conversion::expand): New.
763 (switch_conversion::~switch_conversion): New.
764 * tree-switch-conversion.h: New file.
766 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
768 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
769 tree-vect-patterns.c.
770 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
771 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
772 (vect_recog_sad_pattern): Likewise.
773 (vect_recog_widen_sum_pattern): Likewise.
774 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
775 (vect_recog_widen_shift_pattern): Remove the type_in argument.
776 (vect_recog_rotate_pattern): Likewise.
777 (vect_recog_mult_pattern): Likewise.
778 (vect_recog_vector_vector_shift_pattern): Likewise.
779 (vect_recog_divmod_pattern): Likewise.
780 (vect_recog_mixed_size_cond_pattern): Likewise.
781 (vect_recog_bool_pattern): Likewise.
782 (vect_recog_mask_conversion_pattern): Likewise.
783 (vect_try_gather_scatter_pattern): Likewise.
784 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
785 (vect_recog_over_widening_pattern): Likewise.
786 (vect_recog_gather_scatter_pattern): Likewise.
787 (vect_recog_func_ptr): Move from tree-vectorizer.h
788 (vect_vect_recog_func_ptrs): Move further down the file.
789 (vect_recog_func): Likewise. Remove the third argument.
790 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
791 (vect_pattern_recog_1): Expect the pattern function to do any
792 necessary target tests. Also expect it to provide a vector type.
793 Remove the type_in handling.
795 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
797 * tree-vect-patterns.c (vect_pattern_detected): New function.
798 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
799 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
800 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
801 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
802 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
803 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
804 (vect_recog_mask_conversion_pattern)
805 (vect_try_gather_scatter_pattern): Likewise.
807 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
809 * tree-vect-patterns.c (vect_get_internal_def): New function.
810 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
811 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
812 (search_type_for_mask_1): Use it.
814 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
816 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
817 redundant WIDEN_SUM_EXPR handling.
818 (vect_recog_sad_pattern): Likewise.
820 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
822 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
823 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
824 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
825 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
826 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
828 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
830 * tree-vect-stmts.c (vectorizable_call): Make sure that we
831 use the stmt_vec_info of the original bb statement for the
832 new zero assignment, even if the call is part of a pattern.
834 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
836 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
837 that the sequence is attached to the original statement rather
838 than the pattern statement.
839 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
840 PATTERN_DEF_SEQ from the original statement rather than
841 the main pattern statement.
842 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
843 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
844 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
846 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
848 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
849 definition statements before the early exit for statements that aren't
851 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
853 (vect_transform_loop): ...here. Process pattern definition
854 statements without first checking whether the main pattern
855 statement is live or relevant.
857 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
859 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
860 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
862 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
864 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
865 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
866 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
867 (expand_block_compare): Change select_block_compare_mode call.
868 (expand_strncmp_align_check): Use new functions, fix comment.
869 (emit_final_str_compare_gpr): New function.
870 (expand_strn_compare): Refactor and clean up code.
871 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
873 2018-06-19 Tony Reix <tony.reix@atos.com>
874 Damien Bergamini <damien.bergamini@atos.com>
875 David Edelsohn <dje.gcc@gmail.com>
877 * collect2.c (static_obj): New variable.
878 (static_libs): New variable.
879 (is_in_list): Uncomment declaration.
880 (main): Track AIX libraries linked statically.
881 (is_in_list): Uncomment definition.
882 (scan_prog_file): Don't add AIX shared libraries initializer
883 to constructor list if linking statically.
885 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
887 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
889 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
891 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
893 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
896 2018-06-19 Martin Liska <mliska@suse.cz>
898 * config/i386/i386.c (ix86_can_inline_p): Do not use
899 ipa_fn_summaries::get_create.
900 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
902 (devirtualization_time_bonus): Likewise.
903 (ipcp_propagate_stage): Likewise.
904 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
905 (edge_set_predicate): Likewise.
906 (evaluate_conditions_for_known_args): Likewise.
907 (evaluate_properties_for_edge): Likewise.
908 (ipa_call_summary::reset): Tranform to ...
909 (ipa_call_summary::~ipa_call_summary): ... this.
910 (ipa_fn_summary::reset): Transform to ...
911 (ipa_fn_summary::~ipa_fn_summary): ... this.
912 (ipa_fn_summary_t::remove): Rename to ...
913 (ipa_fn_summary_t::remove_callees): ... this.
914 (ipa_fn_summary_t::duplicate): Use placement new
915 instead of memory copy.
916 (ipa_call_summary_t::duplicate): Likewise.
917 (ipa_call_summary_t::remove): Remove.
918 (dump_ipa_call_summary): Change get_create to get.
919 (ipa_dump_fn_summary): Dump only when summary exists.
920 (analyze_function_body): Use symbol_summary::get instead
922 (compute_fn_summary): Likewise.
923 (estimate_edge_devirt_benefit): Likewise.
924 (estimate_edge_size_and_time): Likewise.
925 (inline_update_callee_summaries): Likewise.
926 (remap_edge_change_prob): Likewise.
927 (remap_edge_summaries): Likewise.
928 (ipa_merge_fn_summary_after_inlining): Likewise.
929 (write_ipa_call_summary): Likewise.
930 (ipa_fn_summary_write): Likewise.
931 (ipa_free_fn_summary): Likewise.
932 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
933 (struct ipa_call_summary): Likewise.
934 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
936 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
937 (estimate_size_after_inlining): Likewise.
938 (estimate_growth): Likewise.
939 (growth_likely_positive): Likewise.
940 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
941 (inline_call): Likewise.
942 * ipa-inline.c (caller_growth_limits): Likewise.
943 (can_inline_edge_p): Likewise.
944 (can_inline_edge_by_limits_p): Likewise.
945 (compute_uninlined_call_time): Likewise.
946 (compute_inlined_call_time): Likewise.
947 (want_inline_small_function_p): Likewise.
948 (edge_badness): Likewise.
949 (update_caller_keys): Likewise.
950 (update_callee_keys): Likewise.
951 (inline_small_functions): Likewise.
952 (inline_to_all_callers_1): Likewise.
953 (dump_overall_stats): Likewise.
954 (early_inline_small_functions): Likewise.
955 (early_inliner): Likewise.
956 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
957 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
958 * ipa-pure-const.c (malloc_candidate_p): Likewise.
959 * ipa-split.c (execute_split_functions): Likewise.
960 * symbol-summary.h: Likewise.
961 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
963 2018-06-19 Richard Biener <rguenther@suse.de>
965 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
966 (vectorize_loops): ... here. Fix dbgcnt handling.
967 (try_vectorize_loop): Wrap try_vectorize_loop_1.
969 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
972 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
973 ieee128 argument takes up only one (vector) register, not two (floating
976 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
978 * gimplify.c (gimplify_init_constructor): Really never clear for an
979 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
981 2018-06-19 Richard Biener <rguenther@suse.de>
983 PR tree-optimization/86179
984 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
985 after failed recognition.
987 2018-06-18 Martin Sebor <msebor@redhat.com>
990 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
991 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
992 Handle integer subtraction.
993 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
994 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
996 2018-06-18 David Malcolm <dmalcolm@redhat.com>
998 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
999 param from rtx to rtx_insn *.
1000 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
1002 (frv_ifcvt_modify_insn): Likwise.
1003 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
1004 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
1005 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
1006 as_a <rtx_insn *> cast to local "unprotected_region" once
1007 it's been established that it's not NULL or pc_rtx.
1008 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
1009 param "sethi" from rtx to rtx_insn *.
1010 (nds32_group_float_insns): Likewise for param "insn".
1011 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
1013 (vax_output_int_subtract): Likewise.
1014 * config/vax/vax.c (vax_output_int_add): Likewise for param
1016 (vax_output_int_subtract): Likewise.
1017 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
1018 (emit_pattern_after): Likewise for param "after".
1019 (emit_insn_after): Likewise.
1020 (emit_jump_insn_after): Likewise.
1021 (emit_call_insn_after): Likewise.
1022 (emit_debug_insn_after): Likewise.
1023 (emit_pattern_before): Likewise for param "before".
1024 (emit_insn_before): Likewise.
1025 (emit_jump_insn_before): Likewise.
1026 * final.c (get_insn_template): Likewise for param "insn", removing
1028 * output.h (get_insn_template): Likewise for 2nd param.
1029 * rtl.h (emit_insn_before): Likewise.
1030 (emit_jump_insn_before): Likewise.
1031 (emit_debug_insn_before_noloc): Likewise.
1032 (emit_insn_after): Likewise.
1033 (emit_jump_insn_after): Likewise.
1034 (emit_call_insn_after): Likewise.
1035 (emit_debug_insn_after): Likewise.
1036 (set_insn_deleted): Likewise for param.
1038 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
1041 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
1042 floating point modes, so that IFmode is numerically greater than
1043 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
1044 to declare the ordering. This prevents IFmode from being
1045 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
1046 machine. Include rs6000-modes.h to share the fractional values
1047 between genmodes* and the rest of the compiler.
1051 * config/rs6000/rs6000-modes.h: New file.
1052 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
1053 meaning of rs6000_long_double_size so that 126..128 selects an
1054 appropriate 128-bit floating point type.
1055 (rs6000_option_override_internal): Likewise.
1056 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
1057 (TARGET_LONG_DOUBLE_128): Change the meaning of
1058 rs6000_long_double_size so that 126..128 selects an appropriate
1059 128-bit floating point type.
1060 (LONG_DOUBLE_TYPE_SIZE): Update comment.
1061 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
1062 source and destination to match the standard usage.
1063 (truncifkf2): Likewise.
1064 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
1065 ISA 2.07 to use an explicit clobber, instead of passing in a
1067 (copysign<mode>3_soft): Likewise.
1069 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1071 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1072 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
1073 (vect_slp_analyze_instance_dependence): Likewise.
1074 (vect_enhance_data_refs_alignment): Likewise.
1075 (vect_analyze_data_refs_alignment): Likewise.
1076 (vect_slp_analyze_and_verify_instance_alignment
1077 (vect_analyze_data_ref_accesses): Likewise.
1078 (vect_prune_runtime_alias_test_list): Likewise.
1079 (vect_analyze_data_refs): Likewise.
1080 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
1081 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
1082 (vect_analyze_scalar_cycles_1): Likewise.
1083 (vect_get_loop_niters): Likewise.
1084 (vect_analyze_loop_form_1): Likewise.
1085 (vect_update_vf_for_slp): Likewise.
1086 (vect_analyze_loop_operations): Likewise.
1087 (vect_analyze_loop): Likewise.
1088 (vectorizable_induction): Likewise.
1089 (vect_transform_loop): Likewise.
1090 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
1091 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1092 (vect_make_slp_decision): Likewise.
1093 (vect_detect_hybrid_slp): Likewise.
1094 (vect_slp_analyze_operations): Likewise.
1095 (vect_slp_bb): Likewise.
1096 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
1097 (vectorizable_bswap): Likewise.
1098 (vectorizable_call): Likewise.
1099 (vectorizable_simd_clone_call): Likewise.
1100 (vectorizable_conversion): Likewise.
1101 (vectorizable_assignment): Likewise.
1102 (vectorizable_shift): Likewise.
1103 (vectorizable_operation): Likewise.
1104 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
1106 2018-06-18 Martin Sebor <msebor@redhat.com>
1108 PR tree-optimization/81384
1109 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
1110 * builtins.c (expand_builtin_strnlen): New function.
1111 (expand_builtin): Call it.
1112 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
1113 * builtins.def (BUILT_IN_STRNLEN): New.
1114 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
1115 Warn for bounds in excess of maximum object size.
1116 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
1117 single-value ranges. Handle strnlen.
1118 (handle_builtin_strlen): Handle strnlen.
1119 (strlen_check_and_optimize_stmt): Same.
1120 * doc/extend.texi (Other Builtins): Document strnlen.
1122 2018-06-18 Maya Rashish <coypu@sdf.org>
1124 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
1125 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
1126 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
1128 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
1130 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
1132 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1134 * tree.c (escaped_string::escape): Replace cast to char * by
1135 const_cast<char *> (unescaped).
1137 2018-06-18 Nick Clifton <nickc@redhat.com>
1140 * tree.c (escaped_string): New class. Converts an unescaped
1141 string into its escaped equivalent.
1142 (warn_deprecated_use): Use the new class to convert the
1143 deprecation message, if present.
1144 (test_escaped_strings): New self test.
1145 (test_c_tests): Add test_escaped_strings.
1146 * doc/extend.texi (deprecated): Add a note that the
1147 deprecation message is affected by the -fmessage-length
1148 option, and that control characters will be escaped.
1149 (#pragma GCC error): Document this pragma.
1150 (#pragma GCC warning): Likewise.
1151 * doc/invoke.texi (-fmessage-length): Document this option's
1152 effect on the #warning and #error preprocessor directives and
1153 the deprecated attribute.
1155 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
1157 * tree.c (decl_value_expr_lookup): Revert latest change.
1158 (decl_value_expr_insert): Likewise.
1160 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
1162 * gimplify.c (nonlocal_vlas): Delete.
1163 (nonlocal_vla_vars): Likewise.
1164 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
1166 (gimplify_body): Do not create and destroy nonlocal_vlas.
1167 * tree-nested.c: Include diagnostic.h.
1168 (use_pointer_in_frame): Tweak.
1169 (lookup_field_for_decl): Add assertion and declare the transformation.
1170 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
1171 internal error when the reference is in a wrong context. Do not
1172 create a debug decl by default.
1173 (note_nonlocal_block_vlas): Delete.
1174 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
1175 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
1176 create a debug decl by default.
1177 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
1178 call is in a wrong context.
1179 (fixup_vla_decls): New function.
1180 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
1181 debug variables were created.
1182 * tree.c (decl_value_expr_lookup): Add checking assertion.
1183 (decl_value_expr_insert): Likewise.
1185 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1188 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
1189 * tree-scalar-evolution.c (interpret_expr): Likewise.
1190 (expression_expensive_p): Likewise.
1191 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
1192 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
1193 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
1194 (ssa_defined_by_minus_one_stmt_p): New.
1196 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1199 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
1200 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
1201 * dojump.c (do_jump): Likewise.
1202 * expr.c (expand_expr_real_2): Check operand type's sign.
1203 * fold-const.c (const_unop): Handle ABSU_EXPR.
1204 (fold_abs_const): Likewise.
1205 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
1206 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
1207 (strip_sign_op_1): Likesise.
1208 * match.pd: Add new pattern to generate ABSU_EXPR.
1209 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
1210 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1211 * tree-eh.c (operation_could_trap_helper_p): Likewise.
1212 * tree-inline.c (estimate_operator_cost): Likewise.
1213 * tree-pretty-print.c (dump_generic_node): Likewise.
1214 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
1215 * tree.def (ABSU_EXPR): New.
1217 2018-06-16 Jakub Jelinek <jakub@redhat.com>
1220 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
1221 documented as preserved for backward compatibility only.
1222 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
1224 PR rtl-optimization/86108
1225 * bb-reorder.c (create_forwarder_block): Renamed to ...
1226 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
1227 jump from new landing pad to the second part.
1228 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
1231 2018-06-15 Jakub Jelinek <jakub@redhat.com>
1234 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
1235 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
1236 Only call store_expr for halves if the mode is the same.
1239 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
1240 Fix up comment formatting.
1242 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
1244 * typed-splay-tree.h (typed_splay_tree::remove): New function.
1245 (typed_splay_tree::closure,
1246 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
1247 (typed_splay_tree::typed_splay_tree,
1248 typed_splay_tree::operator =): Declared private.
1249 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
1250 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
1251 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
1252 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
1253 typed_splay_tree::splay_tree_splay,
1254 typed_splay_tree::splay_tree_foreach_helper,
1255 typed_splay_tree::splay_tree_insert,
1256 typed_splay_tree::splay_tree_remove,
1257 typed_splay_tree::splay_tree_lookup,
1258 typed_splay_tree::splay_tree_predecessor,
1259 typed_splay_tree::splay_tree_successor,
1260 typed_splay_tree::splay_tree_min,
1261 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
1262 (typed_splay_tree::root, typed_splay_tree::comp,
1263 typed_splay_tree::delete_key,
1264 typed_splay_tree::delete_value): New data members.
1265 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
1266 typed_splay_tree::remove.
1268 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
1270 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
1271 -mginv and -mno-ginv to the assembler.
1272 * config/mips/mips.opt (-mcrc): New option.
1274 * doc/invoke.text (-mcrc): Document.
1277 2018-06-15 Nick Clifton <nickc@redhat.com>
1280 * tree.c (escaped_string): New class. Converts an unescaped
1281 string into its escaped equivalent.
1282 (warn_deprecated_use): Use the new class to convert the
1283 deprecation message, if present.
1284 (test_escaped_strings): New self test.
1285 (test_c_tests): Add test_escaped_strings.
1286 * doc/extend.texi (deprecated): Add a note that the
1287 deprecation message is affected by the -fmessage-length
1288 option, and that control characters will be escaped.
1289 (#pragma GCC error): Document this pragma.
1290 (#pragma GCC warning): Likewise.
1291 * doc/invoke.texi (-fmessage-length): Document this option's
1292 effect on the #warning and #error preprocessor directives and
1293 the deprecated attribute.
1295 2018-06-15 Richard Biener <rguenther@suse.de>
1297 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
1298 here, also noting vector size used.
1299 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
1300 size used in MSG_OPTIMIZED_LOCATIONS dump.
1301 (pass_slp_vectorize::execute): Adjust.
1303 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
1306 * config/arc/arc.c (arc_return_address_register): Fix
1309 2018-06-15 Richard Biener <rguenther@suse.de>
1312 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
1313 leave useless conversion stripping to force_gimple_operand_gsi.
1314 (gimplify_build2): Likewise.
1315 (gimplify_build1): Likewise.
1317 2018-06-15 Richard Biener <rguenther@suse.de>
1320 * tree-cfg.c (move_stmt_op): unshare invariant addresses
1321 before adjusting their block.
1323 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
1325 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
1326 multilibs for *-*-rtems*.
1327 * config/riscv/t-rtems: New file.
1329 2018-06-14 Jakub Jelinek <jakub@redhat.com>
1332 * match.pd ((A +- CST1) +- CST2): Punt if last resort
1333 unsigned_type_for returns NULL.
1336 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
1337 subregs of multi-word pseudos unless the float mode has word size.
1339 2018-06-14 Richard Biener <rguenther@suse.de>
1342 * tree-vect-generic.c (build_word_mode_vector_type): Remove
1343 duplicate and harmful type_hash_canon.
1344 * tree.c (type_hash_canon): Assert we didn't find ourselves.
1346 2018-06-14 Richard Biener <rguenther@suse.de>
1349 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
1352 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
1354 * config/rtems.h (STDINT_LONG32): Define.
1356 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
1357 Prachi Godbole <prachi.godbole@imgtec.com>
1359 * config/mips/mips-cpus.def: Define P6600.
1360 * config/mips/mips-tables.opt: Regenerate.
1361 * config/mips/mips.c (mips_ucbranch_type): New enum.
1362 (mips_rtx_cost_data): Add support for P6600.
1363 (mips_issue_rate): Likewise.
1364 (mips_multipass_dfa_lookahead): Likewise.
1365 (mips_avoid_hazard): Likewise.
1366 (mips_reorg_process_insns): Likewise.
1367 (mips_classify_branch_p6600): New function.
1368 * config/mips/mips.h (TUNE_P6600): New define.
1369 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
1370 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
1371 * config/mips/mips.md: Include p6600.md.
1372 (processor): Add p6600.
1373 * config/mips/p6600.md: New file.
1374 * doc/invoke.texi: Add p6600 to supported architectures.
1376 2018-06-13 Martin Sebor <msebor@redhat.com>
1378 PR tree-optimization/86114
1379 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
1381 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
1383 2018-06-13 Richard Biener <rguenther@suse.de>
1385 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
1386 Properly set vector type of the intermediate stmt.
1387 * tree-vect-stmts.c (vectorizable_operation): The destination
1388 var always has vectype_out type.
1390 2018-06-13 Jeff Law <law@redhat.com>
1392 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
1393 integer 0 for argument to print_rtl_with_bb.
1394 (rl78_reorg): Likewise.
1396 2018-06-13 David Malcolm <dmalcolm@redhat.com>
1398 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
1399 from rtx to rtx_insn *.
1400 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
1402 (add_sched_insns_for_speculation): Likewise for local "target",
1403 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1404 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
1405 from rtx_insn ** to rtx_code_label **.
1406 (reorg_emit_nops): Likewise.
1407 (c6x_reorg): Likewise for local "call_labels".
1408 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
1410 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
1411 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
1412 the loops over LABEL_REFS.
1413 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
1415 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1416 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
1417 (split_branches): Strengthen local "olabel" from rtx to
1418 rtx_insn *, adding a safe_as_a cast.
1419 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
1421 (add_insn_after): Likewise for first two params.
1422 (add_insn_before): Likewise.
1423 (remove_insn): Likewise for param.
1424 (emit_pattern_before_noloc): Likewise for second and third params.
1425 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
1426 (emit_call_insn_before_noloc): Likewise.
1427 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
1429 (emit_barrier_before): Likewise.
1430 (emit_label_before): Strengthen "label" param from "rtx" to
1431 "rtx_code_label *". Strengthen "before" param from "rtx" to
1433 (emit_insn_after_1): Strengthen "after" param from "rtx" to
1435 (emit_pattern_after_noloc): Likewise.
1436 (emit_insn_after_noloc): Likewise.
1437 (emit_jump_insn_after_noloc): Likewise.
1438 (emit_call_insn_after_noloc): Likewise.
1439 (emit_debug_insn_after_noloc): Likewise.
1440 (emit_barrier_after): Likewise.
1441 (emit_label_after): Likewise for both params.
1442 (emit_pattern_after_setloc): Likewise for "after" param. Convert
1443 "loc" param from "int" to "location_t".
1444 (emit_insn_after_setloc): Likewise.
1445 (emit_jump_insn_after_setloc): Likewise.
1446 (emit_call_insn_after_setloc): Likewise.
1447 (emit_debug_insn_after_setloc): Likewise.
1448 (emit_pattern_before_setloc): Likewise for "before" param. Convert
1449 "loc" param from "int" to "location_t".
1450 (emit_pattern_before): Convert NULL_RTX to NULL.
1451 (emit_insn_before_setloc): Convert "loc" param from "int" to
1453 (emit_jump_insn_before_setloc): Likewise.
1454 (emit_call_insn_before_setloc): Likewise.
1455 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
1456 rtx_insn *. Convert "loc" param from "int" to "location_t".
1457 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
1458 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
1459 Convert 3rd param from "int" to "location_t".
1460 (emit_barrier_before, emit_barrier_after, next_real_insn):
1461 Strengthen param from rtx to rtx_insn *.
1462 (emit_label_before): Strengthen 1st param from "rtx" to
1463 "rtx_code_label *". Strengthen 2nd param from "rtx" to
1465 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
1466 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
1467 Strengthen 2nd param from "rtx" to "rtx_insn *".
1468 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
1469 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
1470 Likewise. Convert 3rd param from "int" to "location_t".
1471 (emit_label_after): Strengthen 1st param from "rtx" to
1473 (next_real_insn, remove_insn): Strengthen param from "rtx" to
1475 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
1476 from "rtx" to "rtx_insn *".
1478 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
1480 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
1481 bodies streamed in with -Q.
1482 * dumpfile.c (dump_files): Add lto-stream-out dump file.
1483 * dumpfile.h (tree_dump_index): Add lto_stream_out.
1484 * gimple-streamer-out.c: Include gimple-pretty-print.h
1485 (output_bb): Dump stmts streamed.
1486 * lto-section-out.c: Include print-tree.h
1487 (lto_begin_section): Dump sections created.
1488 (lto_output_decl_index): Dump decl encoded.
1489 * lto-streamer-out.c: Include print-tree.h
1490 (create_output_block): Dump output block created.
1491 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
1492 (output_function): Dump function output.
1493 (output_constructor): Dump constructor streamed.
1494 (write_global_stream): Output indexes encoded.
1495 (produce_asm_for_decls): Dump streams encoded.
1496 * lto-streamer.c (streamer_dump_file): New global var.
1497 * lto-streamer.h (streamer_dump_file): Declare.
1498 * passes.c (ipa_write_summaries): Initialize streamer dump.
1499 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1502 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1505 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1506 offsets for register save directives. Emit a second batch of save
1507 directives, if need be, when the function accesses prior frames.
1509 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1511 * config/arc/fpu.md (fmasf4): Force operand to register.
1512 (fnmasf4): Likewise.
1514 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1516 * config/arc/arc-protos.h (arc_pad_return): Remove.
1517 * config/arc/arc.c (machine_function): Remove force_short_suffix
1519 (arc_print_operand): Adjust printing of '&'.
1520 (arc_verify_short): Remove conditional printing of short suffix.
1521 (arc_final_prescan_insn): Remove reference to size_reason.
1522 (pad_return): New function.
1523 (arc_reorg): Call pad_return.
1524 (arc_pad_return): Remove.
1525 (arc_init_machine_status): Remove reference to force_short_suffix.
1526 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
1527 (attr length): When attribute iscompact is true force to 2
1528 regardless; in the case of maybe check if we want to force the
1529 instruction to have 4 bytes length.
1530 (nopv): Change it to generate 4 byte long nop as well.
1531 (blockage): New pattern.
1532 (simple_return): Remove call to arc_pad_return.
1533 (p_return_i): Likewise.
1535 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1537 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1539 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1541 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1544 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1546 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1547 for ARC700 and ARCv2.
1549 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1552 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1553 operands[2] instead of operands[1].
1556 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1558 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1559 case, check whether the outer register overlaps an unallocatable
1560 register, not just whether it fits the required class.
1562 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1564 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1566 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1567 (memrefs_conflict_p): Likewise.
1568 (init_alias_analysis): Likewise.
1569 * cfgexpand.c (expand_debug_expr): Likewise.
1570 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1571 * cse.c (fold_rtx): Likewise.
1572 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1573 * expr.c (emit_block_move_hints): Likewise.
1574 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1575 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1576 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1577 (emit_group_store): Likewise.
1578 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1579 to read the PRE/POST_MODIFY increment.
1580 * calls.c (store_one_arg): Use strip_offset.
1581 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1583 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1585 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1586 (simplify_binary_operation_1): Extend CONST_INT handling to
1588 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1589 than a HOST_WIDE_INT.
1590 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1592 (adjust_mems, add_stores): Update accodingly.
1593 (vt_canonicalize_addr): Track polynomial offsets.
1594 (emit_note_insn_var_location): Likewise.
1595 (vt_add_function_parameter): Likewise.
1596 (vt_initialize): Likewise.
1598 2018-06-12 Jeff Law <law@redhat.com>
1600 * config.gcc (alpha*-*-freebsd*): Remove.
1601 * config/alpha/freebsd.h: Remove.
1603 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1606 * spellcheck-tree.c (levenshtein_distance): Rename to...
1607 (get_edit_distance): ...this, and update for underlying renaming.
1608 * spellcheck-tree.h (levenshtein_distance): Rename to...
1609 (get_edit_distance): ...this.
1610 * spellcheck.c (levenshtein_distance): Rename to...
1611 (get_edit_distance): ...this. Convert from Levenshtein distance
1612 to Damerau-Levenshtein distance by supporting transpositions of
1613 adjacent characters. Rename "v1" to "v_next" and "v0" to
1615 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1616 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1617 update for underlying renaming.
1618 (selftest::levenshtein_distance_unit_test): Rename to...
1619 (selftest::test_get_edit_distance_unit): ...this, and update for
1620 underlying renaming.
1621 (selftest::test_find_closest_string): Add example from PR 69968
1622 where transposition helps
1623 (selftest::test_metric_conditions): Update for renaming.
1624 (selftest::test_metric_conditions): Likewise.
1625 (selftest::spellcheck_c_tests): Likewise.
1626 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1627 (get_edit_distance): ...this.
1628 (best_match::consider): Update for renaming.
1630 2018-06-12 Martin Sebor <msebor@redhat.com>
1632 PR tree-optimization/85259
1633 * builtins.c (compute_objsize): Handle constant offsets.
1634 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1635 true iff a warning has been issued.
1636 * gimple.h (gimple_nonartificial_location): New function.
1637 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1638 gimple_nonartificial_location and handle -Wno-system-headers.
1639 (handle_builtin_stxncpy): Same.
1641 2018-06-12 Martin Sebor <msebor@redhat.com>
1644 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1646 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1648 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1649 BUILTIN_VEC_XST entries for pointer to double and long long.
1651 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1654 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1656 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1659 2018-06-12 Martin Liska <mliska@suse.cz>
1661 * doc/options.texi: Document IntegerRange.
1663 2018-06-12 Martin Liska <mliska@suse.cz>
1665 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1666 * opt-functions.awk: Handle Deprecated flag.
1667 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1669 (read_cmdline_option): Report warning for a deprecated option.
1670 * opts.h (struct cl_option): Add new field cl_deprecated.
1671 (CL_ERR_DEPRECATED): New.
1673 2018-06-12 Martin Liska <mliska@suse.cz>
1675 * doc/options.texi: Document Deprecated option flag.
1677 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1679 * config/arc/arc-arch.h (arc_extras): New enum.
1680 (arc_cpu_t):Add field extra.
1681 (arc_cpu_types): Consider the extras.
1682 * config/arc/arc-cpus.def: Add extras info.
1683 * config/arc/arc-opts.h (processor_type): Consider extra field.
1684 * config/arc/arc.c (arc_override_options): Handle extra field.
1686 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1688 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1689 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1691 * config/arc/arc.c (arc_init): Update pic warning.
1692 (irq_range): Update irq range parsing warnings.
1693 (arc_override_options): Update various warning messages.
1694 (arc_handle_aux_attribute): Likewise.
1696 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1698 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1700 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1702 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1703 compiler messages directives.
1705 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1707 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1708 * config/mips/mips-tables.opt: Regenerate.
1709 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1711 * doc/invoke.texi: Document -march=i6500.
1713 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1715 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1716 (i6400_gpmul): Add cpu_unit.
1717 (i6400_gpdiv): Likewise.
1718 (i6400_msa_add_d): Update reservations.
1719 (i6400_msa_int_add) Likewise.
1720 (i6400_msa_short_logic3) Likewise.
1721 (i6400_msa_short_logic2) Likewise.
1722 (i6400_msa_short_logic) Likewise.
1723 (i6400_msa_move) Likewise.
1724 (i6400_msa_cmp) Likewise.
1725 (i6400_msa_short_float2) Likewise.
1726 (i6400_msa_div_d) Likewise.
1727 (i6400_msa_long_logic1) Likewise.
1728 (i6400_msa_long_logic2) Likewise.
1729 (i6400_msa_mult) Likewise.
1730 (i6400_msa_long_float2) Likewise.
1731 (i6400_msa_long_float4) Likewise.
1732 (i6400_msa_long_float5) Likewise.
1733 (i6400_msa_long_float8) Likewise.
1734 (i6400_fpu_fadd): Include frint type.
1735 (i6400_fpu_store): New define_insn_reservation.
1736 (i6400_fpu_load): Likewise.
1737 (i6400_fpu_move): Likewise.
1738 (i6400_fpu_fcmp): Likewise.
1739 (i6400_fpu_fmadd): Likewise.
1740 (i6400_int_mult): Include imul3nc type and update reservation.
1741 (i6400_int_div): Include idiv3 type and update reservation.
1742 (i6400_int_load): Update to check type not move_type.
1743 (i6400_int_store): Likewise.
1744 (i6400_int_prefetch): Set zero latency.
1746 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1748 * gcc.c: Document new %@{...} sequence.
1749 (LINK_COMMAND_SPEC): Use it for the -L switches.
1750 (cpp_unique_options): Use it for the -I switches.
1751 (at_file_argbuf): New global variable.
1752 (in_at_file): Likewise.
1753 (alloc_args): Create at_file_argbuf.
1754 (clear_args): Truncate at_file_argbuf.
1755 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1756 (open_at_file): New function.
1757 (close_at_file): Likewise.
1758 (create_at_file): Delete.
1759 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1762 (validate_switches_from_spec): Deal with %@{...} sequence.
1763 (validate_switches): Likewise.
1764 (driver::finalize): Call clear_args.
1766 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1768 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1770 2018-06-11 Martin Sebor <msebor@redhat.com>
1772 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1773 (Wno-class-memaccess): Add @opindex.
1774 (Wno-templates, Wno-multiple-inheritance): Same.
1775 (Wno-virtual-inheritance, Wno-namespaces): Same.
1776 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1777 (Wno-format-overflow, Wno-format-truncation): Same.
1778 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1779 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1780 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1781 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1782 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1783 (Wno-misspelled-isr): Same.
1785 2018-06-11 Martin Sebor <msebor@redhat.com>
1787 * PR tree-optimization/86083
1788 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1790 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1792 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1794 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1797 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1798 on the correct operand.
1799 (*movdi_internal64): Ditto.
1801 2018-06-11 Martin Liska <mliska@suse.cz>
1803 PR tree-optimization/86089
1804 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1806 2018-06-11 Julia Koval <julia.koval@intel.com>
1808 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1809 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1810 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1812 2018-06-11 Olivier Hainque <hainque@adacore.com>
1814 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1815 for Ada with strict dwarf2.
1817 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1820 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1823 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1825 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1827 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1829 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1832 2018-06-08 Martin Liska <mliska@suse.cz>
1834 * tree-cfg.h (debug_function): Fix argument type to match
1837 2018-06-08 Martin Liska <mliska@suse.cz>
1839 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1840 Remove usage of MPX-related (and removed) fields.
1841 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1843 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1845 * cfg.c (debug): Use TDF_NONE rather than 0.
1846 * cfghooks.c (debug): Likewise.
1847 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1848 (struct dump_option_value_info): Convert to...
1849 (struct kv_pair): ...this template type.
1850 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1852 (optinfo_verbosity_options): Likewise.
1853 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1855 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1856 than int for "optgroup_flags" param.
1857 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1859 (dump_dec): Likewise.
1860 (dump_finish): Use TDF_NONE rather than 0.
1861 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1862 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1863 than 0. Update for change to option_ptr.
1864 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1865 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1866 0. Update for changes to optinfo_verbosity_options and
1868 (opt_info_switch_p): Convert optgroup_flags from int to
1870 (dump_basic_block): Use dump_flags_t rather than int
1871 for "dump_kind" param.
1872 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1873 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1874 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1875 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1876 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1877 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1878 TDF_NONE): Convert from macros to...
1879 (enum dump_flag): ...this new enum.
1880 (dump_flags_t): Update to use enum.
1881 (operator|, operator&, operator~, operator|=, operator&=):
1882 Implement for dump_flags_t.
1883 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1884 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1885 Convert from macros to...
1886 (enum optgroup_flag): ...this new enum.
1887 (optgroup_flags_t): New typedef.
1888 (operator|, operator|=): Implement for optgroup_flags_t.
1889 (struct dump_file_info): Convert field "alt_flags" to
1890 dump_flags_t. Convert field "optgroup_flags" to
1892 (dump_basic_block): Use dump_flags_t rather than int for param.
1893 (dump_generic_expr_loc): Likewise.
1894 (dump_dec): Likewise.
1895 (dump_register): Convert param "optgroup_flags" to
1897 (opt_info_enable_passes): Likewise.
1898 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1900 * gimple-pretty-print.c (debug): Likewise.
1901 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1902 (merged_store_group::apply_stores): Likewise.
1903 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1904 * gimple.c (verify_gimple_pp): Likewise.
1905 * graphite-poly.c (print_pbb_body): Likewise.
1906 * passes.c (pass_manager::register_one_dump_file): Convert
1907 local "optgroup_flags" to optgroup_flags_t.
1908 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1910 (debug_body): Likewise.
1911 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1912 to optgroup_flags_t.
1913 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1915 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1916 (convert_mult_to_fma): Likewise.
1917 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1918 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1919 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1920 "dump_kind" to dump_flags_t.
1922 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1924 * config/rs6000/rs6000.c (min, max): Delete.
1926 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1928 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1929 -mabi=spe and -mabi=no-spe.
1931 2018-06-08 Martin Liska <mliska@suse.cz>
1933 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1934 where we expect an existing summary.
1936 2018-06-08 Martin Liska <mliska@suse.cz>
1938 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1939 * ipa-inline.h (estimate_edge_growth): Likewise.
1941 2018-06-08 Martin Liska <mliska@suse.cz>
1943 * cgraph.c (function_version_hasher::hash): Use
1944 cgraph_node::get_uid ().
1945 (function_version_hasher::equal):
1946 * cgraph.h (cgraph_node::get_uid): New method.
1947 * ipa-inline.c (update_caller_keys): Use
1948 cgraph_node::get_uid ().
1949 (update_callee_keys): Likewise.
1950 * ipa-utils.c (searchc): Likewise.
1951 (ipa_reduced_postorder): Likewise.
1952 * lto-cgraph.c (input_node): Likewise.
1953 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1954 * symbol-summary.h (symtab_insertion): Likewise.
1955 (symtab_removal): Likewise.
1956 (symtab_duplication): Likewise.
1957 * tree-pretty-print.c (dump_function_header): Likewise.
1958 * tree-sra.c (convert_callers_for_node): Likewise.
1960 2018-06-08 Martin Liska <mliska@suse.cz>
1962 * cgraph.c (symbol_table::create_edge): Always assign a new
1964 (symbol_table::free_edge): Do not recycle numbers.
1965 * cgraph.h (cgraph_edge::get): New method.
1966 * symbol-summary.h (symtab_removal): Use it.
1967 (symtab_duplication): Likewise.
1968 (call_summary::hashable_uid): Remove.
1970 2018-06-08 Martin Liska <mliska@suse.cz>
1972 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1973 (initialize_growth_caches): Remove.
1974 (free_growth_caches): Likewise.
1975 (do_estimate_edge_time): Use edge_growth_cache.
1976 (do_estimate_edge_size): Likewise.
1977 (do_estimate_edge_hints): Likewise.
1978 * ipa-inline.c (reset_edge_caches): Likewise.
1979 (recursive_inlining): Likewise.
1980 (inline_small_functions): Likewise.
1981 * ipa-inline.h (initialize_growth_caches): Remove.
1982 (estimate_edge_size): Likewise.
1983 (estimate_edge_time): Likewise.
1984 (estimate_edge_hints): Likewise.
1985 (reset_edge_growth_cache): Likewise.
1986 * symbol-summary.h (call_summary::remove): New method.
1988 2018-06-08 Martin Liska <mliska@suse.cz>
1990 * ipa-cp.c (class edge_clone_summary): New summary.
1991 (grow_edge_clone_vectors): Remove.
1992 (ipcp_edge_duplication_hook): Remove.
1993 (class edge_clone_summary_t): New call_summary class.
1994 (ipcp_edge_removal_hook): Remove.
1995 (edge_clone_summary_t::duplicate): New function.
1996 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1997 (create_specialized_node): Likewise.
1998 (ipcp_driver): Initialize edge_clone_summaries and do not
2001 2018-06-08 Martin Liska <mliska@suse.cz>
2003 * symbol-summary.h (get): New function.
2004 (call_summary::m_initialize_when_cloning): New class member.
2006 2018-06-08 Martin Liska <mliska@suse.cz>
2008 * cgraph.c (cgraph_node::remove): Do not recycle uid.
2009 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
2010 (symbol_table::allocate_cgraph_symbol): Do not set uid.
2011 * passes.c (uid_hash_t): Record removed_nodes by their uids.
2012 (remove_cgraph_node_from_order): Use the removed_nodes set.
2013 (do_per_function_toporder): Likwise.
2014 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
2015 instead of summary_uid.
2016 (symtab_removal): Likewise.
2017 (symtab_duplication): Likewise.
2019 2018-06-08 Martin Liska <mliska@suse.cz>
2021 * ipa-cp.c (ipcp_store_bits_results): Use
2022 ipcp_transformation_sum.
2023 (ipcp_store_vr_results): Likewise.
2024 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
2026 (ipcp_transformation_initialize): ... this.
2027 (ipa_set_node_agg_value_chain):
2028 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
2029 (write_ipcp_transformation_info): Likewise.
2030 (read_ipcp_transformation_info): Likewise.
2031 (ipcp_update_bits): Likewise.
2032 (ipcp_update_vr): Likewise.
2033 (ipcp_transform_function): Likewise.
2034 * ipa-prop.h: Rename ipcp_transformation_summary to
2035 ipcp_transformation.
2036 (class ipcp_transformation_t): New function summary.
2037 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
2038 (ipa_get_agg_replacements_for_node): Likewise.
2040 2018-06-08 Martin Liska <mliska@suse.cz>
2042 * ipa-pure-const.c (struct funct_state_d): Do it class instead
2044 (class funct_state_summary_t): New function_summary class.
2045 (has_function_state): Remove.
2046 (get_function_state): Likewise.
2047 (set_function_state): Likewise.
2048 (add_new_function): Likewise.
2049 (funct_state_summary_t::insert): New function.
2050 (duplicate_node_data): Remove.
2051 (remove_node_data): Remove.
2052 (funct_state_summary_t::duplicate): New function.
2053 (register_hooks): Create new funct_state_summaries.
2054 (pure_const_generate_summary): Use it.
2055 (pure_const_write_summary): Likewise.
2056 (pure_const_read_summary): Likewise.
2057 (propagate_pure_const): Likewise.
2058 (propagate_nothrow): Likewise.
2059 (dump_malloc_lattice): Likewise.
2060 (propagate_malloc): Likewise.
2061 (execute): Do not register hooks, just remove summary
2063 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
2066 2018-06-08 Martin Liska <mliska@suse.cz>
2068 * ipa-reference.c (remove_node_data): Remove.
2069 (duplicate_node_data): Likewise.
2070 (class ipa_ref_var_info_summary_t): New class.
2071 (class ipa_ref_opt_summary_t): Likewise.
2072 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
2073 (get_reference_optimization_summary): Use
2074 ipa_ref_opt_sum_summaries.
2075 (set_reference_vars_info): Remove.
2076 (set_reference_optimization_summary): Likewise.
2077 (ipa_init): Create summaries.
2078 (init_function_info): Use function summary.
2079 (ipa_ref_opt_summary_t::duplicate): New function.
2080 (ipa_ref_opt_summary_t::remove): New function.
2081 (get_read_write_all_from_node): Fix GNU coding style.
2082 (propagate): Use function summary.
2083 (write_node_summary_p): Fix GNU coding style.
2084 (stream_out_bitmap): Likewise.
2085 (ipa_reference_read_optimization_summary): Use function summary.
2086 (ipa_reference_c_finalize): Do not release hooks.
2088 2018-06-08 Martin Liska <mliska@suse.cz>
2090 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
2091 (analyze_function_body): Extract multiple calls of get_create.
2092 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2093 * ipa-inline.c (recursive_inlining): Use ::get method.
2094 * ipa-inline.h (estimate_edge_growth): Likewise.
2096 2018-06-08 Martin Liska <mliska@suse.cz>
2098 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
2100 (hsa_function_summary::hsa_function_summary): Use the new enum
2102 (hsa_gpu_implementation_p): Use hsa_summaries::get.
2103 * hsa-gen.c (hsa_get_host_function): Likewise.
2104 (get_brig_function_name): Likewise.
2105 * ipa-hsa.c (process_hsa_functions): Likewise.
2106 (ipa_hsa_write_summary): Likewise.
2107 * symbol-summary.h (symtab_duplication): Use ::get function/
2108 (get): New function.
2110 2018-06-08 Martin Liska <mliska@suse.cz>
2112 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
2114 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
2115 (hsa_register_kernel): Likewise.
2116 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
2117 * hsa-gen.c (hsa_get_host_function): Likewise.
2118 (get_brig_function_name): Likewise.
2119 (generate_hsa): Likewise.
2120 (pass_gen_hsail::execute): Likewise.
2121 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
2122 (devirtualization_time_bonus): Likewise.
2123 (ipcp_propagate_stage): Likewise.
2124 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2125 (edge_set_predicate): Likewise.
2126 (evaluate_conditions_for_known_args): Likewise.
2127 (evaluate_properties_for_edge): Likewise.
2128 (ipa_fn_summary::reset): Likewise.
2129 (ipa_fn_summary_t::duplicate): Likewise.
2130 (dump_ipa_call_summary): Likewise.
2131 (ipa_dump_fn_summary): Likewise.
2132 (analyze_function_body): Likewise.
2133 (compute_fn_summary): Likewise.
2134 (estimate_edge_devirt_benefit): Likewise.
2135 (estimate_edge_size_and_time): Likewise.
2136 (estimate_calls_size_and_time): Likewise.
2137 (estimate_node_size_and_time): Likewise.
2138 (inline_update_callee_summaries): Likewise.
2139 (remap_edge_change_prob): Likewise.
2140 (remap_edge_summaries): Likewise.
2141 (ipa_merge_fn_summary_after_inlining): Likewise.
2142 (ipa_update_overall_fn_summary): Likewise.
2143 (read_ipa_call_summary): Likewise.
2144 (inline_read_section): Likewise.
2145 (write_ipa_call_summary): Likewise.
2146 (ipa_fn_summary_write): Likewise.
2147 (ipa_free_fn_summary): Likewise.
2148 * ipa-hsa.c (process_hsa_functions): Likewise.
2149 (ipa_hsa_write_summary): Likewise.
2150 (ipa_hsa_read_section): Likewise.
2151 * ipa-icf.c (sem_function::merge): Likewise.
2152 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2153 (do_estimate_edge_time): Likewise.
2154 (estimate_size_after_inlining): Likewise.
2155 (estimate_growth): Likewise.
2156 (growth_likely_positive): Likewise.
2157 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2158 (inline_call): Likewise.
2159 * ipa-inline.c (caller_growth_limits): Likewise.
2160 (can_inline_edge_p): Likewise.
2161 (can_inline_edge_by_limits_p): Likewise.
2162 (compute_uninlined_call_time): Likewise.
2163 (compute_inlined_call_time): Likewise.
2164 (want_inline_small_function_p): Likewise.
2165 (edge_badness): Likewise.
2166 (update_caller_keys): Likewise.
2167 (update_callee_keys): Likewise.
2168 (recursive_inlining): Likewise.
2169 (inline_small_functions): Likewise.
2170 (inline_to_all_callers_1): Likewise.
2171 (dump_overall_stats): Likewise.
2172 (early_inline_small_functions): Likewise.
2173 (early_inliner): Likewise.
2174 * ipa-inline.h (estimate_edge_growth): Likewise.
2175 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2176 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2177 * ipa-prop.h (IPA_NODE_REF): Likewise.
2178 (IPA_EDGE_REF): Likewise.
2179 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2180 (propagate_malloc): Likewise.
2181 * ipa-split.c (execute_split_functions): Likewise.
2182 * symbol-summary.h: Rename get to get_create.
2184 (get_create): Likewise.
2185 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2187 2018-06-08 Martin Liska <mliska@suse.cz>
2189 * symbol-summary.h (release): Move definition out of class
2191 (symtab_removal): Likewise.
2192 (symtab_duplication): Likewise.
2194 2018-06-08 Martin Liska <mliska@suse.cz>
2196 * symbol-summary.h (function_summary): Move constructor
2197 implementation out of class declaration.
2198 (release): Likewise.
2199 (symtab_insertion): Likewise.
2200 (symtab_removal): Likewise.
2201 (symtab_duplication): Likewise.
2204 2018-06-08 Martin Liska <mliska@suse.cz>
2206 * Makefile.in: Remove support for MPX (macros, related functions,
2207 fields in cgraph_node, ...).
2208 * builtin-types.def (BT_BND): Likewise.
2209 (BT_FN_BND_CONST_PTR): Likewise.
2210 (BT_FN_CONST_PTR_BND): Likewise.
2211 (BT_FN_VOID_PTR_BND): Likewise.
2212 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
2213 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
2214 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
2215 (expand_builtin_mempcpy_with_bounds): Likewise.
2216 (expand_builtin_memset_with_bounds): Likewise.
2217 (expand_builtin_memset_args): Likewise.
2218 (std_expand_builtin_va_start): Likewise.
2219 (expand_builtin): Likewise.
2220 (expand_builtin_with_bounds): Likewise.
2221 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
2222 (DEF_LIB_BUILTIN_CHKP): Likewise.
2223 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
2224 (DEF_CHKP_BUILTIN): Likewise.
2225 (BUILT_IN_MEMCPY): Likewise.
2226 (BUILT_IN_MEMMOVE): Likewise.
2227 (BUILT_IN_MEMPCPY): Likewise.
2228 (BUILT_IN_MEMSET): Likewise.
2229 (BUILT_IN_STPCPY): Likewise.
2230 (BUILT_IN_STRCAT): Likewise.
2231 (BUILT_IN_STRCHR): Likewise.
2232 (BUILT_IN_STRCPY): Likewise.
2233 (BUILT_IN_STRLEN): Likewise.
2234 (BUILT_IN_MEMCPY_CHK): Likewise.
2235 (BUILT_IN_MEMMOVE_CHK): Likewise.
2236 (BUILT_IN_MEMPCPY_CHK): Likewise.
2237 (BUILT_IN_MEMSET_CHK): Likewise.
2238 (BUILT_IN_STPCPY_CHK): Likewise.
2239 (BUILT_IN_STRCAT_CHK): Likewise.
2240 (BUILT_IN_STRCPY_CHK): Likewise.
2241 * calls.c (store_bounds): Likewise.
2242 (emit_call_1): Likewise.
2243 (special_function_p): Likewise.
2244 (maybe_warn_nonstring_arg): Likewise.
2245 (initialize_argument_information): Likewise.
2246 (finalize_must_preallocate): Likewise.
2247 (compute_argument_addresses): Likewise.
2248 (expand_call): Likewise.
2249 * cfgexpand.c (expand_call_stmt): Likewise.
2250 (expand_return): Likewise.
2251 (expand_gimple_stmt_1): Likewise.
2252 (pass_expand::execute): Likewise.
2253 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2254 (cgraph_node::remove): Likewise.
2255 (cgraph_node::dump): Likewise.
2256 (cgraph_node::verify_node): Likewise.
2257 * cgraph.h (chkp_function_instrumented_p): Likewise.
2258 (symtab_node::get_alias_target): Likewise.
2259 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
2260 (cgraph_local_p): Likewise.
2261 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
2262 (cgraph_edge::rebuild_references): Likewise.
2263 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
2264 (walk_polymorphic_call_targets): Likewise.
2265 (cgraph_node::expand_thunk): Likewise.
2266 (symbol_table::output_weakrefs): Likewise.
2267 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
2268 (ix86_handle_option): Likewise.
2269 * config/i386/constraints.md: Likewise.
2270 * config/i386/i386-builtin-types.def (BND): Likewise.
2274 * config/i386/i386-builtin.def (BDESC_END): Likewise.
2275 (BDESC_FIRST): Likewise.
2277 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
2278 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
2279 * config/i386/i386.c (enum reg_class): Likewise.
2280 (ix86_target_string): Likewise.
2281 (ix86_option_override_internal): Likewise.
2282 (ix86_conditional_register_usage): Likewise.
2283 (ix86_valid_target_attribute_inner_p): Likewise.
2284 (ix86_set_indirect_branch_type): Likewise.
2285 (ix86_set_current_function): Likewise.
2286 (ix86_function_arg_regno_p): Likewise.
2287 (init_cumulative_args): Likewise.
2288 (ix86_function_arg_advance): Likewise.
2289 (ix86_function_arg): Likewise.
2290 (ix86_pass_by_reference): Likewise.
2291 (ix86_function_value_regno_p): Likewise.
2292 (ix86_function_value_1): Likewise.
2293 (ix86_function_value_bounds): Likewise.
2294 (ix86_return_in_memory): Likewise.
2295 (ix86_setup_incoming_vararg_bounds): Likewise.
2296 (ix86_va_start): Likewise.
2297 (indirect_thunk_need_prefix): Likewise.
2298 (print_reg): Likewise.
2299 (ix86_print_operand): Likewise.
2300 (ix86_expand_call): Likewise.
2301 (ix86_output_function_return): Likewise.
2302 (reg_encoded_number): Likewise.
2303 (BDESC_VERIFYS): Likewise.
2304 (ix86_init_mpx_builtins): Likewise.
2305 (ix86_init_builtins): Likewise.
2306 (ix86_emit_cmove): Likewise.
2307 (ix86_emit_move_max): Likewise.
2308 (ix86_expand_builtin): Likewise.
2309 (ix86_builtin_mpx_function): Likewise.
2310 (ix86_get_arg_address_for_bt): Likewise.
2311 (ix86_load_bounds): Likewise.
2312 (ix86_store_bounds): Likewise.
2313 (ix86_load_returned_bounds): Likewise.
2314 (ix86_store_returned_bounds): Likewise.
2315 (ix86_class_likely_spilled_p): Likewise.
2316 (ix86_hard_regno_mode_ok): Likewise.
2317 (x86_order_regs_for_local_alloc): Likewise.
2318 (ix86_mitigate_rop): Likewise.
2319 (ix86_bnd_prefixed_insn_p): Likewise.
2320 (ix86_mpx_bound_mode): Likewise.
2321 (ix86_make_bounds_constant): Likewise.
2322 (ix86_initialize_bounds): Likewise.
2323 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
2324 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
2325 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
2326 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
2327 (TARGET_CHKP_BOUND_MODE): Likewise.
2328 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
2329 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
2330 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
2331 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
2332 * config/i386/i386.h (TARGET_MPX): Likewise.
2333 (TARGET_MPX_P): Likewise.
2334 (VALID_BND_REG_MODE): Likewise.
2335 (FIRST_BND_REG): Likewise.
2336 (LAST_BND_REG): Likewise.
2337 (enum reg_class): Likewise.
2338 (BND_REG_P): Likewise.
2339 (BND_REGNO_P): Likewise.
2340 (BNDmode): Likewise.
2341 (ADJUST_INSN_LENGTH): Likewise.
2342 * config/i386/i386.md: Likewise.
2343 * config/i386/i386.opt: Likewise.
2344 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
2345 (defined): Likewise.
2346 (LINK_MPX): Likewise.
2347 (MPX_SPEC): Likewise.
2348 (LIBMPX_SPEC): Likewise.
2349 (LIBMPXWRAPPERS_SPEC): Likewise.
2350 (CHKP_SPEC): Likewise.
2351 * config/i386/predicates.md: Likewise.
2352 * dbxout.c (dbxout_type): Likewise.
2353 * doc/extend.texi: Likewise.
2354 * doc/invoke.texi: Likewise.
2355 * doc/md.texi: Likewise.
2356 * doc/tm.texi: Likewise.
2357 * doc/tm.texi.in: Likewise.
2358 * dwarf2out.c (is_base_type): Likewise.
2359 (gen_formal_types_die): Likewise.
2360 (gen_subprogram_die): Likewise.
2361 (gen_type_die_with_usage): Likewise.
2362 (gen_decl_die): Likewise.
2363 (dwarf2out_late_global_decl): Likewise.
2364 * expr.c (expand_assignment): Likewise.
2365 (emit_storent_insn): Likewise.
2366 (store_expr_with_bounds): Likewise.
2367 (store_expr): Likewise.
2368 (expand_expr_real_1): Likewise.
2369 * expr.h (store_expr_with_bounds): Likewise.
2370 * function.c (use_register_for_decl): Likewise.
2371 (struct bounds_parm_data): Likewise.
2372 (assign_parms_augmented_arg_list): Likewise.
2373 (assign_parm_find_entry_rtl): Likewise.
2374 (assign_parm_is_stack_parm): Likewise.
2375 (assign_parm_load_bounds): Likewise.
2376 (assign_bounds): Likewise.
2377 (assign_parms): Likewise.
2378 (expand_function_start): Likewise.
2379 * gcc.c (CHKP_SPEC): Likewise.
2380 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
2381 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
2382 (wrestrict_dom_walker::check_call): Likewise.
2383 * gimple.c (gimple_build_call_from_tree): Likewise.
2384 * gimple.h (enum gf_mask): Likewise.
2385 (gimple_call_with_bounds_p): Likewise.
2386 (gimple_call_set_with_bounds): Likewise.
2387 * gimplify.c (gimplify_init_constructor): Likewise.
2388 * ipa-cp.c (initialize_node_lattices): Likewise.
2389 (propagate_constants_across_call): Likewise.
2390 (find_more_scalar_values_for_callers_subset): Likewise.
2391 * ipa-hsa.c (process_hsa_functions): Likewise.
2392 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
2393 * ipa-icf.c (sem_function::merge): Likewise.
2394 * ipa-inline.c (early_inliner): Likewise.
2395 * ipa-pure-const.c (warn_function_noreturn): Likewise.
2396 (warn_function_cold): Likewise.
2397 (propagate_pure_const): Likewise.
2398 * ipa-ref.h (enum GTY): Likewise.
2399 * ipa-split.c (find_retbnd): Likewise.
2400 (consider_split): Likewise.
2401 (split_function): Likewise.
2402 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2403 * ipa.c (walk_polymorphic_call_targets): Likewise.
2404 (symbol_table::remove_unreachable_nodes): Likewise.
2405 (process_references): Likewise.
2406 (cgraph_build_static_cdtor_1): Likewise.
2407 * lto-cgraph.c (lto_output_node): Likewise.
2408 (output_refs): Likewise.
2409 (compute_ltrans_boundary): Likewise.
2410 (input_overwrite_node): Likewise.
2411 (input_node): Likewise.
2412 (input_cgraph_1): Likewise.
2413 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
2414 * passes.c (pass_manager::execute_early_local_passes): Likewise.
2415 (class pass_chkp_instrumentation_passes): Likewise.
2416 (make_pass_chkp_instrumentation_passes): Likewise.
2417 * passes.def: Likewise.
2418 * rtl.h (struct GTY): Likewise.
2419 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
2420 * stor-layout.c (layout_type): Likewise.
2421 * symtab.c: Likewise.
2422 * target.def: Likewise.
2423 * targhooks.c (default_chkp_bound_type): Likewise.
2424 (default_chkp_bound_mode): Likewise.
2425 (default_builtin_chkp_function): Likewise.
2426 (default_chkp_function_value_bounds): Likewise.
2427 (default_chkp_make_bounds_constant): Likewise.
2428 (default_chkp_initialize_bounds): Likewise.
2429 * targhooks.h (default_chkp_bound_type): Likewise.
2430 (default_chkp_bound_mode): Likewise.
2431 (default_builtin_chkp_function): Likewise.
2432 (default_chkp_function_value_bounds): Likewise.
2433 (default_chkp_make_bounds_constant): Likewise.
2434 (default_chkp_initialize_bounds): Likewise.
2435 * toplev.c (compile_file): Likewise.
2436 (process_options): Likewise.
2437 * tree-core.h (DEF_BUILTIN): Likewise.
2438 (DEF_BUILTIN_CHKP): Likewise.
2439 * tree-inline.c (declare_return_variable): Likewise.
2440 (remap_gimple_stmt): Likewise.
2441 (copy_bb): Likewise.
2442 (initialize_inlined_parameters): Likewise.
2443 (expand_call_inline): Likewise.
2444 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
2445 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
2446 (make_pass_ipa_chkp_produce_thunks): Likewise.
2447 (make_pass_chkp): Likewise.
2448 (make_pass_chkp_opt): Likewise.
2449 (make_pass_chkp_instrumentation_passes): Likewise.
2450 * tree-pretty-print.c (dump_generic_node): Likewise.
2451 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
2452 * tree-ssa-dce.c (propagate_necessity): Likewise.
2453 (eliminate_unnecessary_stmts): Likewise.
2454 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
2455 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2456 * tree-ssa-sccvn.h: Likewise.
2457 * tree-ssa-strlen.c (get_string_length): Likewise.
2458 (valid_builtin_call): Likewise.
2459 (adjust_last_stmt): Likewise.
2460 (handle_builtin_strchr): Likewise.
2461 (handle_builtin_strcpy): Likewise.
2462 (handle_builtin_stxncpy): Likewise.
2463 (handle_builtin_memcpy): Likewise.
2464 (handle_builtin_strcat): Likewise.
2465 (strlen_check_and_optimize_stmt): Likewise.
2466 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
2467 * tree-streamer-in.c: Likewise.
2468 * tree-streamer.c (record_common_node): Likewise.
2469 * tree.c (tree_code_size): Likewise.
2470 (wide_int_to_tree_1): Likewise.
2471 (type_contains_placeholder_1): Likewise.
2472 (build_common_tree_nodes): Likewise.
2473 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
2474 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
2475 (POINTER_BOUNDS_P): Likewise.
2476 (BOUNDED_TYPE_P): Likewise.
2477 (BOUNDED_P): Likewise.
2478 (CALL_WITH_BOUNDS_P): Likewise.
2479 (pointer_bounds_type_node): Likewise.
2480 * value-prof.c (gimple_ic): Likewise.
2481 * var-tracking.c (vt_add_function_parameters): Likewise.
2482 * varasm.c (make_decl_rtl): Likewise.
2483 (assemble_start_function): Likewise.
2484 (output_constant): Likewise.
2485 (maybe_assemble_visibility): Likewise.
2486 * varpool.c (ctor_for_folding): Likewise.
2487 * chkp-builtins.def: Remove.
2488 * ipa-chkp.c: Remove.
2489 * ipa-chkp.h: Remove.
2490 * rtl-chkp.c: Remove.
2491 * rtl-chkp.h: Remove.
2492 * tree-chkp-opt.c: Remove.
2493 * tree-chkp.c: Remove.
2494 * tree-chkp.h: Remove.
2496 2018-06-07 Carl Love <cel@us.ibm.com>
2498 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
2499 vextract_fp_from_shortl): Add BE support.
2501 2018-06-07 Paul Koning <ni1d@arrl.net>
2503 * compare-elim.c (try_merge_compare): Don't merge compare if
2504 address contains a side effect.
2505 (try_eliminate_compare): Likewise.
2507 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2509 * config.gcc: Support "tremont".
2510 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2511 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2513 * config/i386/i386.c (m_TREMONT): Define.
2514 (processor_target_table): Add "tremont".
2515 (PTA_TREMONT): Define.
2516 (ix86_lea_outperforms): Add TARGET_TREMONT.
2517 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
2518 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
2519 and M_INTEL_GOLDMONT_PLUS.
2520 (fold_builtin_cpu): Add "tremont".
2521 (ix86_add_stmt_cost): Add TARGET_TREMONT.
2522 (ix86_option_override_internal): Add "tremont".
2523 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
2524 (processor_type): Add PROCESSOR_TREMONT.
2525 * config/i386/x86-tune.def: Add m_TREMONT.
2526 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
2528 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2530 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2531 symbol defined for msp430i* devices to be lower case.
2533 gcc/testsuite/gcc.target/msp430/
2534 * msp430i-device-symbol.c: New test.
2535 * msp430f-device-symbol.c: New test.
2536 * msp430.h: New test header file.
2538 2018-06-07 Richard Biener <rguenther@suse.de>
2540 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2541 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2542 Properly wrap signed arithmetic if overflow wraps.
2544 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2546 PR tree-optimization/69615
2547 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2548 of a cast from a same precision integral SSA_NAME in a bb dominated
2549 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2550 cast to utype if rhs2 has already a compatible type.
2552 2018-06-07 Richard Biener <rguenther@suse.de>
2554 PR tree-optimization/85935
2555 * graphite-scop-detection.c (find_params_in_bb): Analyze
2556 condition operands with respect to the correct loop. Assert
2557 the analysis doesn't fail.
2559 2018-06-04 Carl Love <cel@us.ibm.com>
2561 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2562 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2563 as it is slightly cheaper.
2564 (first_match_or_eos_index_<mode>):
2565 Calculate index using natural element order.
2566 (first_match_index_<mode>):
2567 Calculate index using natural element order.
2568 (first_match_or_eos_index_<mode>):
2569 Calculate index using natural order.
2570 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2571 for BE and LE modes.
2572 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2573 P9V_BUILTIN_VCLZLSBB_V16QI.
2574 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2577 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2579 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2580 indentation and line wrap for many prototypes. Add missing
2581 @smallexample directives around block of prototypes for vec_xl and
2584 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2586 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2587 track if we pass or return IEEE 128-bit floating point.
2588 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2589 C++ mangling that is compatible with GCC 8.1.
2590 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2591 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2592 floating point types.
2593 (rs6000_function_arg_advance_1): Likewise.
2594 (rs6000_mangle_type): Optionally generate mangled names that match
2595 what GCC 8.1 generated for IEEE 128-bit floating point types.
2596 (rs6000_globalize_decl_name): If we have an external function that
2597 passes or returns IEEE 128-bit types, generate a weak reference
2598 from the mangled name used in GCC 8.1 to the current mangled
2600 (rs6000_init_builtins): Make __ibm128 use the long double type if
2601 long double is IBM extended double. Make __float128 use the long
2602 double type if long double is IEEE 128-bit.
2605 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2606 macro for __ibm128 built-in functions.
2607 (PACK_IF): Add __ibm128 pack/unpack functions.
2608 (UNPACK_IF): Likewise.
2609 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2610 enable long double built-in functions if long double is IEEE
2611 128-bit floating point.
2612 (rs6000_invalid_builtin): Update long double built-in function
2614 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2615 functions, adjust the built-in function to use the long double
2616 built-in function if __ibm128 and long double are the same type.
2617 * doc/extend.texi (PowerPC builtins): Update documention for
2618 __builtin_{,un}pack_longdouble. Add documentation for
2619 __builtin_{,un}pack_ibm128.
2621 2018-06-06 Jim Wilson <jimw@sifive.com>
2623 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2624 (struct machine_function): New field interrupt_mode.
2625 (riscv_handle_type_attribute): New function. Add forward declaration.
2626 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2627 (riscv_expand_epilogue): Check interrupt_mode field.
2628 (riscv_set_current_function): Check interrupt attribute args and
2629 set interrupt_mode field.
2630 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2631 (riscv_sret, riscv_uret): New.
2632 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2633 new arguments to interrupt attribute.
2635 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2638 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2639 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2641 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2643 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2644 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2645 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2646 several redundant entries.
2648 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2650 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2651 type from "rtx" to "rtx_insn *".
2652 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2653 for local "call_insn", removing cast.
2654 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2656 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2658 PR tree-optimization/86066
2659 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2660 for BIT_INSERT_EXPR stores.
2662 2018-06-06 Richard Biener <rguenther@suse.de>
2664 PR tree-optimization/86062
2665 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2666 component refs ontop
2667 of to be offsetted base.
2669 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2671 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2672 to be static and remove check on interrupt attribute name.
2674 gcc/testsuite/gcc.target/msp430/
2675 * function-attributes-4.c: New test.
2676 * static-interrupts.c: New test.
2678 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2680 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2681 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2683 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2686 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2688 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2689 Remove second argument, change how error is called.
2690 (aarch64_layout_arg): Remove second argument from
2691 aarch64_err_no_fpadvsimd call.
2692 (aarch64_init_cumulative_args): Ditto.
2693 (aarch64_gimplify_va_arg_expr): Ditto.
2694 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2696 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2698 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2699 (*simple_return_indirect_internal<mode>): Rename from
2700 simple_return_indirect_internal. Use W mode iterator.
2701 (rstorssp): New expander.
2702 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2703 (clrssbsy): New expander.
2704 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2706 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2708 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2710 (cmse_check_pointed_object): Likewise.
2712 2018-06-05 Martin Liska <mliska@suse.cz>
2714 PR gcov-profile/47618
2715 * doc/invoke.texi: Document how -fprofile-dir format
2718 2018-06-05 Richard Biener <rguenther@suse.de>
2720 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2721 removal pretend DOM info isn't available so we do not update
2722 it and only remove edges, not dominated blocks. Actually free
2723 DOM info in case we removed something. Remove unreachable blocks.
2724 (mfb_keep_latches): Work with either DOM info or marked backedges.
2725 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2726 first. Mark backedges if DOM info isn't available.
2727 (Re-)compute DOM info after cleanup_control_flow_pre.
2729 2018-06-05 Richard Biener <rguenther@suse.de>
2731 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2732 (locus_discrim_hasher::hash): Adjust.
2733 (locus_discrim_hasher::equal): Likewise.
2734 (next_discriminator_for_locus): Work on line directly.
2735 (same_line_p): Pass in expanded locus1 as well.
2736 (assign_discriminators): Avoid redundant location expansions.
2738 2018-06-05 Richard Biener <rguenther@suse.de>
2740 PR tree-optimization/86046
2741 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2742 if required after clearing TREE_ADDRESSABLE.
2744 2018-06-05 Richard Biener <rguenther@suse.de>
2746 PR tree-optimization/86047
2747 * tree-ssa-loop.c (for_each_index): Glob handling of all
2748 decls and constants and really handle all of them.
2750 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2753 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2754 qualifier_void_pointer and qualifier_const_void_pointer.
2755 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2756 (arm_init_builtins): Handle the above.
2757 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2758 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2759 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2762 2018-06-05 Martin Liska <mliska@suse.cz>
2764 * auto-profile.c (read_autofdo_file): Do not use
2765 gcov_ctr_summary struct.
2766 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2767 * coverage.c (struct counts_entry): Likewise.
2768 (read_counts_file): Read just single summary entry.
2769 (get_coverage_counts): Use gcov_summary struct.
2770 * coverage.h (get_coverage_counts): Likewise.
2771 * gcov-dump.c (dump_working_sets): Likewise.
2772 (tag_summary): Dump just single summary.
2773 * gcov-io.c (gcov_write_summary): Write just histogram
2775 (gcov_read_summary): Read just single summary.
2776 (compute_working_sets): Use gcov_summary struct.
2777 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2778 of GCOV_COUNTERS_SUMMABLE.
2779 (GCOV_COUNTERS_SUMMABLE): Remove.
2780 (GCOV_FIRST_VALUE_COUNTER): Replace with
2781 GCOV_COUNTER_V_INTERVAL.
2782 (struct gcov_ctr_summary): Remove.
2783 (struct gcov_summary): Directly use fields of former
2785 (compute_working_sets): Use gcov_summary struct.
2786 * gcov.c (read_count_file): Do not use ctrs fields.
2787 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2789 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2791 * profile.c: Likewise.
2792 * profile.h: Likewise.
2794 2018-06-05 Martin Liska <mliska@suse.cz>
2796 PR gcov-profile/84846
2797 * gcov.c (output_lines): Print working directory only
2798 in intermediate format.
2800 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2802 * config/s390/s390-builtin-types.def: Add void function type.
2803 * config/s390/s390-builtins.def: Use the function type for the
2806 2018-06-04 Jim Wilson <jimw@sifive.com>
2808 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2810 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2811 and maybe_eh_return. Change regno to unsigned int. Use new args to
2812 handle EH_RETURN_DATA_REGNO registers properly.
2813 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2814 (riscv_expand_epilogue): Update comment. Change argument name and
2815 type. Update code to use new name and type. Pass new args to
2816 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2818 * config/riscv/riscv.md (NORMAL_RETURN): New.
2819 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2820 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2821 (eh_return): Call gen_eh_return_internal and emit barrier.
2822 (eh_return_internal): Call riscv_expand_epilogue.
2824 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2826 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2827 bit_insertion field and declare can_be_merged_into method.
2828 (merged_store_group::can_be_merged_into): New method.
2829 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2830 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2831 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2833 2018-06-04 Richard Biener <rguenther@suse.de>
2835 PR tree-optimization/85955
2836 * builtins.c (fold_builtin_sincos): Convert pointers to
2837 destination to appropriate type before dereferencing.
2839 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2841 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2843 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2845 * expr.c (expand_expr_real_1): Force the operand into memory if
2846 its TYPE_MODE is BLKmode and if there is no integer mode for
2847 the number of bits being extracted.
2849 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2853 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2854 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2856 2018-06-04 Richard Biener <rguenther@suse.de>
2858 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2859 (cleanup_tree_cfg_noloop): ... single caller. Do
2860 start_recording_case_labels later.
2862 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2864 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2865 to _IMMINTRIN_H_INCLUDED.
2866 * config/i386/pconfigintrin.h: Ditto.
2867 * config/i386/waitpkgintrin.h: Ditto.
2868 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2869 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2870 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2871 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2872 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2873 waitpkgintrin.h and cldemoteintrin.h.
2875 2018-06-04 Richard Biener <rguenther@suse.de>
2877 PR tree-optimization/86038
2878 * tracer.c (find_best_successor): Check probability for
2879 being initialized, bail out if not.
2881 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2884 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2885 of bits to ignore when comparing architectures.
2887 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2889 PR tree-optimization/69615
2890 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2891 maximum or minimum of the type, try to merge it also as if
2892 range1 is + [-, x - 1] or + [x + 1, -].
2895 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2897 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2899 PR tree-optimization/86034
2900 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2901 the unsigned bitfield type in a bit insertion sequence if it does not
2902 have a larger precision than the bitfield size.
2903 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2905 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2907 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2909 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2911 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2912 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2913 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2914 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2916 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2918 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2919 Disable -fdelete-null-pointer-checks for ELF toolchain.
2921 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2922 Kito Cheng <kito.cheng@gmail.com>
2924 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2925 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2926 (nds32*-*-*): Add glibc and uclibc conditions.
2927 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2928 (TARGET_EXCEPT_UNWIND_INFO): Define.
2929 * config/nds32/elf.h: New file.
2930 * config/nds32/linux.h: New file.
2931 * config/nds32/nds32-elf.opt: New file.
2932 * config/nds32/nds32-linux.opt: New file.
2933 * config/nds32/nds32-fp-as-gp.c
2934 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2935 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2937 (nds32_asm_file_end): Ditto.
2938 (nds32_print_operand): Ditto.
2939 (nds32_insert_attributes): Ditto.
2940 (nds32_init_libfuncs): New function.
2941 (TARGET_HAVE_TLS): Define.
2942 (TARGET_INIT_LIBFUNCS): Define.
2943 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2945 (TARGET_ELF): Apply different mcmodel setting.
2946 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2947 been migrated into elf.h and linux.h files.
2948 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2949 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2950 (mcmodel): The content has been migrated into nds32-elf.opt and
2951 nds32-linux.opt files.
2952 * config/nds32/t-elf: New file.
2953 * config/nds32/t-linux: New file.
2955 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2956 Shiva Chen <shiva0217@gmail.com>
2958 * config/nds32/constants.md (unspec_volatile_element): Add
2959 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2960 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2962 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2963 (make_pass_nds32_fp_as_gp): Declare.
2964 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2966 (nds32_asm_function_end_prologue): Remove unused asm output.
2967 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2968 (nds32_asm_file_start): Output necessary fp_as_gp information.
2969 (nds32_option_override): Adjust register usage.
2970 (nds32_expand_prologue): Consider fp_as_gp situation.
2971 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2972 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2975 (simple_return): Ditto.
2976 (omit_fp_begin): Output special directive for fp_as_gp.
2977 (omit_fp_end): Output special directive for fp_as_gp.
2978 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2979 mforbid-fp-as-gp): New options.
2981 2018-06-01 Mark Wielaard <mark@klomp.org>
2983 * dwarf2out.c (dwarf2out_finish): Remove generation of
2984 DW_AT_loclists_base.
2986 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2988 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2989 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2990 (struct merged_store_group): Add bit_insertion field.
2991 (dump_char_array): Use standard hexadecimal format.
2992 (merged_store_group::merged_store_group): Set bit_insertion to false.
2993 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2994 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2995 also print the mask in the dump file.
2996 (pass_store_merging::gate): Minor tweak.
2997 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2998 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2999 stores with INTEGER_CST stores.
3000 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
3001 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
3002 and use it throughout. Generate bit insertion sequences if need be.
3003 (pass_store_merging::process_store): Remove redundant condition.
3004 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
3006 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3008 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3009 the 128-bit floating point types. Fix function comment.
3011 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3013 * config/aarch64/aarch64-simd.md
3014 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
3016 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
3019 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3021 PR tree-optimization/85989
3022 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3024 (backprop::intersect_uses): Check it when deciding whether this
3025 is a backedge reference.
3026 (backprop::process_block): Add each phi to m_visited_phis
3027 after visiting it, then clear it at the end.
3029 2018-06-01 Richard Biener <rguenther@suse.de>
3031 * tree-vectorizer.h (vect_dr_stmt): New function.
3032 (vect_get_load_cost): Adjust.
3033 (vect_get_store_cost): Likewise.
3034 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
3035 Use vect_dr_stmt instead of DR_SMTT.
3036 (vect_record_base_alignments): Likewise.
3037 (vect_calculate_target_alignment): Likewise.
3038 (vect_compute_data_ref_alignment): Likewise and make static.
3039 (vect_update_misalignment_for_peel): Likewise.
3040 (vect_verify_datarefs_alignment): Likewise.
3041 (vector_alignment_reachable_p): Likewise.
3042 (vect_get_data_access_cost): Likewise. Pass down
3043 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
3044 (vect_get_peeling_costs_all_drs): Likewise.
3045 (vect_peeling_hash_get_lowest_cost): Likewise.
3046 (vect_enhance_data_refs_alignment): Likewise.
3047 (vect_find_same_alignment_drs): Likewise.
3048 (vect_analyze_data_refs_alignment): Likewise.
3049 (vect_analyze_group_access_1): Likewise.
3050 (vect_analyze_group_access): Likewise.
3051 (vect_analyze_data_ref_access): Likewise.
3052 (vect_analyze_data_ref_accesses): Likewise.
3053 (vect_vfa_segment_size): Likewise.
3054 (vect_small_gap_p): Likewise.
3055 (vectorizable_with_step_bound_p): Likewise.
3056 (vect_prune_runtime_alias_test_list): Likewise.
3057 (vect_analyze_data_refs): Likewise.
3058 (vect_supportable_dr_alignment): Likewise.
3059 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3060 (vect_gen_prolog_loop_niters): Likewise.
3061 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3062 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
3064 (vect_recog_mask_conversion_pattern): Likewise.
3065 (vect_try_gather_scatter_pattern): Likewise.
3066 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
3067 to vect_get_store_cost.
3068 (vect_get_store_cost): Get stmt_info instead of DR.
3069 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
3070 (vect_get_load_cost): Get stmt_info instead of DR.
3072 2018-06-01 Richard Biener <rguenther@suse.de>
3075 * gimple-fold.c (var_decl_component_p): Also allow offsetted
3076 vars wrapped in MEM_REFs.
3078 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3080 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3081 Fix subreg tests so that we only return a choice between
3082 GENERAL_REGS and FP_REGS if the original classes included both.
3084 2018-06-01 Richard Biener <rguenther@suse.de>
3087 * tree-ssa-structalias.c (get_function_part_constraint):
3088 Handle NULL fi->decl.
3089 (find_func_aliases_for_call): Properly handle indirect
3090 fi from direct call.
3091 (find_func_clobbers): Likewise.
3092 (ipa_pta_execute): Likewise.
3093 (create_variable_info_for): For functions that are ifunc_resolver
3094 resolve to a varinfo that contains the result of the resolver call.
3095 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
3098 2018-05-31 Michael Collison <michael.collison@arm.com>
3100 * config/aarch64/aarch64.md:
3101 (*fix_to_zero_extenddfdi2): New pattern.
3102 * gcc.target/aarch64/fix_extend1.c: New testcase.
3104 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
3108 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
3109 and BUILT_IN_STRNCMP_EQ.
3110 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
3111 BUILT_IN_STRNCMP_EQ.
3112 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
3113 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3114 (gimple_fold_builtin): Likewise.
3115 * tree-ssa-strlen.c (compute_string_length): New function.
3116 (determine_min_obsize): New function.
3117 (handle_builtin_string_cmp): New function to handle calls to
3118 string compare functions.
3119 (strlen_optimize_stmt): Add handling to builtin string compare
3121 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3122 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3123 * tree.c (build_common_builtin_nodes): Add new defines of
3124 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
3126 2018-05-31 Jakub Jelinek <jakub@redhat.com>
3129 * bb-reorder.c (pass_partition_blocks::gate): Return false for
3130 functions with naked attribute.
3132 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
3134 * config/i386/sse.md (avx_vec_concat<mode>):
3135 Substitute concat_tg_mode mode attribute with xtg_mode.
3136 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
3137 (concat_tg_mode): Remove mode attribute.
3139 2018-05-31 Martin Sebor <msebor@redhat.com>
3142 * calls.c (alloc_max_size): Correct a logic error/typo.
3143 Treat excessive arguments as infinite. Warn for invalid arguments.
3144 * doc/invoke.texi (-Walloc-size-larger-than): Update.
3146 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
3149 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
3150 and movx for Haswell.
3152 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
3153 Cesar Philippidis <cesar@codesourcery.com>
3156 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
3157 when emitting error on private/firstprivate reductions.
3158 * omp-low.c (lower_omp_target): Avoid reference-type processing
3159 on pointers for firstprivate clause.
3161 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3163 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
3166 * config/aarch64/aarch64-simd.md
3167 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
3168 (aarch64_ld1_x3_<mode>): Likewise
3169 (aarch64_st1x2<VALLDIF:mode>): Likewise
3170 (aarch64_st1_x2_<mode>): Likewise
3171 (aarch64_st1x3<VALLDIF:mode>): Likewise
3172 (aarch64_st1_x3_<mode>): Likewise
3173 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
3174 (vld1_s8_x3): Likewise.
3175 (vld1_u16_x3): Likewise.
3176 (vld1_s16_x3): Likewise.
3177 (vld1_u32_x3): Likewise.
3178 (vld1_s32_x3): Likewise.
3179 (vld1_u64_x3): Likewise.
3180 (vld1_s64_x3): Likewise.
3181 (vld1_f16_x3): Likewise.
3182 (vld1_f32_x3): Likewise.
3183 (vld1_f64_x3): Likewise.
3184 (vld1_p8_x3): Likewise.
3185 (vld1_p16_x3): Likewise.
3186 (vld1_p64_x3): Likewise.
3187 (vld1q_u8_x3): Likewise.
3188 (vld1q_s8_x3): Likewise.
3189 (vld1q_u16_x3): Likewise.
3190 (vld1q_s16_x3): Likewise.
3191 (vld1q_u32_x3): Likewise.
3192 (vld1q_s32_x3): Likewise.
3193 (vld1q_u64_x3): Likewise.
3194 (vld1q_s64_x3): Likewise.
3195 (vld1q_f16_x3): Likewise.
3196 (vld1q_f32_x3): Likewise.
3197 (vld1q_f64_x3): Likewise.
3198 (vld1q_p8_x3): Likewise.
3199 (vld1q_p16_x3): Likewise.
3200 (vld1q_p64_x3): Likewise.
3201 (vst1_s64_x2): Likewise.
3202 (vst1_u64_x2): Likewise.
3203 (vst1_f64_x2): Likewise.
3204 (vst1_s8_x2): Likewise.
3205 (vst1_p8_x2): Likewise.
3206 (vst1_s16_x2): Likewise.
3207 (vst1_p16_x2): Likewise.
3208 (vst1_s32_x2): Likewise.
3209 (vst1_u8_x2): Likewise.
3210 (vst1_u16_x2): Likewise.
3211 (vst1_u32_x2): Likewise.
3212 (vst1_f16_x2): Likewise.
3213 (vst1_f32_x2): Likewise.
3214 (vst1_p64_x2): Likewise.
3215 (vst1q_s8_x2): Likewise.
3216 (vst1q_p8_x2): Likewise.
3217 (vst1q_s16_x2): Likewise.
3218 (vst1q_p16_x2): Likewise.
3219 (vst1q_s32_x2): Likewise.
3220 (vst1q_s64_x2): Likewise.
3221 (vst1q_u8_x2): Likewise.
3222 (vst1q_u16_x2): Likewise.
3223 (vst1q_u32_x2): Likewise.
3224 (vst1q_u64_x2): Likewise.
3225 (vst1q_f16_x2): Likewise.
3226 (vst1q_f32_x2): Likewise.
3227 (vst1q_f64_x2): Likewise.
3228 (vst1q_p64_x2): Likewise.
3229 (vst1_s64_x3): Likewise.
3230 (vst1_u64_x3): Likewise.
3231 (vst1_f64_x3): Likewise.
3232 (vst1_s8_x3): Likewise.
3233 (vst1_p8_x3): Likewise.
3234 (vst1_s16_x3): Likewise.
3235 (vst1_p16_x3): Likewise.
3236 (vst1_s32_x3): Likewise.
3237 (vst1_u8_x3): Likewise.
3238 (vst1_u16_x3): Likewise.
3239 (vst1_u32_x3): Likewise.
3240 (vst1_f16_x3): Likewise.
3241 (vst1_f32_x3): Likewise.
3242 (vst1_p64_x3): Likewise.
3243 (vst1q_s8_x3): Likewise.
3244 (vst1q_p8_x3): Likewise.
3245 (vst1q_s16_x3): Likewise.
3246 (vst1q_p16_x3): Likewise.
3247 (vst1q_s32_x3): Likewise.
3248 (vst1q_s64_x3): Likewise.
3249 (vst1q_u8_x3): Likewise.
3250 (vst1q_u16_x3): Likewise.
3251 (vst1q_u32_x3): Likewise.
3252 (vst1q_u64_x3): Likewise.
3253 (vst1q_f16_x3): Likewise.
3254 (vst1q_f32_x3): Likewise.
3255 (vst1q_f64_x3): Likewise.
3256 (vst1q_p64_x3): Likewise.
3258 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3260 * config/msp430/msp430.c (msp430_output_labelref): Prepend
3261 user_label_prefix to name.
3263 * tree-core.h: Update comment about the format of NAME string
3264 passed to handler in attribute_spec.
3266 * config/msp430/msp430.md: Remove erroneous subreg expression from
3267 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
3268 zero_extend{q,h}isi2.
3270 2018-05-30 Borislav Petkov <bp@suse.de>
3272 * doc/extend.texi: Document some architecture specific
3273 constraints and sort entries.
3275 2018-05-30 Martin Sebor <msebor@redhat.com>
3278 * builtins.c (expand_builtin_stpcpy_1): New function.
3279 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
3280 only if the former succeeds.
3282 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3284 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
3287 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3289 * doc/invoke.texi (-flinker-output): Document
3291 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3293 * passes.c (ipa_write_summaries): Only modify statements if body
3295 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
3296 incrementally linking.
3297 (ipa_passes): Likewise.
3298 * lto-cgraph.c (lto_output_node): When incrementally linking do not
3299 pass down resolution info.
3300 * common.opt (flag_incremental_link): Update info.
3301 * gcc.c (plugin specs): Turn flinker-output=* to
3302 -plugin-opt=-linker-output-known
3303 * toplev.c (compile_file): Also cut compilation when doing incremental
3305 * flag-types. (enum lto_partition_model): Add
3306 LTO_LINKER_OUTPUT_NOLTOREL.
3307 (invoke.texi): Add -flinker-output docs.
3308 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
3309 link same way as WPA; do not stream in dead initializers.
3311 * dwarf2out.c (dwarf2out_die_ref_for_decl,
3312 darf2out_register_external_decl): Support incremental link.
3314 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3316 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
3318 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3320 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
3321 it down to simple_object_copy_lto_debug_sections.
3322 (run_gcc): Determine incremental LTO link time and configure
3323 lto1 into non-wpa mode, disable renaming of debug sections.
3325 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
3327 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3328 descriptions of various incorrectly documented functions.
3330 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
3333 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3334 address check not strict.
3336 2018-05-30 Richard Biener <rguenther@suse.de>
3338 PR tree-optimization/85964
3339 * tracer.c (better_p): Drop initialized count check, we only
3340 call the function with initialized counts now.
3341 (find_best_successor): Do find a best edge if one
3342 has uninitialized count.
3343 (find_best_predecessor): Likewise. Do BB frequency check only
3344 if count is initialized.
3346 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
3348 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
3349 (aarch64_ldrstr_offset_compare): New.
3350 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
3351 load/store orderings.
3352 (aarch64_gen_adjusted_ldpstp): Likewise.
3354 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
3356 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3357 Check for subset of GENERAL_REGS and FP_REGS.
3358 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
3361 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
3363 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
3364 and wi::to_poly_offset. Add the current offset and then check
3365 whether the sum fits, rather than using an unchecked addition of
3366 a checked term. Check for a shwi rather than a uhwi.
3367 * expr.c (get_bit_range): Use tree_to_poly_uint64.
3368 (store_constructor): Use poly_int_tree_p.
3369 (expand_expr_real_1): Likewise.
3370 * function.c (assign_temp): Likewise.
3371 * fold-const.c (const_binop): Use poly_int_tree_p and
3373 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
3375 * ipa-icf-gimple.c (func_checker::compare_operand): Use
3376 to_poly_offset for MEM offsets.
3377 * ipa-icf.c (sem_variable::equals): Likewise.
3378 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
3379 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
3380 wi::to_poly_offset for BIT_FIELD_REF offsets.
3381 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
3383 * var-tracking.c (emit_note_insn_var_location): Use
3384 tree_to_poly_uint64.
3386 2018-05-29 Jim Wilson <jimw@sifive.com>
3388 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
3390 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
3393 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
3394 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
3396 (sse4_1_round<mode>2): Use nonimmediate_operand
3397 for operand 1 predicate.
3399 2018-05-29 Martin Sebor <msebor@redhat.com>
3400 Richard Biener <rguenther@suse.de>
3403 * calls.c (get_size_range): Call determine_value_range instead
3404 of get_value_range..
3405 * tree-vrp.h (determine_value_range): Declared new function.
3406 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
3408 2018-05-29 Richard Biener <rguenther@suse.de>
3410 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
3411 sure to use non-pattern stmts for get_earlier_stmt arguments.
3412 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
3413 called on pattern stmts.
3414 (get_later_stmt): Likewise.
3416 2018-05-29 Martin Liska <mliska@suse.cz>
3418 PR gcov-profile/85759
3419 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
3422 2018-05-29 Jakub Jelinek <jakub@redhat.com>
3424 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
3426 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
3430 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3431 VEC_PACK_FLOAT_EXPR): New tree codes.
3432 * tree-pretty-print.c (op_code_prio): Handle
3433 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3434 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3435 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3436 * tree-inline.c (estimate_operator_cost): Likewise.
3437 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
3438 * fold-const.c (const_binop): Likewise.
3439 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
3440 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3441 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3442 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
3443 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3444 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3445 * expr.c (expand_expr_real_2): Likewise.
3446 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
3447 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
3448 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
3450 * optabs.c (expand_widen_pattern_expr): For
3451 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
3452 sign from result type rather than operand's type.
3453 (expand_binop_directly): For vec_packu_float_optab and
3454 vec_packs_float_optab allow result type to be different from operand's
3456 * optabs-tree.c (optab_for_tree_code): Handle
3457 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3458 VEC_PACK_FLOAT_EXPR. Formatting fixes.
3459 * tree-vect-generic.c (expand_vector_operations_1): Handle
3460 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3461 VEC_PACK_FLOAT_EXPR.
3462 * tree-vect-stmts.c (supportable_widening_operation): Handle
3464 (supportable_narrowing_operation): Handle FLOAT_EXPR.
3465 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
3466 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
3467 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
3468 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
3470 (vec_pack<floatprefix>_float_<mode>): New expander.
3471 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
3473 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
3474 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
3475 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
3476 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
3477 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
3479 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
3480 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
3481 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3482 VEC_PACK_FLOAT_EXPR): Document.
3484 2018-05-29 Richard Biener <rguenther@suse.de>
3486 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
3488 (stmt_vec_info_vec): Make pointer.
3489 (init_stmt_vec_info_vec): Remove.
3490 (free_stmt_vec_info_vec): Likewise.
3491 (set_stmt_vec_info_vec): New function.
3492 (free_stmt_vec_infos): Likewise.
3493 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
3494 (set_vinfo_for_stmt): Likewise.
3495 (get_earlier_stmt): Likewise.
3496 (get_later_stmt): Likewise.
3497 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
3498 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
3499 (vec_info::~vec_info): Free stmt_vec_infos.
3500 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3501 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3502 (pass_slp_vectorize::execute): Likewise.
3503 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
3504 (free_stmt_vec_info_vec): Likewise.
3505 (set_stmt_vec_info_vec): New function.
3506 (free_stmt_vec_infos): Likewise.
3507 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3508 the global stmt_vec_info_vec.
3509 * tree-parloops.c (gather_scalar_reductions): Use
3510 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3513 2018-05-29 Richard Biener <rguenther@suse.de>
3515 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
3517 2018-05-29 Martin Liska <mliska@suse.cz>
3518 David Malcolm <dmalcolm@redhat.com>
3520 * vec.c (test_reverse): New.
3521 (vec_c_tests): Add new test.
3522 * vec.h (vl_ptr>::reverse): New function.
3524 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
3526 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
3528 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3531 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3533 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3535 2018-05-28 Richard Biener <rguenther@suse.de>
3537 PR tree-optimization/85933
3538 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3539 look at stmts marked as vectorizable.
3541 2018-05-28 Richard Biener <rguenther@suse.de>
3543 PR tree-optimization/85934
3544 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3545 vector boolean check before scalar optimization.
3547 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3549 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3552 2018-05-28 Mark Wielaard <mark@klomp.org>
3554 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3555 if it is an expression containing a minus sign.
3557 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3559 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3561 2018-05-27 Paul Koning <ni1d@arrl.net>
3563 * config/pdp11/pdp11.md (truncsihi2): Remove.
3565 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3566 Chung-Ju Wu <jasonwucj@gmail.com>
3568 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3570 (unaligned_store_dw): Ditto.
3571 * config/nds32/nds32-memory-manipulation.c
3572 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3573 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3574 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3575 (emit_setmem_word_loop): Rename to ...
3576 (emit_setmem_doubleword_loop): ... this.
3577 (nds32_gen_dup_4_byte_to_word_value): New function.
3578 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3579 (nds32_expand_setmem_loop): Refine implementation.
3580 (nds32_expand_setmem_loop_v3m): Ditto.
3581 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3584 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3586 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3588 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3590 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3591 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3592 machine->attr_no_prologue_p.
3593 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3594 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3595 (nds32_expand_epilogue): Consider attr_naked_p.
3596 (nds32_expand_epilogue_v3pop): Likewise.
3597 (nds32_can_use_return_insn): Likewise.
3598 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3599 attr_no_prologue_p fields.
3600 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3602 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3605 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3607 * config/i386/sse.md
3608 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3610 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3612 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3614 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3616 (*<floatsuffix>floatv2div2sf2): Rename to ...
3617 (*float<floatunssuffix>v2div2sf2): ... this.
3618 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3619 (float<floatunssuffix>v2div2sf2_mask): ... this.
3620 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3621 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3622 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3624 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3626 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3628 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3630 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3632 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3634 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3635 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3636 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3637 gen_ufix_truncv8dfv8si2.
3638 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3639 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3640 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3641 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3642 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3643 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3644 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3645 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3647 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3651 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3653 2018-05-25 Jim Wilson <jimw@sifive.com>
3655 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3656 * config/riscv/riscv.c (struct machine_function): Add
3657 interrupt_handler_p and attribute_checked_p fields.
3658 (riscv_attribute_table): Add interrupt.
3659 (riscv_interrupt_type_p): New.
3660 (riscv_save_reg_p): Save extra regs for interrupt handler.
3661 (riscv_use_save_libcall): Return false for interrupt handler.
3662 (riscv_first_stack_step): Add forward declaration.
3663 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3664 for interrupt handler with large frame. Use it for saved reg list.
3665 (riscv_expand_prologue): Move flag_stack_usage_info support to
3666 eliminate duplication.
3667 (riscv_expand_epilogue): Generate mret for interrupt handler.
3668 (riscv_epilogue_uses): New.
3669 (riscv_can_use_return_insn): Return false for interrupt handler.
3670 (riscv_function_ok_for_sibcall): Likewise.
3671 (riscv_set_current_function): Add interrupt handler support.
3672 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3673 * config/riscv/riscv.md (UNSPECV_MRET): New.
3675 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3677 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3679 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3681 PR tree-optimization/85712
3682 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3683 this candidate has already been replaced in-situ by a copy.
3685 2018-05-25 Jason Merrill <jason@redhat.com>
3687 PR c++/80485 - inline function non-zero address.
3688 * symtab.c (nonzero_address): Check DECL_COMDAT.
3690 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3693 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3694 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3695 extension of SImode operation. Use const123_operand predicate.
3696 (*saddsi_1): Remove.
3697 (*saddl_se_1): Ditto.
3699 (*ssubl_se_1): Ditto.
3700 * config/alpha/predicates.md (const123_operand): New predicate.
3701 * config/alpha/constraints.md (P): Use IN_RANGE.
3703 2018-05-25 Richard Biener <rguenther@suse.de>
3705 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3707 (ref_maybe_used_by_stmt_p): Likewise.
3708 (stmt_may_clobber_ref_p): Likewise.
3709 (stmt_may_clobber_ref_p_1): Likewise.
3710 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3712 (ref_maybe_used_by_stmt_p): Likewise.
3713 (stmt_may_clobber_ref_p): Likewise.
3714 (stmt_may_clobber_ref_p_1): Likewise.
3715 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3716 the alias oracle to disambiguate DRs with stmts DR analysis
3718 (vect_analyze_data_refs): Do not give up on not analyzable
3719 DRs for BB vectorization. Remove code truncating the dataref
3722 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3725 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3726 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3727 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3729 2018-05-25 Richard Biener <rguenther@suse.de>
3731 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3732 function, combining stmt data ref gathering and fatal analysis
3734 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3735 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3736 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3737 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3738 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3740 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3742 PR tree-optimization/85720
3743 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3744 SCC if all partitions are builtins.
3745 (version_loop_by_alias_check): New parameter. Generate cancelable
3746 runtime alias check if all partitions are builtins.
3747 (distribute_loop): Update call to above function.
3749 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3751 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3752 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3753 (parm_default_def_partition_arg): Ditto.
3754 (set_parm_default_def_partition): Ditto.
3755 (get_parm_default_def_partitions): Ditto and make it static.
3756 (get_undefined_value_partitions): Ditto and make it static.
3757 (remove_ssa_form): Refactor call to init_var_map here.
3758 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3759 computation for loop region.
3760 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3761 (register_default_def): Delete.
3762 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3763 (parm_default_def_partition_arg): Ditto.
3764 (set_parm_default_def_partition): Ditto.
3765 (get_parm_default_def_partitions): Ditto and make it static.
3766 (get_undefined_value_partitions): Ditto and make it static.
3767 (coalesce_with_default, coalesce_with_default): Update comment.
3768 (create_coalesce_list_for_region): New func factored out from
3769 create_outofssa_var_map.
3770 (populate_coalesce_list_for_outofssa): New func factored out from
3771 create_outofssa_var_map and coalesce_ssa_name.
3772 (create_outofssa_var_map): Delete.
3773 (coalesce_ssa_name): Refactor to support live range computation.
3774 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3775 (get_parm_default_def_partitions): Delete.
3776 (get_undefined_value_partitions): Ditto.
3777 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3778 computation for loop region.
3779 (new_tree_live_info, loe_visit_block): Ditto.
3780 (live_worklist, set_var_live_on_entry): Ditto.
3781 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3782 * tree-ssa-live.h (struct _var_map): New fields.
3783 (init_var_map): Change decl.
3784 (region_contains_p): New.
3786 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3788 * tree-ssa-live.h (live_merge_and_clear): Delete.
3790 2018-05-25 Richard Biener <rguenther@suse.de>
3793 * tree-dump.c (dequeue_and_dump): Remove access to removed
3794 operand 2 of a SWITCH_EXPR.
3796 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3798 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3799 multiplication and division.
3800 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3801 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3802 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3803 (cond_udiv_optab, cond_umod_optab): New optabs.
3804 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3805 (IFN_COND_RDIV): New internal functions.
3806 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3807 TRUNC_MOD_EXPR and RDIV_EXPR.
3808 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3809 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3811 (SVE_INT_BINARY): Include mult.
3812 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3813 (optab, sve_int_op): Handle mult.
3814 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3816 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3817 for SVE_INT_BINARY_SD.
3819 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3821 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3822 (optab, sve_int_op): Handle div and udiv.
3823 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3824 for SVE_INT_BINARY_SD.
3825 (*<optab><mode>3): New insn for the same.
3827 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3829 * tree-vect-patterns.c: Include predict.h.
3830 (vect_recog_divmod_pattern): Restrict check for division support
3831 to when optimizing for size.
3833 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3835 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3836 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3837 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3838 (gimple_match_op::set_op): Likewise.
3839 (gimple_resimplify4): Declare.
3840 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3841 (expr::gen_transform): Likewise.
3842 (decision_tree::gen): Generate a simplification routine for 4 operands.
3843 * gimple-match-head.c (gimple_simplify): Add an overload for
3844 4 operands. In the top-level function, handle up to 4 call
3845 arguments and call gimple_resimplify4.
3846 (gimple_resimplify4): New function.
3847 (build_call_internal): Pass a fourth operand.
3848 (maybe_push_to_seq): Likewise.
3849 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3850 Fold VEC_COND_EXPRs of an operation and a default value into
3851 an IFN_COND_* function if possible.
3852 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3854 (SVE_COND_FP_BINARY): Include them.
3855 (optab, sve_fp_op): Handle them.
3856 (SVE_INT_BINARY_REV): New code iterator.
3857 (SVE_COND_FP_BINARY_REV): New int iterator.
3858 (commutative): New int attribute.
3859 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3861 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3863 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3864 (*cond_<optab><mode>): New patterns for reversed operands.
3866 2018-05-25 Richard Biener <rguenther@suse.de>
3868 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3869 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3870 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3871 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3872 * tree-vect-data-refs.c (everywhere): Adjust users.
3873 * tree-vect-loop.c (everywhere): Likewise.
3874 * tree-vect-slp.c (everywhere): Likewise.
3875 * tree-vect-stmts.c (everywhere): Likewise.
3876 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3878 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3880 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3882 (gcc_cv_as_section_exclude): ... this.
3883 Try Solaris as #exclude syntax.
3884 * configure: Regenerate.
3885 * config.in: Regenerate.
3886 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3888 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3889 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3891 * varasm.c (default_elf_asm_named_section): Don't check if
3892 HAVE_GAS_SECTION_EXCLUDE is defined.
3894 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3896 * doc/md.texi: Update the documentation of the cond_* optabs
3897 to mention the new final operand. Fix GET_MODE_NUNITS call.
3898 Describe the scalar case too.
3899 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3900 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3902 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3903 (get_conditional_internal_fn): Update comment.
3904 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3905 accumulator value as a final argument to conditional functions.
3906 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3907 a define_expand and add an "else" operand. Assert for now that
3908 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3909 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3910 (*cond_<optab><mode>): New patterns.
3911 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3912 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3913 (UNSPEC_COND_EOR): Delete.
3914 (optab): Remove associated mappings.
3915 (SVE_INT_BINARY): New code iterator.
3916 (sve_int_op): Remove int attribute and add "minus" to the code
3918 (SVE_COND_INT_OP): Delete.
3919 (SVE_COND_FP_OP): Rename to...
3920 (SVE_COND_FP_BINARY): ...this.
3922 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3924 * optabs.c (can_reuse_operands_p): New function.
3925 (maybe_legitimize_operands): Try to reuse the results for
3928 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3930 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3931 Add {q} suffix to insn mnemonic.
3933 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3935 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3936 (msp430_warn_func_return): New.
3938 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3940 * fold-const.c (tree_nonzero_bits): New function.
3941 * fold-const.h (tree_nonzero_bits): Likewise.
3942 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3943 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3945 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3949 * varasm.c (assemble_alias): Check ifunc_resolver only on
3952 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3955 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3956 when memory input operand is handled.
3958 2018-05-24 Luis Machado <luis.machado@linaro.org>
3960 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3962 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3964 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3966 * match.pd: Delay FMA folds until after vectorization.
3968 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3971 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3972 address check not strict.
3974 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3976 * gimple-match.h (gimple_match_op): New class.
3977 (mprts_hook): Replace parameters with a gimple_match_op *.
3978 (maybe_build_generic_op): Likewise.
3979 (gimple_simplified_result_is_gimple_val): Replace parameters with
3980 a const gimple_match_op *.
3981 (gimple_simplify): Replace code_helper * and tree * parameters with
3982 a gimple_match_op * parameter.
3983 (gimple_resimplify1): Replace code_helper *, tree and tree *
3984 parameters with a gimple_match_op * parameter.
3985 (gimple_resimplify2): Likewise.
3986 (gimple_resimplify3): Likewise.
3987 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3988 parameters with a gimple_match_op * parameter.
3989 * gimple-match-head.c (gimple_simplify): Change prototypes of
3990 auto-generated functions to take a gimple_match_op * instead of
3991 separate code_helper * and tree * parameters. Make the same
3992 change in the top-level overload and update calls to the
3993 gimple_resimplify routines. Update calls to the auto-generated
3994 functions and to maybe_push_res_to_seq in the publicly-facing
3995 operation-specific gimple_simplify overloads.
3996 (gimple_match_op::MAX_NUM_OPS): Define.
3997 (gimple_resimplify1): Replace rcode and ops with a single res_op
3998 parameter. Update call to gimple_simplify.
3999 (gimple_resimplify2): Likewise.
4000 (gimple_resimplify3): Likewise.
4001 (mprts_hook): Replace parameters with a gimple_match_op *.
4002 (maybe_build_generic_op): Likewise.
4003 (build_call_internal): Replace type, nargs and ops with
4004 a gimple_match_op *.
4005 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
4006 with a single gimple_match_op *. Update calls to mprts_hook,
4007 build_call_internal and gimple_simplified_result_is_gimple_val.
4008 Factor out code that is common to the tree_code and combined_fn cases.
4009 * genmatch.c (expr::gen_transform): Replace tem_code and
4010 tem_ops with a gimple_match_op called tem_op. Update calls
4011 to the gimple_resimplify functions and maybe_push_res_to_seq.
4012 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
4013 res_ops. Update call to the gimple_resimplify functions.
4014 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
4015 (decision_tree::gen): Make the functions take a gimple_match_op *
4016 called res_op instead of separate res_code and res_ops parameters.
4017 Update call accordingly.
4018 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
4019 and ops with a single res_op parameter. Update calls to
4020 maybe_build_generic_op and maybe_push_res_to_seq.
4021 (fold_stmt_1): Update calls to gimple_simplify and
4022 replace_stmt_with_simplification.
4023 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
4024 and gimple_simplified_result_is_gimple_val.
4025 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
4027 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
4028 with a gimple_match_op *.
4029 (vn_nary_build_or_lookup): Likewise. Update call to
4030 vn_nary_build_or_lookup_1.
4031 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
4032 gimple_match_op *. Update calls to the gimple_resimplify routines
4033 and to gimple_simplified_result_is_gimple_val.
4034 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
4035 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
4036 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
4037 (visit_nary_op): Likewise.
4038 (visit_reference_op_load): Likewise.
4040 2018-05-23 Luis Machado <luis.machado@linaro.org>
4042 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
4043 modifier for printing the step amount.
4045 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
4048 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
4051 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
4053 * doc/sourcebuild.texi (Endianness): New subsubsection.
4055 2018-05-23 Luis Machado <luis.machado@linaro.org>
4057 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4058 <prefetch_dynamic_strides>: New const bool field.
4059 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4060 prefetch_dynamic_strides.
4061 (exynosm1_prefetch_tune): Likewise.
4062 (thunderxt88_prefetch_tune): Likewise.
4063 (thunderx_prefetch_tune): Likewise.
4064 (thunderx2t99_prefetch_tune): Likewise.
4065 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
4067 (aarch64_override_options_internal): Update to set
4068 PARAM_PREFETCH_DYNAMIC_STRIDES.
4069 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4070 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4071 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4072 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4073 prefetch-dynamic-strides setting.
4075 2018-05-23 Luis Machado <luis.machado@linaro.org>
4077 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4078 <minimum_stride>: New const int field.
4079 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4080 minimum_stride field defaulting to -1.
4081 (exynosm1_prefetch_tune): Likewise.
4082 (thunderxt88_prefetch_tune): Likewise.
4083 (thunderx_prefetch_tune): Likewise.
4084 (thunderx2t99_prefetch_tune): Likewise.
4085 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
4086 <default_opt_level>: Set to 3.
4087 (aarch64_override_options_internal): Update to set
4088 PARAM_PREFETCH_MINIMUM_STRIDE.
4089 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4090 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4091 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4092 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4093 stride is constant and is below the minimum stride threshold.
4095 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4097 * config/arm/arm-cpus.in (mode26): Delete.
4098 (armv4): Delete mode26 reference.
4099 * config/arm/arm.c (arm_configure_build_target): Delete use of
4102 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4104 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
4106 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
4107 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
4109 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
4110 (floatunsdidf2): Ditto.
4112 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4114 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
4115 (fixuns_trunc<mode>si2_avx512f): Ditto.
4116 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
4117 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
4118 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
4120 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4122 PR rtl-optimization/79985
4123 * df-scan.c (df_insn_refs_collect): Remove special case for
4124 global registers and asm statements.
4126 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4128 * extend.texi (Global Register Variables): Rewrite the bullet list.
4129 Note that the register is available for allocation. Note that access
4130 via inline asm must use constraints. Add note about async-signal
4131 handlers. Remove paragraph about automagic register selection.
4133 2018-05-23 Richard Biener <rguenther@suse.de>
4135 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
4136 of fixed offset from memset VN.
4138 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
4140 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
4142 (alloc_cand_and_find_basis): Initialize first_interp field.
4143 (slsr_process_mul): Modify first_interp field.
4144 (slsr_process_add): Likewise.
4145 (slsr_process_cast): Modify first_interp field for each new
4147 (slsr_process_copy): Likewise.
4148 (dump_candidate): Dump first_interp field.
4149 (replace_mult_candidate): Process all interpretations, not just
4151 (replace_rhs_if_not_dup): Likewise.
4152 (replace_one_candidate): Likewise.
4154 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
4156 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
4158 (aarch64_needs_frame_chain): New function.
4159 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
4161 2018-05-23 Sudakshina Das <sudi.das@arm.com>
4164 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
4165 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
4166 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
4167 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
4168 as true for strict-align.
4169 (aarch64_can_inline_p): Perform checks even when callee has no
4170 attributes to check for strict alignment.
4171 * doc/extend.texi (AArch64 Function Attributes): Document
4173 * doc/invoke.texi: (AArch64 Options): Likewise.
4175 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
4177 PR tree-optimization/85853
4178 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
4179 the handling of the root of the node to...
4180 (vect_slp_analyze_node_operations_1): ...this new function,
4181 and run the whole thing with the child nodes' def types
4182 set according to their SLP node's def type.
4184 2018-05-23 Richard Biener <rguenther@suse.de>
4187 * tree-data-ref.c (create_runtime_alias_checks): Defer
4188 and ignore overflow warnings.
4190 2018-05-23 Yury Gribov <tetra2005@gmail.com>
4192 PR tree-optimization/85822
4193 * tree-vrp.c (is_masked_range_test): Fix handling of negative
4196 2018-05-23 Richard Biener <rguenther@suse.de>
4198 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4199 memset constants via native_interpret_expr.
4201 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
4204 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
4206 (cgraph_node::create_alias): Likewise.
4207 (cgraph_node::get_availability): Check ifunc_resolver instead
4208 of looking up ifunc attribute.
4209 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4210 * varasm.c (do_assemble_alias): Likewise.
4211 (assemble_alias): Likewise.
4212 (default_binds_local_p_3): Likewise.
4213 * cgraph.h (cgraph_node): Add ifunc_resolver.
4214 (cgraph_node::only_called_directly_or_aliased_p): Return false
4216 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
4218 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
4219 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
4220 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
4221 instead of looking up ifunc attribute.
4223 2018-05-22 Luis Machado <luis.machado@linaro.org>
4225 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
4227 2018-05-22 Martin Sebor <msebor@redhat.com>
4230 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
4231 only when expasion succeeds.
4232 (expand_builtin_strcmp): Same.
4233 (expand_builtin_strncmp): Same.
4235 2018-05-22 Martin Sebor <msebor@redhat.com>
4237 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
4239 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4240 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4242 * config/aarch64/aarch64-ldpstp.md: Replace uses of
4243 aarch64_mem_pair_operand with memory_operand and delete operand swapping
4245 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4246 Add check for legitimate_address.
4247 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
4248 (aarch64_swap_ldrstr_operands): New.
4249 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
4252 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4253 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4255 * config/aarch64/aarch64.md: New patterns to generate stp
4257 (store_pair_sw, store_pair_dw): New patterns to generate stp for
4258 single words and double words.
4259 (load_pair_sw, load_pair_dw): Likewise.
4260 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
4262 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
4264 * config/aarch64/aarch64-ldpstp.md: Modify peephole
4265 for different mode ldpstp and add peephole for merged zero stores.
4267 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4269 (aarch64_gen_store_pair): Rename calls to match new patterns.
4270 (aarch64_gen_load_pair): Rename calls to match new patterns.
4271 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
4272 (load_pair<DREG:mode><DREG2:mode>): ... This.
4273 (store_pair<mode>): Rename to...
4274 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
4275 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
4277 (V_INT_EQUIV): Handle SImode.
4278 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
4281 2018-05-22 Martin Sebor <msebor@redhat.com>
4284 * calls.c (maybe_warn_nonstring_arg): Use string length to set
4285 or ajust the presumed bound on an operation to avoid unnecessary
4288 2018-05-22 Martin Sebor <msebor@redhat.com>
4290 PR tree-optimization/85826
4291 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
4292 assuming that a DECL necesarily has a constant size.
4294 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
4297 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
4299 2018-05-22 Richard Biener <rguenther@suse.de>
4301 PR tree-optimization/85834
4302 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
4303 non-constant and non-zero memset arguments.
4305 2018-05-22 Martin Liska <mliska@suse.cz>
4308 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
4310 2018-05-22 Richard Biener <rguenther@suse.de>
4312 PR tree-optimization/85863
4313 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
4314 comparisons when vectype is specified.
4315 (vectorizable_condition): Do not specify vectype for
4316 vect_is_simple_cond when SLP vectorizing.
4318 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
4321 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
4322 define __ibm128 as long double.
4323 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
4324 as a distinct type when IEEE 128-bit support is enabled.
4325 (init_float128_ieee): Fix up conversions between IFmode and IEEE
4326 128-bit types to use the correct functions.
4327 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
4328 convert between 128-bit floating point types that have different
4329 modes but the same representation, instead of using gen_lowpart to
4331 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
4333 (IFKF_reg): New attributes to give the register constraints for
4335 (extend<mode>tf2_internal): New insns to mark an explicit
4336 conversion between 128-bit floating point types that have a
4337 different mode but share the same representation.
4339 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
4341 PR tree-optimization/85814
4342 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
4343 a null return from get_strinfo when unsharing the next
4344 strinfo in the chain.
4346 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4349 * varasm.c (weak_finish): Clean up weak_decls.
4351 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4353 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
4354 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
4355 UNSPEC_UADALP values.
4356 * config/aarch64/iterators.md (ABAL): New int iterator.
4359 (sur): Add mappings for the above.
4360 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
4362 (aarch64_<sur>abal<mode>_4): Likewise.
4363 (aarch64_<sur>adalp<mode>_3): Likewise.
4364 (<sur>sadv16qi): New define_expand.
4366 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
4368 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
4369 (*movdf_internal): Ditto.
4370 (*rcpsf2_sse): Ditto.
4371 (*rsqrtsf2_sse): Ditto.
4372 (*sqrt<mode>2_sse): Ditto.
4374 2018-05-21 Tamar Christina <tamar.christina@arm.com>
4376 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
4378 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
4379 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
4380 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
4381 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
4383 * config/aarch64/arm_neon.h: Likewise.
4384 * config/aarch64/iterators.md (VQ_I): New.
4386 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
4388 * config.gcc: Add arc/t-multilib-linux to tmake_file for
4390 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
4393 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4395 * config/nds32/constraints.md (S): New constraint.
4396 * config/nds32/nds32.md (call_internal): Use constraint S.
4397 (call_value_internal): Likewise.
4398 (sibcall_internal): Likewise.
4399 (sibcall_value_internal): Likewise.
4401 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4402 Chung-Ju Wu <jasonwucj@gmail.com>
4404 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
4407 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4408 Chung-Ju Wu <jasonwucj@gmail.com>
4410 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
4411 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
4412 (nds32_rtx_costs_impl): Simplify.
4413 (nds32_address_cost_impl): Simplify.
4414 (nds32_init_rtx_costs): New function.
4415 (nds32_rtx_costs_speed_prefer): Likewise.
4416 (nds32_rtx_costs_size_prefer): Likewise.
4417 (nds32_address_cost_speed_prefer): Likewise.
4418 (nds32_address_cost_speed_fwprop): Likewise.
4419 (nds32_address_cost_size_prefer): Likewise.
4420 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
4421 * config/nds32/nds32.c (nds32_option_override): Use
4422 nds32_init_rtx_costs function.
4424 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4426 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
4427 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
4428 (TARGET_PIPELINE_N8): Likewise.
4429 (TARGET_PIPELINE_N10): Likewise.
4430 (TARGET_PIPELINE_N13): Likewise.
4431 (TARGET_PIPELINE_GRAYWOLF): Likewise.
4433 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
4435 * config/nds32/nds32-fpu.md: Update copyright year.
4437 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4439 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
4441 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4443 * config/nds32/nds32.c
4444 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
4445 * config/nds32/nds32.opt (minline-asm-r15): New option.
4447 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4449 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
4451 * config/nds32/nds32.md (abssi2): New pattern.
4453 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
4455 * config/i386/i386.md (rex64namesuffix): New mode attribute.
4456 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
4457 Merge insn pattern from sse_cvtsi2ss<round_name> and
4458 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
4459 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
4460 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
4461 using SWI48 mode iterator.
4462 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
4463 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
4464 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
4465 pattern from sse_cvttss2si<round_saeonly_name>
4466 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
4467 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
4468 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
4469 using SWI48 mode iterator.
4470 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
4471 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
4472 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
4473 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
4474 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
4475 using SWI48 mode iterator.
4476 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
4477 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
4478 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
4479 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
4480 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
4481 SWI48 mode iterator.
4482 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
4483 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
4484 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
4485 pattern from sse_cvttsd2si<round_saeonly_name>
4486 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
4488 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4490 * config/nds32/nds32-md-auxiliary.c
4491 (nds32_valid_smw_lwm_base_p): Refine.
4492 (nds32_output_smw_single_word): Refine.
4493 (nds32_output_smw_double_word): New.
4494 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
4496 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4498 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
4499 (nds32_output_stack_pop): Refine.
4500 (nds32_expand_unaligned_load): Refine.
4501 (nds32_expand_unaligned_store): Refine.
4503 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4504 Chung-Ju Wu <jasonwucj@gmail.com>
4506 * config/nds32/constants.md: Add TP_REGNUM constant.
4507 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4508 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4510 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4511 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4512 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4513 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4514 and PIC code generation.
4515 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
4517 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
4519 * config/nds32/nds32.md: Support TLS and PIC.
4520 * config/nds32/nds32.c: Support TLS and PIC.
4521 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
4522 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
4525 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4527 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
4528 mode with E_ prefix.
4530 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4531 Chung-Ju Wu <jasonwucj@gmail.com>
4533 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4534 * config/nds32/nds32-md-auxiliary.c
4535 (symbolic_reference_mentioned_p): New.
4536 (nds32_legitimize_ict_address): New.
4537 (nds32_expand_ict_move): New.
4538 (nds32_indirect_call_referenced_p): New.
4539 (nds32_symbol_binds_local_p): Delete.
4540 (nds32_long_call_p): Modify.
4541 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4542 * config/nds32/nds32-protos.h
4543 (symbolic_reference_mentioned_p): Declare.
4544 (nds32_legitimize_ict_address): Declare.
4545 (nds32_expand_ict_move): Declare.
4546 (nds32_indirect_call_referenced_p): Declare.
4547 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4548 (nds32_relax_group): Use nds32_ict_const_p as condition.
4549 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4550 (nds32_asm_file_start): Output ict_model directive in asm code.
4551 (nds32_legitimate_address_p): Consider indirect call.
4552 (nds32_print_operand): Consider indirect call.
4553 (nds32_print_operand_address): Consider indirect call.
4554 (nds32_insert_attributes): Handle "indirect_call" attribute.
4555 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4556 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4557 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4558 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4559 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4560 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4561 (TARGET_ICT_MODEL_SMALL): Define.
4562 (TARGET_ICT_MODEL_LARGE): Define.
4563 * config/nds32/nds32.md (movsi): Consider ict model.
4564 (call, call_value): Consider ict model.
4565 (sibcall, sibcall_value): Consider ict model.
4566 * config/nds32/nds32.opt (mict-model): New option.
4567 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4570 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4571 Monk Chiang <sh.chiang04@gmail.com>
4572 Jim Wilson <jimw@sifive.com>
4574 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4575 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4576 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4577 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4578 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4579 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4580 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4581 compute save_libcall_adjustment properly.
4582 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4583 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4584 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4585 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4586 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4587 (ABI_SPEC): Handle mabi=ilp32e.
4588 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4589 (RVE): Add RVE mask.
4590 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4591 <-march>: Add rv32e as an example.
4593 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4596 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4597 (intra_create_variable_infos): Handle C++ constructors.
4599 2018-05-18 Martin Liska <mliska@suse.cz>
4601 * passes.def: Remove a redundant pass.
4603 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4606 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4608 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4610 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4612 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4614 (armv2, armv2a, armv3, armv3m): Delete architectures.
4615 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4616 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4617 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4619 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4620 (*mulsidi3adddi): Likewise.
4621 (mulsidi3): Likewise.
4622 (*mulsidi3_nov6): Likewise.
4623 (umulsidi3): Likewise.
4624 (umulsidi3_nov6): Likewise.
4625 (umaddsidi4): Likewise.
4626 (*umulsidi3adddi): Likewise.
4627 (smulsi3_highpart): Likewise.
4628 (*smulsi3_highpart_nov6): Likewise.
4629 (umulsi3_highpart): Likewise.
4630 (*umulsi3_highpart_nov6): Likewise.
4631 * config/arm/arm.h (arm_arch3m): Delete.
4632 * config/arm/arm.c (arm_arch3m): Delete.
4633 (arm_option_override_internal): Update armv3-related comment.
4634 (arm_configure_build_target): Delete use of isa_bit_mode32.
4635 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4636 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4637 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4639 (mulusa3): Likewise.
4640 * config/arm/arm-protos.h (arm_arch3m): Delete.
4641 * config/arm/arm-tables.opt: Regenerate.
4642 * config/arm/arm-tune.md: Likewise.
4643 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4644 deleted architectures.
4646 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4648 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4649 (armv5t, armv5te): New features.
4650 (ARMv5, ARMv5e): Delete fgroups.
4651 (ARMv5t, ARMv5te): Adjust for above changes.
4653 (armv5, armv5e): Delete arches.
4654 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4656 (*call_reg_arm): Likewise.
4657 (*call_value_reg_armv5): Likewise.
4658 (*call_value_reg_arm): Likewise.
4659 (*call_symbol): Likewise.
4660 (*call_value_symbol): Likewise.
4661 (*sibcall_insn): Likewise.
4662 (*sibcall_value_insn): Likewise.
4664 (prefetch): Likewise.
4665 (define_split and define_peephole2 dependent on arm_arch5):
4667 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4669 (TARGET_ARM_QBIT): Likewise.
4670 (TARGET_DSP_MULTIPLY): Likewise.
4671 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4672 (arm_arch5, arm_arch5e): Delete.
4673 (arm_arch5t, arm_arch5te): Declare.
4674 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4675 (arm_arch5t): Declare.
4676 (arm_option_reconfigure_globals): Update for the above.
4677 (arm_options_perform_arch_sanity_checks): Update comment, replace
4678 use of arm_arch5 with arm_arch5t.
4679 (use_return_insn): Likewise.
4680 (arm_emit_call_insn): Likewise.
4681 (output_return_instruction): Likewise.
4682 (arm_final_prescan_insn): Likewise.
4683 (arm_coproc_builtin_available): Likewise.
4684 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4685 arm_arch5e with arm_arch5t and arm_arch5te.
4686 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4687 (arm_arch5t, arm_arch5te): Declare.
4688 * config/arm/arm-tables.opt: Regenerate.
4689 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4690 * config/arm/t-multilib: Likewise.
4691 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4692 instead of arm_arch5.
4693 (*call_reg_thumb1): Likewise.
4694 (*call_value_reg_thumb1_v5): Likewise.
4695 (*call_value_reg_thumb1): Likewise.
4696 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4698 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4700 2018-05-18 Martin Liska <mliska@suse.cz>
4702 PR gcov-profile/84846
4703 * doc/gcov.texi: Document -t option of gcov tool.
4705 2018-05-18 Martin Liska <mliska@suse.cz>
4707 PR gcov-profile/84846
4708 * gcov.c (print_usage): Add new -t option.
4709 (process_args): Handle the option.
4710 (generate_results): Use stdout as output when requested by
4713 2018-05-18 Martin Liska <mliska@suse.cz>
4715 PR gcov-profile/84846
4716 * coverage.c (coverage_init): Write PWD to .gcno file.
4717 * doc/gcov.texi: Document how working directory is printed.
4718 * gcov-dump.c (dump_gcov_file): Print PWD.
4719 * gcov.c (output_intermediate_file): Likewise.
4720 (read_graph_file): Read PWD string.
4721 (output_lines): Print PWD.
4723 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4726 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4727 for retval and return false if all args to phi are zero.
4729 2018-05-18 Richard Biener <rguenther@suse.de>
4731 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4733 (evrp_dom_walker::before_dom_children): Call it.
4735 2018-05-18 Richard Biener <rguenther@suse.de>
4737 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4738 results when processing array refs with variable index.
4740 2018-05-18 Toon Moene <toon@moene.org>
4742 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4743 directly after that of -floop-interchange. Indicate that both
4744 options are enabled by default when specifying -O3.
4746 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4748 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4749 iterator. Delete separate integer-mode vec_set<mode> expander.
4750 (aarch64_simd_vec_setv2di): Delete.
4751 (vec_setv2di): Delete.
4752 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4753 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4754 the "w, r" alternative.
4756 2018-05-18 Martin Liska <mliska@suse.cz>
4758 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4759 * tree-pass.h (make_pass_lower_switch_O0): New function.
4760 * tree-switch-conversion.c (node_has_low_bound): Remove.
4761 (node_has_high_bound): Likewise.
4762 (node_is_bounded): Likewise.
4763 (class pass_lower_switch): Make it a template type and create
4765 (pass_lower_switch::execute): Add template argument.
4766 (make_pass_lower_switch): New function.
4767 (make_pass_lower_switch_O0): New function.
4768 (do_jump_if_equal): Remove.
4769 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4770 all the hard work to tree optimization passes.
4772 2018-05-18 Martin Liska <mliska@suse.cz>
4774 * dbgcnt.c (limit_low): Renamed from limit.
4775 (limit_high): New variable.
4776 (dbg_cnt_is_enabled): Check for upper limit.
4777 (dbg_cnt): Adjust dumping.
4778 (dbg_cnt_set_limit_by_index): Add new argument for high
4780 (dbg_cnt_set_limit_by_name): Likewise.
4781 (dbg_cnt_process_single_pair): Parse new format.
4782 (dbg_cnt_process_opt): Use strtok.
4783 (dbg_cnt_list_all_counters): Remove 'value' and add
4785 * doc/invoke.texi: Document changes.
4787 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4789 * doc/sourcebuild.texi (scalar_all_fma): Document.
4790 * tree.def (FMA_EXPR): Delete.
4791 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4792 * internal-fn.c (ternary_direct): New macro.
4793 (expand_ternary_optab_fn): Likewise.
4794 (direct_ternary_optab_supported_p): Likewise.
4795 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4796 * builtins.c (fold_builtin_fma): Delete.
4797 (fold_builtin_3): Don't call it.
4798 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4799 * expr.c (expand_expr_real_2): Likewise.
4800 * fold-const.c (operand_equal_p): Likewise.
4801 (fold_ternary_loc): Likewise.
4802 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4803 * gimple.c (DEFTREECODE): Likewise.
4804 * gimplify.c (gimplify_expr): Likewise.
4805 * optabs-tree.c (optab_for_tree_code): Likewise.
4806 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4807 * tree-eh.c (operation_could_trap_p): Likewise.
4808 (stmt_could_throw_1_p): Likewise.
4809 * tree-inline.c (estimate_operator_cost): Likewise.
4810 * tree-pretty-print.c (dump_generic_node): Likewise.
4811 (op_code_prio): Likewise.
4812 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4813 * tree-ssa-operands.c (get_expr_operands): Likewise.
4814 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4815 * fold-const-call.h (fold_fma): Delete.
4816 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4817 CFN_FNMA and CFN_FNMS.
4819 * genmatch.c (combined_fn): New enum.
4820 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4821 (commutative_op): New function.
4822 (commutate): Use it. Handle more than 2 operands.
4823 (dt_operand::gen_gimple_expr): Use commutative_op.
4824 (parser::parse_expr): Allow :c to be used with non-binary
4825 operators if the commutative operand is known.
4826 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4827 CFN_FMS, CFN_FNMA and CFN_FNMS.
4828 (backprop::process_assign_use): Remove FMA_EXPR handling.
4829 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4830 (gen_hsa_fma): New function.
4831 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4832 IFN_FNMA and IFN_FNMS.
4833 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4834 * gimple-fold.h (follow_all_ssa_edges): Declare.
4835 * gimple-fold.c (follow_all_ssa_edges): New function.
4836 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4837 gimple_build interface and use follow_all_ssa_edges to fold the result.
4838 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4839 instead of checking for optabs directly.
4840 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4841 rather than FMA_EXPRs.
4842 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4843 call to IFN_FMA instead of an FMA_EXPR.
4845 2018-05-17 Jim Wilson <jimw@sifive.com>
4847 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4848 sign extended promoted subreg, and the range does not have the sign bit
4849 set, then do a sign extend.
4851 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4852 test, check for sign extended subreg and/or constant operands, and
4853 do a sign extend in that case.
4855 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4857 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4858 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4860 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4861 Change logics_shift_reg to logics_shift_imm.
4862 (thunderx2t99_fp_loadpair_basic): Delete.
4863 (thunderx2t99_fp_storepair_basic): Delete.
4864 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4865 (thunderx2t99_asimd_polynomial): Delete.
4866 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4867 and neon_fp_mul_d_scalar_q.
4868 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4869 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4870 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4871 (thunderx2t99_asimd_lut): Add missing tbl types.
4872 (thunderx2t99_asimd_ext): Delete.
4873 (thunderx2t99_asimd_load1_1_mult): Delete.
4874 (thunderx2t99_asimd_load1_2_mult): Delete.
4875 (thunderx2t99_asimd_load1_ldp): New.
4876 (thunderx2t99_asimd_load1): New.
4877 (thunderx2t99_asimd_load2): Add missing *load2* types.
4878 (thunderx2t99_asimd_load3): New.
4879 (thunderx2t99_asimd_load4): New.
4880 (thunderx2t99_asimd_store1_1_mult): Delete.
4881 (thunderx2t99_asimd_store1_2_mult): Delete.
4882 (thunderx2t99_asimd_store2_mult): Delete.
4883 (thunderx2t99_asimd_store2_onelane): Delete.
4884 (thunderx2t99_asimd_store_stp): New.
4885 (thunderx2t99_asimd_store1): New.
4886 (thunderx2t99_asimd_store2): New.
4887 (thunderx2t99_asimd_store3): New.
4888 (thunderx2t99_asimd_store4): New.
4890 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4892 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4893 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4895 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4896 Segher Boessenkool <segher@kernel.crashing.org>
4899 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4902 2018-05-17 Richard Biener <rguenther@suse.de>
4904 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4905 for pruning loop and prune defs feeding only already visited PHIs.
4907 2018-05-17 Richard Biener <rguenther@suse.de>
4909 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4911 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4912 Richard Biener <rguenther@suse.de>
4914 PR tree-optimization/85793
4915 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4916 for VMAT_ELEMENTWISE.
4918 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4920 * internal-fn.h (lookup_internal_fn): Declare
4921 * internal-fn.c (lookup_internal_fn): New function.
4922 * gimple.c (gimple_build_call_from_tree): Handle calls to
4924 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4925 internal function names.
4926 * tree-pretty-print.c (dump_generic_node): Likewise.
4927 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4929 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4931 * gimple-fold.h (gimple_build): Make the function forms take
4932 combined_fn rather than built_in_function.
4933 (gimple_simplify): Likewise.
4934 * gimple-match-head.c (gimple_simplify): Likewise.
4935 * gimple-fold.c (gimple_build): Likewise.
4936 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4937 rather than gimple_build_call_internal.
4938 (get_initial_defs_for_reduction): Likewise.
4939 (vect_create_epilog_for_reduction): Likewise.
4940 (vectorizable_live_operation): Likewise.
4942 2018-05-17 Martin Liska <mliska@suse.cz>
4944 * gimple-ssa-sprintf.c (format_directive): Do not use
4945 space in between 'G_' and '('.
4947 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4950 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4951 even if the mask is not all ones.
4954 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4956 (ix86_gimple_fold_builtin): Likewise.
4959 * config/i386/i386.c: Include tree-vector-builder.h.
4960 (ix86_vector_shift_count): New function.
4961 (ix86_fold_builtin): Fold shift builtins by scalar count.
4962 (ix86_gimple_fold_builtin): Likewise.
4964 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4965 _mm512_setzero): New intrinsics.
4967 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4968 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4970 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4971 code generation for cases where splatting a value is not useful.
4972 * simplify-rtx.c (simplify_ternary_operation): Simplify
4973 vec_merge across a vec_duplicate and a paradoxical subreg forming
4974 a vector mode to a vec_concat.
4976 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4978 * config.gcc: Support "goldmont-plus".
4979 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4981 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4982 PROCESSOR_GOLDMONT_PLUS.
4983 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4984 (processor_target_table): Add "goldmont-plus".
4985 (PTA_GOLDMONT_PLUS): Define.
4986 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4987 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4988 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4989 (fold_builtin_cpu): Add "goldmont-plus".
4990 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4991 (ix86_option_override_internal): Add "goldmont-plus".
4992 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4993 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4994 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4995 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4997 2018-05-17 Richard Biener <rguenther@suse.de>
4999 PR tree-optimization/85757
5000 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
5001 remove defs that only feed that PHI from further processing.
5003 2018-05-16 Jim Wilson <jimw@sifive.com>
5005 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
5007 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
5009 2018-05-16 Mark Wielaard <mark@klomp.org>
5011 * dwarf2out.c (count_index_strings): New function.
5012 (output_indirect_strings): Call count_index_strings and generate
5013 header for dwarf_version >= 5.
5015 2018-05-16 Mark Wielaard <mark@klomp.org>
5017 * dwarf2out.c (dwarf_FORM): New function.
5018 (set_indirect_string): Use dwarf_FORM.
5019 (reset_indirect_string): Likewise.
5020 (size_of_die): Likewise.
5021 (value_format): Likewise.
5022 (output_die): Likewise.
5023 (add_skeleton_AT_string): Likewise.
5024 (output_macinfo_op): Likewise.
5025 (index_string): Likewise.
5026 (output_index_string_offset): Likewise.
5027 (output_index_string): Likewise.
5028 (count_index_strings): Likewise.
5030 2018-05-16 Carl Love <cel@us.ibm.com>
5032 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5033 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
5035 2018-05-16 Martin Jambor <mjambor@suse.cz>
5037 * ipa-prop.c (ipa_free_all_edge_args): Remove.
5038 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
5040 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
5042 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
5043 (fnma<mode>4): Likewise.
5044 (fms<mode>4): Likewise.
5045 (fnms<mode>4): Likewise.
5046 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
5047 (aarch64_fnma<mode>4): Likewise.
5048 (aarch64_fms<mode>4): Likewise.
5049 (aarch64_fnms<mode>4): Likewise.
5050 (aarch64_fnmadd<mode>4): Likewise.
5052 2018-05-16 Jason Merrill <jason@redhat.com>
5054 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
5056 2018-05-16 Richard Biener <rguenther@suse.de>
5058 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
5059 (dump_stmt_cost): Declare.
5060 (add_stmt_cost): Dump cost we add.
5061 (add_stmt_costs): New function.
5062 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
5064 (vect_analyze_stmt): Adjust prototype.
5065 (vectorizable_condition): Likewise.
5066 (vectorizable_live_operation): Likewise.
5067 (vectorizable_reduction): Likewise.
5068 (vectorizable_induction): Likewise.
5069 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
5070 cost vector to pass to vectorizable_ and record afterwards.
5071 (vect_model_reduction_cost): Take cost vector argument and adjust.
5072 (vect_model_induction_cost): Likewise.
5073 (vectorizable_reduction): Likewise.
5074 (vectorizable_induction): Likewise.
5075 (vectorizable_live_operation): Likewise.
5076 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
5077 SLP_TREE_NUMBER_OF_VEC_STMTS.
5078 (vect_analyze_slp_cost_1): Remove.
5079 (vect_analyze_slp_cost): Likewise.
5080 (vect_slp_analyze_node_operations): Take visited args and
5081 a target cost vector. Avoid processing already visited stmt sets.
5082 (vect_slp_analyze_operations): Use a local cost vector to gather
5083 costs and register those of non-discarded instances.
5084 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
5085 (vect_schedule_slp_instance): Remove copying of
5086 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
5088 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
5089 adding cost. Record cost entry location.
5090 (vect_prologue_cost_for_slp_op): Function to compute cost of
5091 a constant or invariant generated for SLP vect in the prologue,
5092 split out from vect_analyze_slp_cost_1.
5093 (vect_model_simple_cost): Make static. Adjust for SLP costing.
5094 (vect_model_promotion_demotion_cost): Likewise.
5095 (vect_model_store_cost): Likewise, make static.
5096 (vect_model_load_cost): Likewise.
5097 (vectorizable_bswap): Add cost vector arg and adjust.
5098 (vectorizable_call): Likewise.
5099 (vectorizable_simd_clone_call): Likewise.
5100 (vectorizable_conversion): Likewise.
5101 (vectorizable_assignment): Likewise.
5102 (vectorizable_shift): Likewise.
5103 (vectorizable_operation): Likewise.
5104 (vectorizable_store): Likewise.
5105 (vectorizable_load): Likewise.
5106 (vectorizable_condition): Likewise.
5107 (vectorizable_comparison): Likewise.
5108 (can_vectorize_live_stmts): Likewise.
5109 (vect_analyze_stmt): Likewise.
5110 (vect_transform_stmt): Adjust calls to vectorizable_*.
5111 * tree-vectorizer.c: Include gimple-pretty-print.h.
5112 (dump_stmt_cost): New function.
5114 2018-05-16 Richard Biener <rguenther@suse.de>
5116 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
5117 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
5118 * tree-ssa-dse.c: Include tree-ssa-loop.h.
5119 (check_name): New callback.
5120 (dse_classify_store): Track cycles via a visited bitmap of PHI
5121 defs and simplify handling of in-loop and across loop dead stores
5122 and properly fail for loop-variant refs. Handle byte-tracking with
5123 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
5126 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
5128 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
5129 (vect_get_mask_type_for_stmt): Likewise.
5130 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
5132 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
5133 to determine the statement's vector type and the vector type that
5134 should be used for calculating nunits. Deal with cases in which
5135 the type has to be deferred.
5136 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
5137 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
5138 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
5139 (vect_determine_vf_for_stmt): New functions, split out from...
5140 (vect_determine_vectorization_factor): ...here.
5141 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
5142 (vect_get_mask_type_for_stmt): New functions, split out from
5143 vect_determine_vectorization_factor.
5145 2018-05-16 Richard Biener <rguenther@suse.de>
5147 * tree-cfg.c (verify_gimple_assign_ternary): Properly
5148 verify the [VEC_]COND_EXPR embedded comparison.
5150 2018-05-15 Martin Sebor <msebor@redhat.com>
5152 PR tree-optimization/85753
5153 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
5154 RECORD_TYPE in addition to ARRAY_TYPE.
5156 2018-05-15 Martin Sebor <msebor@redhat.com>
5159 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
5161 2018-05-15 Richard Biener <rguenther@suse.de>
5163 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
5164 add by_clobber_p one. Change algorithm to collect all defs
5165 representing uses we need to walk and try reducing them to
5166 a single one before failing.
5167 (dse_dom_walker::dse_optimize_stmt): Adjust.
5169 2018-05-13 Mark Wielaard <mark@klomp.org>
5171 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
5172 (size_of_loc_descr): Likewise.
5173 (output_loc_operands): Likewise.
5174 (output_loc_operands_raw): Likewise.
5175 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
5176 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
5177 (hash_loc_operands): Likewise.
5178 (compare_loc_operands): Likewise.
5180 2018-05-14 Mark Wielaard <mark@klomp.org>
5182 * dwarf2out.c (count_index_addrs): New function.
5183 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
5185 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5187 PR tree-optimization/83648
5188 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
5189 return value as malloc candidate.
5191 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5194 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
5195 param as true in call to suggest_attribute.
5197 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
5199 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
5200 -mreadonly-in-sdata.
5202 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5204 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
5206 (aarch64_crypto_aesd_fused): Likewise.
5208 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
5210 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
5211 (movsi_aarch64): Likewise.
5212 (load_pairsi): Likewise.
5213 (load_pairdi): Likewise.
5214 (store_pairsi): Likewise.
5215 (store_pairdi): Likewise.
5216 (load_pairsf): Likewise.
5217 (load_pairdf): Likewise.
5218 (store_pairsf): Likewise.
5219 (store_pairdf): Likewise.
5220 (zero_extend): Likewise.
5221 (trunc): Swap alternatives.
5222 (fcvt_target): Add '?' to prefer w over r.
5224 2018-05-14 Jakub Jelinek <jakub@redhat.com>
5227 * config/i386/i386.md: Disallow non-commutative arithmetics in
5228 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
5229 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
5230 in the peephole2 before it.
5232 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
5234 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
5235 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
5236 (ix86_handle_option): Handle -mcldemote.
5237 * config.gcc: New header.
5238 * config/i386/cldemoteintrin.h: New file.
5239 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
5240 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5242 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5243 OPTION_MASK_ISA_CLDEMOTE.
5244 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
5245 (ix86_valid_target_attribute_inner_p): Ditto.
5246 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
5247 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
5248 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
5249 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
5250 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
5252 * config/i386/i386.opt: Add -mcldemote.
5253 * config/i386/x86intrin.h: New header.
5254 * doc/invoke.texi: Add -mcldemote.
5256 2018-05-14 Richard Biener <rguenther@suse.de>
5258 * doc/match-and-simplify.texi: Adjust :s documentation.
5260 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
5262 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
5263 intended memcpy size.
5264 (REORDER_45): Likewise.
5266 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
5268 * sort.cc: New file.
5269 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
5270 * vec.c (qsort_chk): Use gcc_qsort.
5271 * Makefile.in (OBJS-libcommon): Add sort.o.
5272 (build/sort.o): New target. Use it...
5273 (BUILD_RTL): ... here, and...
5274 (build/gencfn-macros): ... here, and...
5275 (build/genmatch): ... here.
5277 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5278 Chung-Ju Wu <jasonwucj@gmail.com>
5280 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
5281 * config/nds32/nds32-graywolf.md: New file.
5282 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
5283 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
5285 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
5286 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
5287 * config/nds32/nds32.md (pipeline_model): Add graywolf.
5288 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
5289 * config/nds32/pipelines.md: Include n15 settings.
5291 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5292 Chung-Ju Wu <jasonwucj@gmail.com>
5294 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
5295 * config/nds32/nds32-n13.md: New file.
5296 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
5297 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
5299 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
5300 * config/nds32/nds32.md (pipeline_model): Add n13.
5301 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
5302 * config/nds32/pipelines.md: Include n13 settings.
5304 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5305 Chung-Ju Wu <jasonwucj@gmail.com>
5307 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
5308 * config/nds32/nds32-n10.md: New file.
5309 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
5310 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
5312 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
5313 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
5314 * config/nds32/nds32.md (pipeline_model): Add n10.
5315 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
5316 * config/nds32/pipelines.md: Include n10 settings.
5318 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
5319 Kito Cheng <kito.cheng@gmail.com>
5320 Chung-Ju Wu <jasonwucj@gmail.com>
5322 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
5323 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5324 Add enum values for DSP extension instructions.
5325 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
5327 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
5328 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
5330 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
5331 * config/nds32/nds32-dspext.md: New file for DSP implementation.
5332 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
5333 * config/nds32/nds32-intrinsic.md: Likewise.
5334 * config/nds32/nds32_intrinsic.h: Likewise.
5335 * config/nds32/nds32-md-auxiliary.c: Likewise.
5336 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
5337 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
5338 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
5339 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
5340 * config/nds32/nds32-protos.h: New declarations for DSP extension.
5341 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
5342 TYPE_DMAC in switch statement.
5343 * config/nds32/nds32.c: New checking and implementation for DSP
5344 extension instructions.
5345 * config/nds32/nds32.h: Likewise.
5346 * config/nds32/nds32.md: Likewise.
5347 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
5348 * config/nds32/predicates.md: Implement new predicates for DSP
5351 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5353 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
5354 Reformat alternatives and attributes so it is easier to identify
5355 which constraints/attributes go with which instruction.
5356 (mov<mode>_hardfloat32, FMOVE64): Likewise.
5357 (mov<mode>_softfloat32, FMOVE64): Likewise.
5358 (mov<mode>_hardfloat64, FMOVE64): Likewise.
5359 (mov<mode>_softfloat64, FMOVE64): Likewise.
5361 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
5363 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5365 (Basic PowerPC Built-in Functions): The new name of the
5366 subsection previously known as "PowerPC Built-in Functions".
5367 (Basic PowerPC Built-in Functions Available on all Configurations):
5369 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
5370 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
5371 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
5372 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
5374 2018-05-11 Martin Jambor <mjambor@suse.cz>
5377 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
5380 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5383 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
5385 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
5387 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
5388 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
5389 (ix86_handle_option): Handle -mwaitpkg.
5390 * config.gcc: New header.
5391 * config/i386/cpuid.h (bit_WAITPKG): New bit.
5392 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
5393 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
5395 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5396 OPTION_MASK_ISA_WAITPKG.
5397 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
5398 (ix86_option_override_internal): Add PTA_WAITPKG.
5399 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
5400 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
5401 IX86_BUILTIN_TPAUSE.
5402 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
5403 __builtin_ia32_umwait and __builtin_ia32_tpause.
5404 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
5405 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
5406 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
5407 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
5408 UNSPECV_TPAUSE): New.
5409 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
5410 * config/i386/i386.opt: Add -mwaitpkg.
5411 * config/i386/waitpkgintrin.h: New file.
5412 * config/i386/x86intrin.h: New header.
5413 * doc/invoke.texi: Add -mwaitpkg.
5415 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5418 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
5420 (cortex-m0): Use armv6s-m isa.
5421 (cortex-m0plus): Likewise.
5422 (cortex-m1): Likewise.
5423 (cortex-m0.small-multiply): Likewise.
5424 (cortex-m0plus.small-multiply): Likewise.
5425 (cortex-m1.small-multiply): Likewise.
5427 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
5428 Jakub Jelinek <jakub@redhat.com>
5430 PR tree-optimization/85692
5431 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
5432 source permute as well.
5434 2018-05-11 Martin Liska <mliska@suse.cz>
5437 * doc/extend.texi: Document LLVM style format for no_sanitize
5440 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5442 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
5443 mode_supports_vsx_dform_quad to mode_supports_dq_form.
5444 (mode_supports_vsx_dform_quad): Likewise.
5445 (mode_supports_vmx_dform): Move these functions to be next to the
5446 other mode_supports functions.
5447 (mode_supports_dq_form): Likewise.
5448 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
5449 mode_supports_dq_form.
5450 (reg_offset_addressing_ok_p): Likewise.
5451 (offsettable_ok_by_alignment): Likewise.
5452 (rs6000_legitimate_offset_address_p): Likewise.
5453 (legitimate_lo_sum_address_p): Likewise.
5454 (rs6000_legitimize_address): Likewise.
5455 (rs6000_legitimize_reload_address): Likewise.
5456 (rs6000_secondary_reload_inner): Likewise.
5457 (rs6000_preferred_reload_class): Likewise.
5458 (rs6000_output_move_128bit): Likewise.
5460 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5462 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
5463 Generate SImode target register for null target.
5464 <case IX86_BUILTIN_XGETBV>: Ditto.
5465 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
5466 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
5468 2018-05-10 Carl Love <cel@us.ibm.com>
5470 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5471 dcbtt and dcbtstt if operands[2] is 0.
5473 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5476 * config/i386/sse.md (usadv64qi): New expander.
5478 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
5480 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
5481 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
5482 -maltivec=be support.
5483 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
5484 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
5485 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
5486 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
5487 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
5488 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
5489 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
5490 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
5491 altivec_vsumsws): Adjust.
5492 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
5493 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
5494 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
5496 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
5497 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
5498 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
5499 (altivec_lve<VI_char>x): Delete expand.
5500 (*altivec_lve<VI_char>x_internal): Rename to...
5501 (altivec_lve<VI_char>x): ... this.
5502 (altivec_lvxl_<mode>): Delete expand.
5503 (*altivec_lvxl_<mode>_internal): Rename to ...
5504 (altivec_lvxl_<mode>): ... this.
5505 (altivec_stvxl_<mode>): Delete expand.
5506 (*altivec_stvxl_<mode>_internal): Rename to ...
5507 (altivec_stvxl_<mode>): ... this.
5508 (altivec_stve<VI_char>x): Delete expand.
5509 (*altivec_stve<VI_char>x_internal): Rename to ...
5510 (altivec_stve<VI_char>x): ... this.
5511 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5512 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5513 reduc_plus_scal_<mode>): Adjust.
5514 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
5516 (rs6000_cpu_cpp_builtins): Adjust.
5517 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
5518 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
5519 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
5520 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
5521 -maltivec=be support.
5522 (rs6000_split_vec_extract_var): Adjust.
5523 (rs6000_split_v4si_init): Adjust.
5524 (swap_selector_for_mode): Delete.
5525 (altivec_expand_lvx_be, altivec_expand_stvx_be,
5526 altivec_expand_stvex_be): Delete.
5527 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
5528 -maltivec=be support.
5529 (rs6000_gimple_fold_builtin): Ditto.
5530 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5532 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5533 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5534 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5535 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5536 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5537 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5538 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5539 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5540 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5541 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5542 anonymous split): Adjust.
5543 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5544 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5546 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5548 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5549 when --with-gxx-include-dir is also specified.
5550 * configure: Regenerate.
5552 2018-05-09 Jim Wilson <jimw@sifive.com>
5555 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5556 * config/riscv/t-withmultilib: New.
5557 * config/riscv/withmultilib.h: New.
5558 * doc/install.texi: Document RISC-V --with-multilib-list support.
5560 2018-05-09 Richard Biener <rguenther@suse.de>
5562 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5564 (vect_bb_vectorization_profitable_p): Adjust. Compute
5565 actual scalar cost using the cost vector and the add_stmt_cost
5568 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5570 PR rtl-optimization/85645
5571 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5572 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5574 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5576 PR rtl-optimization/85645
5577 * shrink-wrap.c (spread_components): Return a boolean saying if
5578 anything was changed.
5579 (try_shrink_wrapping_separate): Iterate spread_components until
5580 nothing changes anymore.
5582 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5584 PR rtl-optimization/85645
5585 * regrename.c (build_def_use): Also kill the chains that include the
5586 destination of a REG_CFA_REGISTER note.
5588 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5590 PR rtl-optimization/85645
5591 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5592 insn that has a REG_CFA_REGISTER note.
5594 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5596 * cfgexpand.c (expand_clobber): New function.
5597 (expand_gimple_stmt_1): Use it.
5598 * tree-vect-stmts.c (vect_clobber_variable): New function,
5600 (vectorizable_simd_clone_call): ...here.
5601 (vectorizable_store): Emit a clobber either side of an
5602 IFN_STORE_LANES sequence.
5603 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5605 2018-05-09 Tom de Vries <tom@codesourcery.com>
5608 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5609 (define_insn "trap_if_false"): Add exit after trap.
5611 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5613 PR rtl-optimization/85638
5614 * bb-reorder.c: Include common/common-target.h.
5615 (create_forwarder_block): New function extracted from...
5616 (fix_up_crossing_landing_pad): ...here. Rename into...
5617 (dw2_fix_up_crossing_landing_pad): ...this.
5618 (sjlj_fix_up_crossing_landing_pad): New function.
5619 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5620 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5621 from both partitions and exit the loop after one iteration.
5623 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5626 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5628 (Basic PowerPC Built-in Functions): The new name of the
5629 subsection previously known as "PowerPC Built-in Functions".
5630 (Basic PowerPC Built-in Functions Available on all Configurations):
5632 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5634 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5636 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5638 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5641 2018-05-08 Jim Wilson <jimw@sifive.com>
5643 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5644 (LD_EMUL_SUFFIX): New.
5645 (LINK_SPEC): Use it.
5647 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5649 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5651 (Basic PowerPC Built-in Functions): The new name of the
5652 subsection previously known as "PowerPC Built-in Functions".
5653 (Basic PowerPC Built-in Functions Available on all Configurations):
5655 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5657 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5659 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5661 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5664 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5667 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5668 after cmpelim optimization.
5670 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5672 * config.gcc: Support "goldmont".
5673 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5674 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5676 * config/i386/i386.c (m_GOLDMONT): Define.
5677 (processor_target_table): Add "goldmont".
5678 (PTA_GOLDMONT): Define.
5679 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5680 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5681 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5682 (fold_builtin_cpu): Add "goldmont".
5683 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5684 (ix86_option_override_internal): Add "goldmont".
5685 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5686 (processor_type): Add PROCESSOR_GOLDMONT.
5687 * config/i386/i386.md: Add CPU "glm".
5688 * config/i386/glm.md: New file.
5689 * config/i386/x86-tune.def: Add m_GOLDMONT.
5690 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5692 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5695 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5697 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5698 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5699 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5702 * config/i386/i386.c (ix86_fold_builtin): Handle
5703 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5706 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5707 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5709 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5712 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5713 (check_arch): Likewise.
5714 (check_fpu): Return the result rather than printing it.
5715 (end arch): Fix operator precedence.
5716 (end cpu): Likewise.
5717 (END): Print the result from check_fpu.
5719 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5720 Alan Hayward <alan.hayward@arm.com>
5721 David Sherwood <david.sherwood@arm.com>
5723 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5724 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5725 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5726 (*fcmuo<mode>_and): New patterns.
5728 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5730 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5731 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5732 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5733 (cmp_op, sve_imm_con): New code attributes.
5734 (SVE_COND_INT_CMP, imm_con): Delete.
5735 (cmp_op): Remove above unspecs from int attribute.
5736 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5738 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5739 comparison-specific unspecs.
5740 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5741 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5742 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5743 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5744 (*vec_fcm<cmp_op><mode>): Rename to...
5745 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5746 (*vec_fcmuo<mode>): Rename to...
5747 (*fcmuo<mode>): ...this and adjust likewise.
5748 (*pred_fcm<cmp_op><mode>): New pattern.
5749 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5750 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5752 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5754 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5755 (aarch64_emit_sve_predicated_cond): New function.
5756 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5757 (aarch64_emit_unspec_cond_or): Replace with...
5758 (aarch64_emit_sve_or_conds): ...this new function. Use
5759 aarch64_emit_sve_ptrue_op for the individual comparisons and
5760 aarch64_emit_binop to OR them together.
5761 (aarch64_emit_inverted_unspec_cond): Replace with...
5762 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5763 aarch64_emit_sve_ptrue_op for the comparison and
5764 aarch64_emit_unop to invert the result.
5765 (aarch64_expand_sve_vec_cmp_float): Update after the above
5766 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5768 2018-05-07 Nathan Sidwell <nathan@acm.org>
5770 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5771 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5772 (Backwards Compatibility): Likewise.
5774 2018-05-07 Luis Machado <luis.machado@linaro.org>
5778 2018-05-07 Luis Machado <luis.machado@linaro.org>
5780 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5781 <prefetch_dynamic_strides>: New const bool field.
5782 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5783 prefetch_dynamic_strides.
5784 (exynosm1_prefetch_tune): Likewise.
5785 (thunderxt88_prefetch_tune): Likewise.
5786 (thunderx_prefetch_tune): Likewise.
5787 (thunderx2t99_prefetch_tune): Likewise.
5788 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5790 (aarch64_override_options_internal): Update to set
5791 PARAM_PREFETCH_DYNAMIC_STRIDES.
5792 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5793 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5794 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5795 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5796 prefetch-dynamic-strides setting.
5798 2018-05-07 Luis Machado <luis.machado@linaro.org>
5800 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5801 <minimum_stride>: New const int field.
5802 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5803 minimum_stride field.
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 minimum_stride to 2048.
5809 (aarch64_override_options_internal): Update to set
5810 PARAM_PREFETCH_MINIMUM_STRIDE.
5811 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5812 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5813 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5814 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5815 stride is constant and is below the minimum stride threshold.
5817 2018-05-07 Luis Machado <luis.machado@linaro.org>
5819 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5822 2018-05-07 Luis Machado <luis.machado@linaro.org>
5824 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5825 <prefetch_dynamic_strides>: New const bool field.
5826 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5827 prefetch_dynamic_strides.
5828 (exynosm1_prefetch_tune): Likewise.
5829 (thunderxt88_prefetch_tune): Likewise.
5830 (thunderx_prefetch_tune): Likewise.
5831 (thunderx2t99_prefetch_tune): Likewise.
5832 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5834 (aarch64_override_options_internal): Update to set
5835 PARAM_PREFETCH_DYNAMIC_STRIDES.
5836 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5837 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5838 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5839 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5840 prefetch-dynamic-strides setting.
5842 2018-05-07 Luis Machado <luis.machado@linaro.org>
5844 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5845 <minimum_stride>: New const int field.
5846 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5847 minimum_stride field.
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 minimum_stride to 2048.
5853 (aarch64_override_options_internal): Update to set
5854 PARAM_PREFETCH_MINIMUM_STRIDE.
5855 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5856 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5857 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5858 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5859 stride is constant and is below the minimum stride threshold.
5861 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5864 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5865 the type is addressable. Don't force op into register if it has
5868 2018-05-05 Roland McGrath <mcgrathr@google.com>
5871 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5872 any section for which we don't know a specific type it should have,
5873 regardless of name. Previously this was done only for the exact
5874 names ".init_array", ".fini_array", and ".preinit_array".
5875 (default_elf_asm_named_section): Add comment about
5876 relationship with default_section_type_flags and SECTION_NOTYPE.
5877 (get_section): Don't consider it a type conflict if one side has
5878 SECTION_NOTYPE and the other doesn't, as long as neither has the
5879 SECTION_BSS et al used in the default_section_type_flags logic.
5881 2018-05-05 Tom de Vries <tom@codesourcery.com>
5884 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5885 (workaround_barsyncs): New function.
5886 (nvptx_reorg): Use workaround_barsyncs.
5887 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5888 (define_expand "nvptx_membar_cta"): New define_expand.
5889 (define_insn "*nvptx_membar_cta"): New insn.
5891 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5893 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5894 To improve optimization opportunities.
5895 * builtin-types.def: The new needed builtin types for the above.
5897 2018-05-04 Richard Biener <rguenther@suse.de>
5899 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5900 * gimple-ssa-store-merging.c
5901 (imm_store_chain_info::output_merged_store): Remove redundant create,
5902 release split_store vector contents on failure.
5903 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5904 scalar stmt vector on cache hit.
5906 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5908 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5910 * config.gcc (powerpc-xilinx-eabi*): Remove.
5911 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5913 (fusion_addis_mem_combo_load): Ditto.
5914 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5916 (rs6000_cpu_cpp_builtins): Ditto.
5917 * config/rs6000/rs6000-linux.c
5918 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5919 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5920 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5922 (rs6000_setup_reg_addr_masks): Ditto.
5923 (rs6000_init_hard_regno_mode_ok): Ditto.
5924 (rs6000_option_override_internal): Ditto.
5925 (legitimate_lo_sum_address_p): Ditto.
5926 (rs6000_legitimize_address): Ditto.
5927 (rs6000_legitimize_reload_address): Ditto.
5928 (rs6000_legitimate_address_p): Ditto.
5929 (abi_v4_pass_in_fpr): Ditto.
5930 (setup_incoming_varargs): Ditto.
5931 (rs6000_gimplify_va_arg): Ditto.
5932 (rs6000_split_multireg_move): Ditto.
5933 (rs6000_savres_strategy): Ditto.
5934 (rs6000_emit_prologue_components): Ditto.
5935 (rs6000_emit_epilogue_components): Ditto.
5936 (rs6000_emit_prologue): Ditto.
5937 (rs6000_emit_epilogue): Ditto.
5938 (rs6000_elf_file_end): Ditto.
5939 (rs6000_function_value): Ditto.
5940 (rs6000_libcall_value): Ditto.
5941 * config/rs6000/rs6000.h: Ditto.
5942 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5943 (TARGET_MINMAX): ... this. New.
5944 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5945 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5946 (*movsi_internal1_single): Delete.
5947 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5948 mfpu=, mxilinx-fpu): Delete.
5949 * config/rs6000/singlefp.h: Delete.
5950 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5951 * config/rs6000/t-rs6000: Ditto.
5952 * config/rs6000/t-xilinx: Delete.
5953 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5954 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5955 (VStype_simple): Delete.
5956 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5957 * config/rs6000/xfpu.h: Delete.
5958 * config/rs6000/xfpu.md: Delete.
5959 * config/rs6000/xilinx.h: Delete.
5960 * config/rs6000/xilinx.opt: Delete.
5961 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5962 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5964 2018-05-04 Tom de Vries <tom@codesourcery.com>
5967 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5970 2018-05-04 Richard Biener <rguenther@suse.de>
5973 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5974 (expand_complex_div_wide): Likewise.
5975 (expand_complex_operations_1): Likewise.
5976 (expand_complex_libcall): Preserve EH info of the original stmt.
5977 (tree_lower_complex): Handle removed blocks.
5978 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5979 on complex multiplication and division libcall builtins.
5981 2018-05-04 Richard Biener <rguenther@suse.de>
5984 * fold-const.c (negate_expr_p): Restrict negation of operand
5985 zero of a division to when we know that can happen without
5987 (fold_negate_expr_1): Likewise.
5989 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5992 * real.h (real_nextafter): Declare.
5993 * real.c (real_nextafter): New function.
5994 * fold-const-call.c (fold_const_nextafter): New function.
5995 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5996 CASE_CFN_NEXTTOWARD.
5997 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5998 even when arg1_mode is different from arg0_mode.
6000 2018-05-03 Nathan Sidwell <nathan@acm.org>
6002 * doc/extend.texi (Deprecated Features): Remove
6004 (Backwards Compatibility): Likewise.
6005 * doc/invoke.texi (C++ Language Options): Likewise.
6006 (C++ Dialect Options): Likewise.
6008 2018-05-03 Jakub Jelinek <jakub@redhat.com>
6011 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
6012 _mm512_mask_mullox_epi64): New intrinsics.
6014 2018-05-03 Tom de Vries <tom@codesourcery.com>
6017 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6018 dump files): Add offload-tree.
6020 2018-05-03 Richard Biener <rguenther@suse.de>
6022 PR tree-optimization/85615
6023 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
6024 to loops not nested in BBs loop father to avoid creating multi-entry
6027 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6029 PR tree-optimization/70291
6030 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
6031 arguments. Change return type to tree. Emit libcall as a new
6032 statement rather than replacing existing one when inplace_p is true.
6033 (expand_complex_multiplication_components): New function.
6034 (expand_complex_multiplication): Expand floating-point complex
6035 multiplication using the above.
6036 (expand_complex_division): Rename inner_type parameter to type.
6037 Update expand_complex_libcall call-site.
6038 (expand_complex_operations_1): Update expand_complex_multiplication
6039 and expand_complex_division call-sites.
6041 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6044 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6045 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6046 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
6047 the highest significant bit of the shift count mask is clear. In
6048 check whether and[sq]i3 is needed verify that all significant bits
6049 of the shift count other than the highest are set.
6051 2018-05-02 Tom de Vries <tom@codesourcery.com>
6054 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
6055 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
6056 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
6057 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
6058 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
6059 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
6060 __builtin_goacc_parlevel_size.
6062 2018-05-02 Richard Biener <rguenther@suse.de>
6064 PR tree-optimization/85597
6065 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
6066 do not use split vect_get_vec_defs call but call vect_get_slp_defs
6069 2018-05-02 Tom de Vries <tom@codesourcery.com>
6072 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6073 dump files): Add ltrans-tree.
6075 2018-05-02 Tom de Vries <tom@codesourcery.com>
6078 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6079 dump files): Add wpa-ipa.
6081 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
6083 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
6084 powerpc*-*-linux*paired* target.
6085 * config/rs6000/750cl.h: Delete.
6086 * config/rs6000/paired.h: Delete.
6087 * config/rs6000/paired.md: Delete.
6088 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
6090 * config/rs6000/rs6000-builtin.def: Remove paired float support.
6091 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
6092 comment. Remove paired float support.
6093 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
6094 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6096 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
6097 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
6099 * config/rs6000/rs6000.c: Remove paired float support.
6100 (paired_expand_vector_init, paired_expand_vector_move,
6101 paired_emit_vector_compare, paired_emit_vector_cond_expr,
6102 (paired_expand_lv_builtin, paired_expand_stv_builtin,
6103 paired_expand_builtin, paired_expand_predicate_builtin,
6104 paired_init_builtins): Delete.
6105 * config/rs6000/rs6000.h: Remove paired float support.
6106 * config/rs6000/rs6000.md: Remove paired float support.
6107 (move_from_CR_ov_bit): Delete.
6108 * config/rs6000/rs6000.opt (mpaired): Delete.
6109 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
6110 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
6112 2018-05-02 Richard Biener <rguenther@suse.de>
6115 * gimplify.c (gimplify_save_expr): When in SSA form allow
6116 SAVE_EXPRs to compute to SSA vars.
6118 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6121 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6122 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6123 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
6124 clobber operands[2], instead use a new pseudo. Formatting fixes.
6126 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
6128 PR tree-optimization/85586
6129 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
6130 exit early for statements in the same group if the accesses are
6133 2018-05-02 Tom de Vries <tom@codesourcery.com>
6136 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
6139 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
6141 PR tree-optimization/85143
6142 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
6144 2018-05-01 Tom de Vries <tom@codesourcery.com>
6147 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
6148 not found" error message.
6150 2018-05-01 Tom de Vries <tom@codesourcery.com>
6153 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
6154 * vec.c (test_ordered_remove_if): New function.
6155 (vec_c_tests): Call test_ordered_remove_if.
6156 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
6157 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
6158 * tree-vect-patterns.c (vect_pattern_recog_1): Use
6159 VEC_ORDERED_REMOVE_IF.
6161 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6163 PR tree-optimization/82665
6164 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
6165 pointer subtraction where arguments come from a memchr call.
6167 2018-05-01 Jakub Jelinek <jakub@redhat.com>
6169 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
6170 --push-state --as-needed and --pop-state instead of --as-needed and
6171 --no-as-needed if ld supports it.
6172 * configure: Regenerated.
6175 * doc/install.texi2html: Replace _002d with - and _002a with * in
6176 generated html files using sed.
6178 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6181 * gcc-rich-location.c (blank_line_before_p): New function.
6182 (use_new_line): New function.
6183 (gcc_rich_location::add_fixit_insert_formatted): New function.
6184 * gcc-rich-location.h
6185 (gcc_rich_location::add_fixit_insert_formatted): New function.
6187 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6189 * selftest.c (assert_streq): Rename "expected" and "actual" to
6190 "val1" and "val2". Extend NULL-handling to cover both inputs
6191 symmetrically, while still requiring both to be non-NULL for a pass.
6192 * selftest.h (assert_streq): Rename "expected" and "actual" to
6194 (ASSERT_EQ): Likewise.
6195 (ASSERT_EQ_AT): Likewise.
6196 (ASSERT_KNOWN_EQ): Likewise.
6197 (ASSERT_KNOWN_EQ_AT): Likewise.
6198 (ASSERT_NE): Likewise.
6199 (ASSERT_MAYBE_NE): Likewise.
6200 (ASSERT_MAYBE_NE_AT): Likewise.
6201 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
6202 the assertion to pass.
6203 (ASSERT_STREQ_AT): Likewise.
6205 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6207 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
6208 interaction with -pie.
6210 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6212 * selftest.h: Fix alphabetization of per-source-file selftest
6215 2018-04-30 Jason Merrill <jason@redhat.com>
6217 PR c++/61982 - dead stores to destroyed objects.
6218 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
6221 2018-04-30 Jason Merrill <jason@redhat.com>
6223 * tree.c (build_clobber): New.
6224 * tree.h: Declare it.
6225 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
6227 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6229 * diagnostic-show-locus.c (layout::layout): Update for
6230 location_get_source_line returning a char_span.
6231 (struct char_span): Move to input.h.
6232 (struct correction): Update for fields in char_span becoming
6234 (struct source_line): Update for location_get_source_line
6235 returning a char_span.
6236 (layout::print_line): Likewise.
6237 * edit-context.c (edited_file::print_content): Likewise.
6238 (edited_file::print_diff_hunk): Likewise.
6239 (edited_file::print_run_of_changed_lines): Likewise.
6240 (edited_file::get_num_lines): Likewise.
6241 (edited_line::edited_line): Likewise.
6242 * final.c (asm_show_source): Likewise.
6243 * input.c (location_get_source_line): Convert return type
6244 from const char * to char_span, losing the final "line_len"
6246 (dump_location_info): Update for the above.
6247 (get_substring_ranges_for_loc): Likewise. Use a char_span
6248 when handling the literal within the line.
6249 (test_reading_source_line): Update for location_get_source_line
6250 returning a char_span.
6251 * input.h (class char_span): Move here from
6252 diagnostic-show-locus.c, converting from a struct to a class.
6253 Make data members private.
6254 (char_span::operator bool): New.
6255 (char_span::length): New.
6256 (char_span::get_buffer): New.
6257 (char_span::operator[]): New.
6258 (char_span::subspan): Make const.
6259 (char_span::xstrdup): New.
6260 (location_get_source_line): Convert return type from const char *
6261 to char_span, losing the final "line_size" param.
6263 2018-04-30 Jan Hubicka <jh@suse.cz>
6265 * lto-wrapper.c (ltrans_priorities): New static var.
6266 (cmp_priority): New.
6267 (run_gcc): Read priorities and if doing parallel build order
6268 the Makefile by them.
6270 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6272 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
6274 2018-04-30 Richard Biener <rguenther@suse.de>
6276 * tree-cfg.c (verify_address): Remove base argument, add
6277 flag whether to check TREE_ADDRESSABLE and do that.
6278 (verify_expr): Remove.
6279 (verify_types_in_gimple_reference): Add pieces from verify_expr.
6280 (verify_gimple_assign_single): Likewise.
6281 (verify_gimple_switch): Likewise.
6282 (verify_expr_location_1): Dereference tp once. Add (disabled)
6283 piece from verify_expr.
6284 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
6286 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6288 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
6290 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6292 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
6293 (small_data_pattern): Likewise.
6294 (arc_rewrite_small_data): Likewise.
6295 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
6296 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
6297 (get_symbol_alignment): New function.
6298 (legitimate_small_data_address_p): Likewise.
6299 (legitimate_scaled_address): Update, call
6300 legitimate_small_data_address_p.
6301 (output_sdata): New static variable.
6302 (arc_print_operand): Update how we handle small data operands.
6303 (arc_print_operand_address): Likewise.
6304 (arc_legitimate_address_p): Update, use
6305 legitimate_small_data_address_p.
6306 (arc_rewrite_small_data_p): Remove.
6307 (arc_rewrite_small_data_1): Likewise.
6308 (arc_rewrite_small_data): Likewise.
6309 (small_data_pattern): Likewise.
6310 (compact_sda_memory_operand): Update to use
6311 legitimate_small_data_address_p and get_symbol_alignment.
6312 (prepare_move_operands): Don't rewite sdata pattern.
6313 (prepare_extend_operands): Remove.
6314 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
6316 (zero_extendqisi2): Likewise.
6317 (zero_extendhisi2): Likewise.
6318 (extendqihi2): Likewise.
6319 (extendqisi2): Likewise.
6320 (extendhisi2): Likewise.
6324 * config/arc/constraints.md (Usd): Change it to memory constraint.
6326 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6328 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
6329 as source of std instructions.
6330 * config/arc/arc.md (movsi_insn): Update pattern predicate to
6331 allow 6-bit constants as source for store instructions.
6332 (movdi_insn): Update instruction pattern to allow 6-bit constants
6333 as source for store instructions.
6335 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6337 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
6339 2018-04-30 Nathan Sidwell <nathan@acm.org>
6340 Sandra Loosemore <sandra@codesourcery.com>
6342 * dumpfile.c (dump_open): Allow '-' for stdout.
6343 * doc/invoke.texi (Developer Options): Document dump filename
6344 determination early. Document stdin/stdout selection.
6346 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
6348 Microblaze Target: PIC data text relative
6350 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
6351 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
6353 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
6354 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
6355 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
6356 New addressing mode for data-text relative position indepenedent code.
6357 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
6358 'ADDRESS_SYMBOLIC_TXT_REL'.
6359 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
6360 (microblaze_legitimate_pic_operand): Exclude function calls from
6361 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
6362 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
6364 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6365 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6366 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
6367 for 'address + offset'.
6368 (microblaze_expand_prologue): Add new function prologue call for
6370 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
6371 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
6372 table in case of TARGET_PIC_DATA_TEXT_REL.
6373 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
6374 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
6375 Add new macros 'UNSPEC_TEXT',
6376 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
6377 + exclude function calls from 'UNSPEC_PLT' in case of data text
6379 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6380 new target hook for generating address diff vector tables in case of
6382 * doc/tm.texi : Regenerate.
6383 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
6384 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
6385 of addr diff vector generation.
6386 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6387 target hook definition.
6388 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
6389 Add default function for generate_pic_addr_diff_vec -> flag_pic.
6390 * doc/invoke.texi (Add new pic option): Add new microblaze pic
6391 option for data text relative.
6393 2018-04-30 Richard Biener <rguenther@suse.de>
6395 * tree-chrec.h (evolution_function_is_constant_p): Remove
6397 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
6399 2018-04-30 Richard Biener <rguenther@suse.de>
6402 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
6404 2018-04-30 Richard Biener <rguenther@suse.de>
6406 PR tree-optimization/28364
6407 PR tree-optimization/85275
6408 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
6409 copying first exit test.
6411 2018-04-28 Mark Wielaard <mark@klomp.org>
6413 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
6415 (dwarf_AT): Handle DW_AT_addr_base.
6416 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
6418 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
6421 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
6422 (*ashl<dwi>3_doubleword_mask_1): Ditto.
6423 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
6424 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
6426 2018-04-28 Richard Biener <rguenther@suse.de>
6428 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
6429 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
6430 to reflect use. Only add interesting stmts.
6432 2018-04-27 Martin Jambor <mjambor@suse.cz>
6435 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
6436 the jump function allows for passing through aggregate values.
6438 2018-04-27 David Malcolm <dmalcolm@redhat.com>
6440 * input.h (in_system_header_at): Convert from macro to inline
6442 (from_macro_expansion_at): Likewise.
6443 (from_macro_definition_at): Likewise.
6445 2018-04-27 Jeff Law <law@redhat.com>
6447 * config.gcc: Mark tile* targets as deprecated/obsolete.
6449 2018-04-27 Richard Biener <rguenther@suse.de>
6451 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
6454 2018-04-27 Richard Biener <rguenther@suse.de>
6456 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
6458 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
6460 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
6461 with Yd constraint. Set "preferred_for_speed" attribute from
6462 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
6464 (*movdi_internal): Ditto.
6465 (movti_interunit splitters): Remove
6466 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
6467 (movdi_interunit splitters): Ditto.
6468 * config/i386/constraints.md (Ye): Remove.
6469 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
6471 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6474 * config/aarch64/constraints.md (Usg): Limit to 31.
6477 2018-04-27 Jakub Jelinek <jakub@redhat.com>
6479 PR tree-optimization/85529
6480 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
6481 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
6482 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
6483 zero extension or masking of the MSB bit.
6484 (optimize_range_tests): Add FIRST_BB argument, pass it through
6485 to optimize_range_tests_var_bound.
6486 (maybe_optimize_range_tests, reassociate_bb): Adjust
6487 optimize_range_tests callers.
6489 2018-04-26 Richard Biener <rguenther@suse.de>
6490 Jakub Jelinek <jakub@redhat.com>
6492 * cgraph.h (symbol_table): Just declare debug method here.
6493 * symtab.c (symbol_table::debug): Define.
6495 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
6497 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
6499 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6501 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6502 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6503 (*movdi_internal): Substitute Yi and Yj constraint with x
6504 and Ym and Yn constraint with y constraint. Update "isa"
6505 attribute and set "preferred_for_speed" attribute from
6506 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6507 (*movsi_internal): Ditto.
6508 (*movdf_internal): Ditto.
6509 (*movsf_internal): Ditto.
6510 (*zero_extendsidi2): Ditto.
6511 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6512 (sse2_loadld): Ditto.
6513 (*vec_extract<ssevecmodelower>_0): Ditto.
6514 (*vec_extractv4si_0_zext_sse4): Ditto.
6515 (vec_concatv2di): Ditto.
6516 (*vec_dup<mode>): Ditto.
6517 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
6518 * config/i386/constraints.md (Yi): Remove.
6523 2018-04-26 Nathan Sidwell <nathan@acm.org>
6525 * dumpfile.c (dump_open): New.
6526 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
6527 (dump_finish): Detect stdio/stderr by value not name.
6529 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6531 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6533 2018-04-26 Tom de Vries <tom@codesourcery.com>
6536 * config/nvptx/nvptx.c (verify_neutering_jumps)
6537 (verify_neutering_labels): New function
6538 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6540 2018-04-26 Tom de Vries <tom@codesourcery.com>
6543 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6544 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6547 2018-04-26 Richard Biener <rguenther@suse.de>
6548 Tom de Vries <tom@codesourcery.com>
6551 * lto-streamer-out.c (output_function): Fixup loops if required to match
6552 discovery done in the reader.
6554 2018-04-26 Richard Biener <rguenther@suse.de>
6556 PR tree-optimization/85116
6557 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6558 have a loop exit from the single latch predecessor. Remove
6559 case of header with just condition.
6560 (ch_base::copy_headers): Exclude infinite loops from any
6562 (pass_ch::execute): Record exits.
6564 2018-04-26 Richard Biener <rguenther@suse.de>
6566 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6567 prologue cost vector and pass it to vect_get_load_cost.
6568 (vect_get_peeling_costs_all_drs): Likewise.
6569 (vect_peeling_hash_get_lowest_cost): Likewise.
6570 (vect_enhance_data_refs_alignment): Likewise.
6572 2018-04-26 Richard Biener <rguenther@suse.de>
6575 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6576 checking of integer<->pointer conversions.
6577 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6578 sign-/zero-extending pointer types.
6579 (expand_omp_for_static_chunk): Likewise.
6581 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6582 Jean Lee <xiaoyur347@gmail.com>
6584 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6585 (TARGET_ASAN_SHADOW_OFFSET): Define.
6586 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6587 true for -fsanitize=address.
6589 2018-04-25 Mark Wielaard <mark@klomp.org>
6591 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6594 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6596 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6597 than "alu", remove explicit "memory" and "imm_disp" attributes.
6598 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6601 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6602 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6603 gen_lowpart_no_emit.
6605 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6608 * config/i386/i386.c (ix86_expand_builtin): Change memory
6609 operand to XI, extend p0 to Pmode.
6610 * config/i386/i386.md: Change unspec volatile and operand
6611 1 mode to XI, change operand 0 mode to P.
6613 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6615 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6616 GET_MODE_MASK before any checking.
6617 (nds32_can_use_bset_p): Likewise.
6618 (nds32_can_use_btgl_p): Likewise.
6620 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6622 * config/nds32/nds32-doubleword.md: New define_split pattern for
6623 illegal register number.
6625 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6627 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6629 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6631 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6633 2018-04-25 Richard Biener <rguenther@suse.de>
6635 * lto-streamer.h (LTO_major_version): Bump to 8.
6637 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6639 * BASE-VER: Set to 9.0.0.
6641 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6643 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6644 in __abskf2 and __powikf2.
6646 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6649 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6650 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6651 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6652 Use the above on operand 2. Reindent.
6653 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6655 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6658 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6660 * config/i386/i386.opt (mcet): Removed.
6661 * doc/install.texi: Remove -mcet documentation.
6662 * doc/invoke.texi: Likewise.
6664 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6667 * doc/install.texi: Remove -mcet from bootstrap-cet.
6669 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6672 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6673 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6677 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6678 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6679 containing a CONST_VECTOR.
6681 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6683 * doc/install.texi: Update newlib dependency for nvptx.
6685 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6688 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6689 instead of INTVAL when shifting x left.
6691 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6693 PR tree-optimization/85478
6694 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6695 vect_grouped_store_supported for single element vectors.
6697 2018-04-24 Richard Biener <rguenther@suse.de>
6700 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6701 load cost increase to the case of non-constant step.
6703 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6706 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6707 destination if any_malformed_asm.
6709 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6712 * expr.c (store_field): In the bitfield case, if the value comes from
6713 a function call and is returned in registers by means of a PARALLEL,
6714 do not change the mode of the temporary unless BLKmode and VOIDmode.
6716 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6718 PR rtl-optimization/85423
6719 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6720 dependencies to debug insns when the previous insn is non-debug.
6722 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6724 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6725 enums into a single definition.
6726 (fls): Fix predicates and printing.
6729 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6731 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6732 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6733 and short u6 immediate.
6734 (check_if_valid_sleep_operand): Remove.
6735 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6737 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6739 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6740 flag_always_save_lp condition.
6741 * config/nds32/nds32.opt (malways-save-lp): New option.
6743 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6745 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6746 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6747 * config/nds32/nds32.h
6748 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6749 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6751 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6753 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6754 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6756 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6757 Chung-Ju Wu <jasonwucj@gmail.com>
6759 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6761 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6762 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6764 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6766 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6768 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6770 * config/nds32/nds32-protos.h (nds32_data_alignment,
6771 nds32_local_alignment): Declare.
6772 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6773 nds32_local_alignment): New functions.
6774 (TARGET_CONSTANT_ALIGNMENT): Define.
6775 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6777 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6779 * config/nds32/nds32.c
6780 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6781 (TARGET_MODES_TIEABLE_P): Likewise.
6783 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6785 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6788 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6789 Chung-Ju Wu <jasonwucj@gmail.com>
6791 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6792 for unaligned access.
6793 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6795 * config/nds32/nds32-intrinsic.md: Likewise.
6796 * config/nds32/nds32_intrinsic.h: Likewise.
6797 * config/nds32/nds32.h (nds32_builtins): Likewise.
6798 * config/nds32/nds32.opt (munaligned-access): New option.
6799 * config/nds32/nds32.c (nds32_asm_file_start): Display
6800 flag_unaligned_access status.
6802 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6804 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6805 -mno-relax is present.
6806 * config/riscv/linux.h (LINK_SPEC): Ditto.
6808 2018-04-20 Martin Sebor <msebor@redhat.com>
6811 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6813 (gimple_fold_builtin_stxcpy_chk): Same.
6814 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6816 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6819 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6820 __powikf2 when long double is IEEE 128-bit.
6822 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6824 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6825 step to make sure stack always aligned.
6827 2018-04-20 Carl Love <cel@us.ibm.com>
6830 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6831 size check for arg0.
6833 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6834 Tom de Vries <tom@codesourcery.com>
6837 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6838 Emit insns for calls too.
6839 (nvptx_find_par): Always look for worker-level predecessor insn.
6840 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6842 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6843 (nvptx_process_pars): Propagate frames for calls.
6845 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6848 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6850 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6851 (ix86_handle_option): Don't handle OPT_mibt.
6852 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6854 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6856 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6857 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6858 (ix86_target_macros): Define __CET__ with flag_cf_protection
6859 for -fcf-protection.
6860 * config/i386/i386.c (isa2_opts): Remove -mibt.
6861 * config/i386/i386.h (TARGET_IBT): Removed.
6862 (TARGET_IBT_P): Likewise.
6863 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6864 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6865 * config/i386/i386.opt (mcet): Update help message.
6868 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6869 -mcet as an alias for -mshstk.
6871 2018-04-20 Richard Biener <rguenther@suse.de>
6874 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6875 complexity by forcing a single use of the multiply operand.
6877 2018-04-20 Martin Jambor <mjambor@suse.cz>
6880 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6881 recursion dependency to only apply to non-clones.
6883 2018-04-20 Martin Jambor <mjambor@suse.cz>
6886 * ipa-cp.c (create_specialized_node): Check that clones of
6887 self-recursive edges exist during IPA-CP.
6889 2018-04-19 Toon Moene <toon@moene.org>
6891 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6894 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6896 PR tree-optimization/85467
6897 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6898 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6899 VECTOR_CST element to type.
6901 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6904 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6905 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6906 (builtin_longjmp): Likewise.
6907 (save_stack_nonlocal): New pattern.
6908 (restore_stack_nonlocal): Likewise.
6910 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6913 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6914 Replace ASM_OUTPUT_LABEL with fprintf.
6916 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6919 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6920 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6921 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6922 define __IBT__ and __SHSTK__ for -fcf-protection.
6923 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6925 (ix86_trampoline_init): Likewise.
6926 (x86_output_mi_thunk): Likewise.
6927 (ix86_notrack_prefixed_insn_p): Likewise.
6928 (ix86_option_override_internal): Don't disallow -fcf-protection.
6929 * config/i386/i386.md (rdssp<mode>): Also enable for
6931 (incssp<mode>): Likewise.
6932 (nop_endbr): Likewise.
6933 * config/i386/i386.opt (mcet): Change help message to built-in
6937 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6938 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6939 enable CET built-in functions.
6941 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6943 * common/config/i386/i386-common.c
6944 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6945 OPTION_MASK_ISA_MOVDIRI_UNSET,
6946 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6947 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6948 * config.gcc (movdirintrin.h): New header.
6949 * config/i386/cpuid.h (bit_MOVDIRI,
6950 bit_MOVDIR64B): New bits.
6951 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6953 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6954 (VOID, PVOID, PCVOID)): New function types.
6955 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6956 __builtin_ia32_directstoreu_u64,
6957 __builtin_ia32_movdir64b): New builtins.
6958 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6959 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6961 (ix86_valid_target_attribute_inner_p): Ditto.
6962 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6963 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6964 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6965 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6966 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6967 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6968 (movdiri<mode>, movdir64b_<mode>): New.
6969 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6970 * config/i386/immintrin.h: Include movdirintrin.h.
6971 * config/i386/movdirintrin.h: New file.
6972 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6974 2018-04-19 Richard Biener <rguenther@suse.de>
6977 * cfg.c (clear_bb_flags): When loop state says we have
6978 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6980 2018-04-19 Richard Biener <rguenther@suse.de>
6982 PR tree-optimization/84737
6983 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6984 copying restrict info.
6985 (vect_setup_realignment): Use it.
6986 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6987 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6988 the first DR to all generated stores.
6989 (vectorizable_load): Likewise for loads.
6991 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6993 PR tree-optimization/85446
6994 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6995 the integral and pointer types to have the same precision.
6997 * doc/install.texi: Document --disable-cet being the default and
7000 2018-04-18 Martin Liska <mliska@suse.cz>
7002 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
7005 2018-04-18 Martin Liska <mliska@suse.cz>
7008 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7011 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7012 arguments if they are comparable.
7014 2018-04-18 Martin Liska <mliska@suse.cz>
7017 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7020 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7023 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
7026 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
7027 ENDBR after calling __morestack.
7029 2018-04-18 David Malcolm <dmalcolm@redhat.com>
7032 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
7033 by using gcc_base_ver to generate a gcc_driver_version, and use
7034 it when generating GCC_DRIVER_NAME.
7035 * configure: Regenerate.
7037 2018-04-18 Jakub Jelinek <jakub@redhat.com>
7040 * config.gcc: Obsolete powerpc*-*-*spe*.
7042 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7045 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
7046 (stabstr_D): Change type of unum from unsigned int to
7047 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
7050 2018-04-17 Jim Wilson <jimw@sifive.com>
7053 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
7054 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
7055 Set arg_pointer_offset after using pretend_args_size.
7057 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7059 PR rtl-optimization/85431
7060 * dse.c (record_store): Ignore zero width stores.
7063 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
7064 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
7065 __builtin_stack_restore rather than after it.
7066 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
7067 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
7068 argument instead of virtual_dynamic_stack_rtx.
7070 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
7072 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
7074 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
7075 Add note to error message to explain internal mapping of overloaded
7076 built-in function name to non-overloaded built-in function name.
7077 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
7080 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7083 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
7084 where the inputs overlap with the output.
7086 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7088 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
7089 (=v, v) alternative and explicit "memory" attribute.
7090 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
7091 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7093 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
7094 "sselog1" type instead of "sselog".
7095 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
7096 "sselog". Remove explicit "memory" attribute.
7097 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
7098 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7100 (vec_extract_hi_v32hi): Merge all alternatives into one, use
7101 "sselog1" type instead of "sselog". Remove explicit "memory"
7103 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
7104 use "sselog1" type instead of "sselog". Remove explicit "memory"
7106 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
7107 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7109 (vec_extract_hi_v64qi): Merge all alternatives into one, use
7110 "sselog1" type instead of "sselog". Remove explicit "memory"
7112 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
7113 use "sselog1" type instead of "sselog". Remove explicit "memory"
7117 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
7120 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
7123 2018-04-17 Martin Jambor <mjambor@suse.cz>
7126 * ipa-cp.c (create_specialized_node): Call
7127 expand_all_artificial_thunks if necessary.
7129 2018-04-17 Martin Liska <mliska@suse.cz>
7132 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
7133 in message, remote space in between '_G' and '('.
7135 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7138 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
7139 avx512f_vmcmp<mode>3<round_saeonly_name>,
7140 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
7141 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
7142 avx512f_rndscale<mode><round_saeonly_name>,
7143 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
7144 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7145 Use %<iptr>2 instead of %2 for -masm=intel.
7146 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
7147 avx512f_vcvttss2usi<round_saeonly_name>,
7148 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
7150 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
7151 avx512f_vcvttsd2usi<round_saeonly_name>,
7152 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
7153 Use %q1 instead of %1 for -masm=intel.
7154 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
7155 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
7156 of %3 for -masm=intel.
7157 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
7159 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
7161 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
7163 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
7164 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
7166 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
7168 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
7169 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
7170 %g1 and one with %0 and %1.
7171 (avx512er_vmrcp28<mode><round_saeonly_name>,
7172 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
7174 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
7175 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
7176 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
7177 of %0 and %{%4%} for -masm=intel.
7178 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
7179 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
7180 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
7181 order of %0 and %{%5%}%{z%} for -masm=intel.
7183 2018-04-17 Jan Hubicka <jh@suse.cz>
7186 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
7188 2018-04-17 Martin Liska <mliska@suse.cz>
7191 * multiple_target.c (create_dispatcher_calls): Set apostrophes
7192 for target_clone error message. Make default implementation
7193 clone to be a local declaration.
7194 (separate_attrs): Add new argument and check for an empty
7196 (expand_target_clones): Handle it.
7197 (ipa_target_clone): Make redirection just for target_clones
7200 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
7201 Tom de Vries <tom@codesourcery.com>
7204 * omp-expand.c (expand_oacc_for): Add dummy false branch for
7205 tiled basic blocks without omp continue statements.
7207 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
7210 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
7211 vec_extract expression as having side effects to make sure it gets
7214 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
7217 * config/i386/i386.c (get_builtin_code_for_version): Check
7220 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
7223 * gcc/config.gcc: Support "skylake".
7224 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
7226 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
7227 (processor_target_table): Add "skylake".
7228 (ix86_option_override_internal): Add "skylake".
7229 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
7230 PROCESSOR_CANNONLAKE.
7231 (get_builtin_code_for_version): Fix priority for
7232 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
7233 PROCESSOR_SKYLAKE-AVX512.
7234 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
7235 (processor_type): Add PROCESSOR_SKYLAKE.
7237 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
7238 Jason Merrill <jason@redhat.com>
7241 * convert.c (convert_to_integer_1): Use direct recursion for
7242 enumeral types and types with a precision less than the number
7243 of bits in their mode.
7245 2018-04-16 Julia Koval <julia.koval@intel.com>
7248 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
7249 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
7251 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
7254 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
7255 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
7256 and -mno-direct-move.
7258 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
7261 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
7262 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
7263 Ensure negative shifts result in {0}.
7265 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
7267 PR rtl-optimization/79916
7268 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
7269 regs (if any) to define how to gnerate SD moves when LRA is in
7272 2018-04-13 Jakub Jelinek <jakub@redhat.com>
7274 PR rtl-optimization/85393
7275 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
7276 * except.c (expand_dw2_landing_pad_for_region): Make static.
7277 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
7278 a label and unconditional jump to old_bb, rather than
7279 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
7282 PR rtl-optimization/85376
7283 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
7284 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
7285 instead of a specific value.
7287 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7288 Bin Cheng <bin.cheng@arm.com>
7290 PR tree-optimization/82965
7291 PR tree-optimization/83991
7292 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
7293 by_profile_only parameter.
7294 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
7295 information if the loop was predicted to iterate too many times.
7296 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
7298 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7301 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
7304 2018-04-13 Martin Liska <mliska@suse.cz>
7305 Jakub Jelinek <jakub@redhat.com>
7308 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
7309 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
7310 * builtins.c (expand_builtin_memory_copy_args): Use
7311 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
7312 handle dest_addr == pc_rtx.
7314 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
7317 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
7318 asked to not generate direct moves.
7319 (fix_trunc<mode>si2_stfiwx): Similar.
7320 (fix_trunc<mode>si2_internal): Similar.
7322 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7325 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
7326 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
7327 lookup if dest in some wider mode is known to be const0_rtx and
7328 if so, record permanent equivalence for it to be ZERO_EXTEND of
7329 the narrower mode destination.
7331 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7333 * lto-streamer-out.c (output_function): Revert 259346.
7334 * omp-expand.c (expand_oacc_for): Likewise.
7336 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
7338 PR rtl-optimization/85354
7339 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
7340 * sel-sched.c (sel_global_init): ... here.
7342 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
7345 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
7346 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7347 mode for PE-COFF targets.
7348 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
7349 (i386_pe_asm_lto_end): Likewise.
7350 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
7351 (TARGET_ASM_LTO_END): Likewise.
7352 * config/i386/winnt.c (saved_debug_info_level): New static variable.
7353 (i386_pe_asm_lto_start): New function.
7354 (i386_pe_asm_lto_end): Likewise.
7356 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7357 Richard Biener <rguenther@suse.de>
7360 * lto-streamer-out.c (output_function): Fix CFG loop state before
7362 * omp-expand.c (expand_oacc_for): Handle calls to internal
7363 functions like regular functions.
7365 2018-04-12 Richard Biener <rguenther@suse.de>
7368 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
7369 for the early LTO debug to properly generate references to it
7370 during DIE emission. Do not re-use that for the skeleton for
7372 (dwarf2out_early_finish): Likewise.
7374 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7377 * config/i386/sse.md
7378 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
7379 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
7380 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
7381 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
7382 and output is a reg, avoid creating invalid lowpart subreg, but
7383 instead split into a 512-bit move. Don't split if not AVX512VL,
7384 input is xmm16+ reg and output is a mem.
7385 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
7386 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
7387 xmm16+ reg and output is a mem.
7389 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7391 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
7392 also for flag_dwarf2_cfi_asm.
7394 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7396 PR rtl-optimization/85342
7397 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
7398 a bool scalar var inside of the loop instead. Don't try to update
7399 recog_data.operand after failed apply_change_group.
7401 2018-04-12 Tom de Vries <tom@codesourcery.com>
7404 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
7405 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
7406 array with flexible array member as array without given dimension.
7407 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
7408 argument for undefined param to true.
7410 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
7413 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7414 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
7415 from PowerPC section.
7416 * config/rs6000/sysv4.opt (mcall-): Improve help text.
7417 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
7418 help text that is too long.
7419 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
7420 help text that is too long.
7421 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
7422 help text that is too long.
7424 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
7426 * config/alpha/alpha.md (stack_probe_internal): Rename
7427 from "probe_stack". Update all callers.
7429 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7431 PR rtl-optimization/84566
7432 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
7433 sched_macro_fuse_insns.
7435 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7438 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
7439 (compute_block_dependences): ... from here.
7441 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7443 PR tree-optimization/85331
7444 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
7445 from int to HOST_WIDE_INT.
7447 2018-04-11 Martin Jambor <mjambor@suse.cz>
7450 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
7451 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
7452 not the same as the source val.
7453 (cgraph_edge_brings_value_p): New parameter.
7454 (gather_edges_for_value): Pass destination value to
7455 cgraph_edge_brings_value_p.
7456 (perhaps_add_new_callers): Likewise.
7457 (get_info_about_necessary_edges): Likewise and exclude values brought
7458 only by self-recursive edges.
7459 (create_specialized_node): Redirect only clones of self-calling edges.
7460 (+self_recursive_pass_through_p): New function.
7461 (find_more_scalar_values_for_callers_subset): Use it.
7462 (find_aggregate_values_for_callers_subset): Likewise.
7463 (known_aggs_to_agg_replacement_list): Removed.
7464 (decide_whether_version_node): Re-calculate known constants for all
7465 remaining context clones.
7467 2018-04-11 Richard Biener <rguenther@suse.de>
7470 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
7471 from early DWARF output.
7472 (dwarf2out_early_finish): Output line info unconditionally into
7473 early DWARF and add reference to it.
7475 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7478 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
7479 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
7480 other than V2DFmode using iptr mode attribute.
7481 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
7483 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7485 PR rtl-optimization/84659
7486 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
7488 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7491 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
7493 (output_loc_list): Pass address of a dummy size variable even in the
7494 locview handling loop.
7495 (index_location_lists): Add comment on why skip_loc_list_entry can't
7498 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7501 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7504 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7507 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7508 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7509 and -mstring-compare-inline-limit.
7511 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7514 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7515 for stack clash protection in a register whenever we need it to be in
7518 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7520 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
7521 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
7523 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7526 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
7528 (mlong-double-): Ditto.
7529 * config/rs6000/sysv4.opt (msdata=): Ditto.
7530 (mtls-size=): Ditto.
7532 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7534 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7535 erroneous entries for
7536 "vector int vec_ldl (int, long int *)", and
7537 "vector unsigned int vec_ldl (int, unsigned long int *)".
7538 Add comments and entries for
7539 "vector bool char vec_ldl (int, bool char *)",
7540 "vector bool short vec_ldl (int, bool short *)",
7541 "vector bool int vec_ldl (int, bool int *)",
7542 "vector bool long long vec_ldl (int, bool long long *)",
7543 "vector pixel vec_ldl (int, pixel *)",
7544 "vector long long vec_ldl (int, long long *)",
7545 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7546 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7547 type tree bool_long_long_type_node and correct definition of
7548 bool_V2DI_type_node to make reference to this new type tree.
7549 (rs6000_mangle_type): Replace erroneous reference to
7550 bool_long_type_node with bool_long_long_type_node.
7551 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7552 comments to emphasize sign distinctions for char and int types and
7553 replace RS6000_BTI_bool_long constant with
7554 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7555 use of RS6000_BTI_pixel.
7556 (bool_long_type_node): Remove this macro definition.
7557 (bool_long_long_type_node): New macro definition
7559 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7561 PR rtl-optimization/85300
7562 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7563 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7564 simplify_unary_operation fails.
7566 2018-04-10 Martin Liska <mliska@suse.cz>
7568 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7569 cgraph_edge and ipa_ref.
7571 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7575 * config/i386/sse.md
7576 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7577 computation of the VEC_MERGE selector from mask.
7578 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7579 Fix decoding of the VEC_MERGE selector into mask.
7581 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7583 PR tree-optimization/85286
7584 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7586 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7588 * final.c (final_1): Set insn_last_address as well as
7589 insn_current_address.
7591 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7594 * explow.c (emit_stack_probe): Call validize_mem on memory location
7595 before passing it to gen_probe_stack. Create address operand and
7596 legitimize it for the probe_stack_address case.
7598 2018-04-09 Jan Hubicka <jh@suse.cz>
7601 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7602 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7603 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7605 (free_lang_data): Rebuild type inheritance graph.
7607 2018-04-09 Martin Sebor <msebor@redhat.com>
7609 * invoke.texi (-finline-small-functions): Mention other optimization
7611 (-findirect-inlining, -fpartial-inlining): Same.
7612 (-finline-functions-called-once): Same.
7613 (-freorder-blocks-and-partition): Same.
7615 2018-04-09 Jan Hubicka <jh@suse.cz>
7618 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7619 jumps; choose last target that matches the criteria (i.e.
7620 no partition changes for non-crossing jumps).
7621 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7622 support for redirecting crossing jumps to non-crossing.
7624 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7626 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7627 also for naked functions.
7629 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7631 * config/arc/arc.md (add_shift): New pattern.
7632 (add_shift2): Likewise.
7633 (sub_shift): Likewise.
7634 (sub_shift_cmp0_noout): Likewise.
7635 (compare_si_ashiftsi): Likewise.
7636 (xbfu_cmp0_noout): New combine pattern.
7637 (xbfu_cmp0"): Likewise.
7638 (movsi_set_cc_insn): Place the predicable variant first.
7639 (commutative_binary_cmp0_noout): Remove clobber.
7640 (commutative_binary_cmp0): New pattern.
7641 (noncommutative_binary_cmp0): Likewise.
7642 (noncommutative_binary_cmp0_noout): Likewise.
7643 (noncommutative_binary_comparison_result_used): Removed.
7644 (rsub_cmp0): New pattern.
7645 (rsub_cmp0_noout): Likewise.
7646 (extzvsi): Changed, keep only meaningful variants.
7647 (SQH, SEZ): New iterators.
7648 (SQH_postfix): New mode attribute.
7649 (SEZ_prefix): New code attribute.
7650 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7651 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7652 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7654 (noncommutative_operator): Check the availability of barrel
7657 2018-04-09 Richard Biener <rguenther@suse.de>
7659 PR tree-optimization/85284
7660 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7661 Only use the niter constraining form of simple_iv when the exit
7664 2018-04-09 Tom de Vries <tom@codesourcery.com>
7667 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7668 (define_expand "*memory_barrier"): New define_expand.
7669 (define_insn "memory_barrier"): New insn.
7671 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7673 PR rtl-optimization/80463
7674 PR rtl-optimization/83972
7675 PR rtl-optimization/83480
7677 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7678 correct producer for the insn.
7679 (tidy_control_flow): Fixup seqnos in case of debug insns.
7681 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7683 PR rtl-optimization/83913
7685 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7686 different sched-times when merging exprs.
7688 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7690 PR rtl-optimization/83962
7692 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7693 tidy_fallthru_edge and tidy_control_flow.
7695 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7697 PR rtl-optimization/83530
7699 * sel-sched.c (force_next_insn): New global variable.
7700 (remove_insn_for_debug): When force_next_insn is true, also leave only
7701 next insn in the ready list.
7702 (sel_sched_region): When the region wasn't scheduled, make another pass
7703 over it with force_next_insn set to 1.
7705 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7707 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7709 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7710 for interrupt control.
7711 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7712 functions for interrupt control.
7713 * config/nds32/nds32-intrinsic.md: Likewise.
7714 * config/nds32/nds32_intrinsic.h: Likewise.
7715 * config/nds32/nds32.h (nds32_builtins): Likewise.
7717 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7719 * config/nds32/nds32.c (nds32_init_machine_status,
7720 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7721 strict_aligned_p field.
7722 (nds32_expand_to_rtl_hook): New function.
7723 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7724 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7726 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7727 Chung-Ju Wu <jasonwucj@gmail.com>
7729 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7730 * config/nds32/nds32-n7.md: New file.
7731 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7732 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7734 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7735 * config/nds32/nds32.md (pipeline_model): Add n7.
7736 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7737 * config/nds32/pipelines.md: Include n7 settings.
7739 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7740 Chung-Ju Wu <jasonwucj@gmail.com>
7742 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7743 * config/nds32/nds32-e8.md: New file.
7744 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7745 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7747 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7748 * config/nds32/nds32.md (pipeline_model): Add e8.
7749 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7750 * config/nds32/pipelines.md: Include e8 settings.
7752 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7753 Chung-Ju Wu <jasonwucj@gmail.com>
7755 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7756 * config/nds32/nds32-n8.md: New file.
7757 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7758 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7760 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7761 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7762 * config/nds32/nds32.md (pipeline_model): Add n8.
7763 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7764 * config/nds32/pipelines.md: Include n8 settings.
7766 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7767 Chung-Ju Wu <jasonwucj@gmail.com>
7769 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7770 * config/nds32/nds32-n9-2r1w.md: New file.
7771 * config/nds32/nds32-n9-3r2w.md: New file.
7772 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7773 nds32_register_ports): New or modify for cpu n9.
7774 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7776 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7777 * config/nds32/nds32-utils.c: New file.
7778 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7779 TARGET_MUL_SLOW): Define.
7780 * config/nds32/nds32.md (pipeline_model): New attribute.
7781 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7782 New options that support cpu n9.
7783 * config/nds32/pipelines.md: Include n9 settings.
7784 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7786 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7788 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7789 information if necessary.
7790 (output_cond_branch_compare_zero): Likewise.
7791 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7792 (nds32_target_alignment): Refine for alignment.
7793 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7794 (FUNCTION_BOUNDARY): Modify.
7795 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7797 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7799 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7801 * config/nds32/constants.md (unspec_volatile_element): Add values for
7802 TLB operation and data prefetch.
7803 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7804 functions for TLB operation and data prefetch.
7805 * config/nds32/nds32-intrinsic.md: Likewise.
7806 * config/nds32/nds32_intrinsic.h: Likewise.
7807 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7808 (nds32_print_operand): Likewise.
7809 * config/nds32/nds32.h (nds32_builtins): Likewise.
7811 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7812 Andrew Pinski <pinsika@gcc.gnu.org>
7815 * match.pd: Use constant_boolean_node of correct type instead of
7816 boolean_true_node or boolean_false_node for simplifying
7817 pointer comparisons to zero.
7819 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7821 PR tree-optimization/80021
7822 * tree.c (verify_type_variant): Make error call in verify_variant_match
7823 translatable and remove final full stop.
7825 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7827 * config/nds32/constants.md (unspec_volatile_element): Add
7828 UNSPEC_VOLATILE_EH_RETURN.
7829 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7830 nds32_output_stack_pop): Support dwarf exception handling process.
7831 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7832 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7833 exception handling process.
7834 (nds32_compute_stack_frame): Likewise.
7835 (nds32_return_addr_rtx): Likewise.
7836 (nds32_initial_elimination_offset): Likewise.
7837 (nds32_expand_prologue): Likewise.
7838 (nds32_expand_epilogue): Likewise.
7839 (nds32_dynamic_chain_address): New function.
7840 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7842 (DYNAMIC_CHAIN_ADDRESS): Define.
7843 (EH_RETURN_DATA_REGNO): Define.
7844 (EH_RETURN_STACKADJ_RTX): Define.
7845 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7846 patterns for dwarf exception handling.
7848 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7850 * config/nds32/nds32.h: Clean up obsolete macros.
7852 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7854 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7855 Add enum values for particular instructions.
7856 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7857 particular intrinsic functions.
7858 * config/nds32/nds32-intrinsic.md: Likewise.
7859 * config/nds32/nds32_intrinsic.h: Likewise.
7860 * config/nds32/nds32.h (nds32_builtins): Likewise.
7861 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7862 (btst, ave): New patterns for particular instructions.
7864 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7866 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7867 Add enum values for atomic load/store and memory sync.
7868 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7870 * config/nds32/nds32-intrinsic.md: Likewise.
7871 * config/nds32/nds32_intrinsic.h: Likewise.
7872 * config/nds32/nds32.h (nds32_builtins): Likewise.
7874 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7876 PR tree-optimization/85257
7877 * fold-const.c (native_encode_vector): If not all elts could fit
7878 and off is -1, return 0 rather than offset.
7879 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7880 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7881 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7882 adjust buffer in native_interpret_expr call.
7884 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7886 * config/nds32/constants.md (unspec_volatile_element): Add cache
7887 control enum values.
7888 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7889 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7890 * config/nds32/nds32.c (nds32_cctl_names): New.
7891 (nds32_print_operand): Handle cache control register names.
7892 * config/nds32/nds32.h (nds32_builtins): New enum values.
7893 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7895 * config/nds32/nds32.md (type): Add mmu.
7896 * config/nds32/pipelines.md (simple_insn): Add mmu.
7898 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7900 * config/nds32/nds32.md (type): Remove call.
7901 * config/nds32/pipelines.md (simple_insn): Likewise.
7903 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7905 * config/nds32/constants.md (unspec_volatile_element): Add
7906 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7907 UNSPEC_VOLATILE_FMFCFG.
7908 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7909 description for fmfcfg and fmfcsr.
7910 (bdesc_1arg): Add fmtcsr.
7911 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7912 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7913 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7914 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7915 unspec_fmfcfg): New patterns.
7916 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7917 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7918 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7919 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7920 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7921 __nds32__fmfcfg): Define.
7923 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7925 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7926 intrinsic register names.
7927 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7928 intrinsic register enum values and macros.
7930 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7932 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7933 for load/store addressing form.
7934 (nds32_print_operand_address): Likewise.
7936 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7939 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7940 based on LABEL_REF. Remove useless assertion.
7941 (pic_address_needs_scratch): Fix formatting.
7942 (sparc_legitimize_pic_address): Minor tweaks.
7943 (sparc_delegitimize_address): Adjust assertion accordingly.
7944 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7945 into symbolic_operand.
7946 (movsi_high_pic_label_ref): Likewise.
7947 (movsi_lo_sum_pic_label_ref): Likewise.
7948 (movdi_pic_label_ref): Likewise.
7949 (movdi_high_pic_label_ref): Likewise.
7950 (movdi_lo_sum_pic_label_ref): Likewise.
7952 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7954 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7955 custom LIB_SPEC setup.
7957 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7958 Kito Cheng <kito.cheng@gmail.com>
7960 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7961 * config/riscv/freebsd.h: New.
7963 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7965 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7966 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7969 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7970 Kito Cheng <kito.cheng@gmail.com>
7972 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7973 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7974 * config/nds32/nds32-protos.h (nds32_output_call,
7975 nds32_output_return): Declare.
7976 * config/nds32/nds32.md: Refine all the call and return patterns.
7978 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7981 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7982 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7984 PR rtl-optimization/84872
7985 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7986 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7989 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7991 * expr.c (copy_blkmode_to_reg): Revert 254862.
7992 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7994 2018-04-06 Richard Biener <rguenther@suse.de>
7997 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7998 after seeing a component reference with an adjacent field. Treat
7999 refs to arrays at struct end of external decls similar to
8000 refs to unconstrained commons.
8002 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8005 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
8006 look through SAVE_EXPRs with non-side-effects argument. Adjust
8008 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
8011 2018-04-06 Richard Biener <rguenther@suse.de>
8014 * alias.c (find_base_term): New wrapper around find_base_term
8015 unwinding CSELIB_VAL_PTR changes.
8016 (find_base_term): Do not restore CSELIB_VAL_PTR during the
8019 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8021 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
8023 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
8024 constant definitions.
8025 ("nop"): lr 0,0 -> nopr r0
8026 ("nop_lr0", "nop_lr1"): New insn definitions.
8028 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8030 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
8031 NDS32_V3PUSH_AVAILABLE_P macro.
8033 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
8034 Chung-Ju Wu <jasonwucj@gmail.com>
8036 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
8037 (nds32*-*-*): Add float and fpu_config into supported_defaults.
8038 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
8039 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
8040 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
8041 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
8042 * config/nds32/constraints.md: New constraints and checking for hard
8043 float configuration.
8044 * config/nds32/iterators.md: New mode iterator and attribute for hard
8045 float configuration.
8046 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
8048 * config/nds32/nds32-fpu.md: New file.
8049 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
8050 deal with hard float code generation.
8051 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
8053 (abi_type, float_reg_number): New enum type.
8054 * config/nds32/nds32-predicates.c: New predicates for hard float.
8055 * config/nds32/nds32-protos.h: Declare functions for hard float.
8056 * config/nds32/nds32.c: Implementation for hard float configuration.
8057 * config/nds32/nds32.h: Definitions for hard float configuration.
8058 * config/nds32/nds32.md: Include hard float machine description and
8059 modify patterns for hard float configuration.
8060 * config/nds32/nds32.opt: New options for hard float configuration.
8061 * config/nds32/predicates.md: New predicates for hard float
8064 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8066 * common/config/nds32/nds32-common.c
8067 (nds32_option_optimization_table): Enable -mreleax-hint by default.
8069 2018-04-05 Jakub Jelinek <jakub@redhat.com>
8072 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
8073 CONSTRUCTOR_ELT (ctor, ...)->value.
8075 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
8078 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
8080 2018-04-05 Tom de Vries <tom@codesourcery.com>
8083 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
8086 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8087 Kito Cheng <kito.cheng@gmail.com>
8089 * config/nds32/constraints.md (U33): Fine-tune checking condition.
8090 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
8091 * config/nds32/nds32.h (nds32_16bit_address_type): Add
8092 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
8094 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8095 Kito Cheng <kito.cheng@gmail.com>
8097 * config/nds32/constraints.md (Ufe): New memory constraint.
8098 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
8099 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
8100 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
8102 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
8103 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
8105 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8107 * config/nds32/nds32.md: Use optimize_size in the condition for
8108 alu-shift instructions.
8110 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8112 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
8114 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8116 * config/nds32/nds32.md (negsi2): Refine pattern.
8118 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8119 Chung-Ju Wu <jasonwucj@gmail.com>
8121 * config/nds32/iterators.md (shift_rotate): New code iterator.
8122 (shift): New code attribute.
8123 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
8124 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
8125 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
8126 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
8127 bit-wise operations.
8128 (andsi3, *andsi3): Ditto.
8129 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
8130 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
8131 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
8132 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
8133 nds32_ior_operand, nds32_xor_operand): New predicates.
8135 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8137 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
8138 (addsi3, subsi3): ... this.
8140 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8142 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
8144 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8146 * config/nds32/nds32.md: Adjust indention.
8148 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8150 * config/nds32/nds32.md (feature): New attribute.
8152 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8154 * config/nds32/nds32.md (subtype): New attribute.
8156 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
8159 * config/arm/arm-builtins.c (arm_expand_builtin): Change
8160 expansion to perform a bitwise AND of the argument followed by a
8161 boolean negation of the result.
8163 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
8165 PR rtl-optimization/84878
8166 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
8167 the basic block. Assert the use reference is not artificial and that
8168 it has an associated insn.
8170 2018-04-04 Michael Matz <matz@suse.de>
8172 * builtins.c (compute_objsize): Pass correct operand
8173 to array_at_struct_end_p.
8175 2018-04-04 Richard Biener <rguenther@suse.de>
8178 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
8179 from contexts for DINFO_LEVEL_TERSE and below.
8181 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8183 * config/nds32/nds32-doubleword.md (move_<mode>): Require
8184 resiter_operand condition.
8185 * config/nds32/nds32.md (*move<mode>): Ditto.
8187 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8188 Monk Chiang <sh.chiang04@gmail.com>
8190 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
8192 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8194 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
8196 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8197 Kito Cheng <kito.cheng@gmail.com>
8199 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
8200 nds32_cond_code_str, output_cond_branch,
8201 output_cond_branch_compare_zero, nds32_expand_cbranch,
8202 nds32_expand_cstore, nds32_expand_movcc,
8203 nds32_output_cbranchsi4_equality_zero,
8204 nds32_output_cbranchsi4_equality_reg,
8205 nds32_output_cbranchsi4_equality_reg_or_const_int,
8206 nds32_output_cbranchsi4_greater_less_zero: New functions.
8207 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
8208 nds32_expand_cstore, nds32_expand_movcc,
8209 nds32_output_cbranchsi4_equality_zero,
8210 nds32_output_cbranchsi4_equality_reg,
8211 nds32_output_cbranchsi4_equality_reg_or_const_int,
8212 nds32_output_cbranchsi4_greater_less_zero): Declare.
8213 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
8214 nds32_rimm11s_operand): New predicates.
8215 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
8216 * config/nds32/nds32.md: Rewrite all the branch and conditional move
8219 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8221 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
8222 * config/nds32/nds32.md: Ditto.
8223 * config/nds32/pipelines.md: Ditto.
8225 2018-04-04 Richard Biener <rguenther@suse.de>
8227 PR tree-optimization/85168
8228 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
8229 propagating abnormals.
8231 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8233 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
8235 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8236 Kito Cheng <kito.cheng@gmail.com>
8238 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
8239 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
8240 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
8241 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
8242 * config/nds32/nds32.md (sibcall_internal): New.
8243 (sibcall_register): Remove.
8244 (sibcall_immediate): Remove.
8245 (sibcall_value_internal): New.
8246 (sibcall_value_register): Remove.
8247 (sibcall_value_immediate): Remove.
8248 * config/nds32/predicates.md (nds32_general_register_operand): New.
8249 (nds32_call_address_operand): New.
8251 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8253 PR rtl-optimization/85167
8254 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
8255 bb_defs if *split_p, instead preinitialize it to NULL.
8257 PR tree-optimization/85156
8258 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
8259 evaluating the argument multiple times.
8261 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
8263 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
8265 (_mm_cvtpd_ps): Likewise.
8266 (_mm_cvttpd_epi32): Likewise.
8267 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
8268 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
8269 vector, pixel, and bool following altivec.h include.
8271 2018-04-03 Martin Sebor <msebor@redhat.com>
8273 * doc/extend.texi (Common Function Attributes): Clarify.
8274 (const attribute): Likewise.
8275 (pure attribute): Likewise.
8277 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8280 * config/i386/i386.c (ix86_expand_vector_set): Use
8281 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
8283 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
8285 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
8286 instructions when changing rounding bits to preserve precision bits
8287 in the x87 control word.
8289 2018-04-03 Martin Liska <mliska@suse.cz>
8291 PR tree-optimization/82491
8292 * rtl.h (strip_offset_and_add): Replace += suboffset with
8293 poly_uint64 () + suboffset.
8295 2018-03-29 Martin Liska <mliska@suse.cz>
8296 Martin Jambor <mjambor@suse.cz>
8299 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
8300 param_type is not an integral or pointer type.
8302 2018-04-03 Richard Biener <rguenther@suse.de>
8304 * sese.h (recompute_all_dominators): Remove.
8306 2018-04-02 Martin Sebor <msebor@redhat.com>
8308 * doc/invoke.texi (-Wrestrict): Fix typos.
8310 2018-04-02 Jim Wilson <jimw@sifive.com>
8312 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
8313 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
8314 (<optab>di3, <optab>si3_extend): Likewise.
8315 (<optab>si3_mask, <optab>si3_mask_1): New.
8316 (<optab>di3_mask, <optab>di3_mask_1): New.
8317 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
8318 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
8319 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
8321 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
8323 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
8326 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
8328 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
8329 (nds32_canonicalize_comparison): New function.
8331 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8332 Kito Cheng <kito.cheng@gmail.com>
8333 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8335 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
8336 * config/nds32/constants.md (unspec_volatile_element): Add
8337 UNSPEC_VOLATILE_RELAX_GROUP.
8338 * config/nds32/nds32-relax-opt.c: New file.
8339 * config/nds32/nds32-predicates.c
8340 (nds32_symbol_load_store_p): New function.
8341 * config/nds32/nds32-protos.h
8342 (nds32_symbol_load_store_p): Declare function.
8343 (make_pass_nds32_relax_opt): Declare new rtl pass function.
8344 * config/nds32/nds32.c
8345 (nds32_register_pass): New function to register pass.
8346 (nds32_register_passes): New function to register passes.
8347 * config/nds32/nds32.md (relax_group): New pattern.
8348 * config/nds32/nds32.opt (mrelax-hint): New option.
8349 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
8351 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
8353 * config/nds32/t-nds32: Modify files dependency.
8355 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8357 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
8358 (PROFILE_HOOK): Define its implementation.
8360 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8362 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
8363 type and 32-bit size.
8365 2018-04-01 Jakub Jelinek <jakub@redhat.com>
8368 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
8369 (V_128_256): New mode iterator.
8370 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
8371 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
8372 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
8374 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
8375 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
8377 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
8380 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
8381 NaN inputs correctly.
8383 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
8386 * config/rs6000/vsx.md (??r): New mode attribute.
8387 (*vsx_mov<mode>_64bit): Use it.
8388 (*vsx_mov<mode>_32bit): Likewise.
8390 2018-03-30 Martin Sebor <msebor@redhat.com>
8392 PR tree-optimization/84818
8393 * builtins.c (check_access): Use warning_n.
8395 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8398 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
8400 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
8403 2018-03-30 Julia Koval <julia.koval@intel.com>
8406 * x86-tune.def (movx, partial_reg_dependency): Enable for
8409 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
8412 * lra-constraints.c (process_alt_operands): Move setting
8413 this_alternative_matches below.
8415 2018-03-29 Martin Liska <mliska@suse.cz>
8418 * doc/invoke.texi: Document how LTO works with debug info.
8419 Describe auto-load support of binutils. Mention 'x86-64'
8420 as valid option value of -march option.
8422 2018-03-29 Jakub Jelinek <jakub@redhat.com>
8424 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
8427 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
8428 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
8429 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
8432 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8435 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
8436 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
8437 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
8438 for RS6000_BTM_POWERPC64.
8439 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
8440 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
8441 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
8446 2018-03-28 Carl Love <cel@us.ibm.com>
8449 2017-09-27 Carl Love <cel@us.ibm.com>
8451 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
8452 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
8453 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
8456 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8458 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
8459 instead of __vector bool.
8460 (_mm_max_pu8): Likewise.
8461 (_mm_min_pi16): Likewise.
8463 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8466 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
8467 (DIVWEUO): Likewise.
8469 (DIVDEUO): Likewise.
8470 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
8471 DIVWEUO and DIVDEUO.
8472 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
8473 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
8474 (div_extend): Likewise.
8475 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
8477 (__builtin_divweuo): Likewise.
8478 (__builtin_divdeo): Likewise.
8479 (__builtin_divdeuo): Likewise.
8481 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8484 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
8485 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
8487 PR tree-optimization/82004
8488 * gimple-match-head.c (optimize_pow_to_exp): New function.
8489 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
8490 Don't fold to exp if optimize_pow_to_exp is false.
8492 2018-03-28 Martin Liska <mliska@suse.cz>
8495 * calls.c (initialize_argument_information): Fix trailing space.
8496 * common.opt: Fix typo and provide better explanation for
8497 -fsanitize-coverage option.
8498 * config/i386/i386.opt: Fix typo.
8500 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8501 Martin Liska <mliska@suse.cz>
8504 * gimplify.c (asan_poison_variable): Don't do the check for
8505 gimplify_omp_ctxp here.
8506 (gimplify_decl_expr): Do it here.
8507 (gimplify_target_expr): Likewise.
8509 2018-03-28 Martin Liska <mliska@suse.cz>
8512 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8513 chkp_type_bounds_count if MPX is not enabled.
8515 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
8517 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
8519 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8522 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
8523 function to create the function decl for complex long double
8524 multiply and divide for -mabi=ieeelongdouble.
8525 (init_float128_ieee): Call it.
8527 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8530 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8531 -fcf-protection=branch -mibt.
8532 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8534 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8537 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8538 arm_word_relocations.
8540 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8543 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8544 extern array declarations.
8546 2018-03-27 Richard Biener <rguenther@suse.de>
8549 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8550 explicit single_use checks.
8552 2018-03-27 Richard Biener <rguenther@suse.de>
8554 PR tree-optimization/85082
8555 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8558 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8560 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8561 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8562 Turn on fasynchronous-unwind-tables and funwind-tables.
8564 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8567 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8568 (*bmi_blsr_<mode>_ccz): Ditto.
8570 2018-03-26 Tom de Vries <tom@codesourcery.com>
8572 PR tree-optimization/85063
8573 * omp-general.c (offloading_function_p): New function. Factor out
8575 * omp-offload.c (pass_omp_target_link::gate): ... here.
8576 * omp-general.h (offloading_function_p): Declare.
8577 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8578 with attribute omp declare target for offloading functions.
8580 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8582 PR tree-optimization/84005
8583 * tree-data-ref.h (get_base_for_alignment): Declare.
8584 * tree-data-ref.c (get_base_for_alignment_1): New function.
8585 (get_base_for_alignment): Likewise.
8586 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8587 get_base_for_alignment to find a suitable base object, instead
8588 of always using drb->base_address.
8590 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8593 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8594 known size by default.
8596 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8599 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8600 and non BLKmode operands.
8602 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8605 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8606 Clean up attributes.
8608 2018-03-23 Richard Biener <rguenther@suse.de>
8611 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8612 we are going to emit early debug for LTO.
8614 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8617 * function.c (match_asm_constraints_1): Don't optimize if input
8618 doesn't satisfy general_operand predicate for output's mode.
8621 * alias.c (write_dependence_p): Don't require for x_canonicalized
8622 non-VOIDmode if x has VOIDmode.
8625 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8626 just don't try to optimize it rather than assert it never happens.
8628 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8630 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8631 macro expansions for definition of ST_INTERNAL_<mode> and
8632 LD_INTERNAL_<mode> builtins.
8633 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8635 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8637 (altivec_expand_st_builtin): Likewise.
8638 (altivec_expand_builtin): Remove calls to deleted functions.
8639 (rs6000_address_for_altivec): Delete this function.
8640 * config/rs6000/vector.md: Remove expands for
8641 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8643 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8646 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8647 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8648 re-computing once computed.
8649 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8650 (arm_init_machine_status): Initialize
8651 machine->static_chain_stack_bytes.
8653 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8656 * doc/extend.texi: Add four new prototypes for vec_ld.
8657 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8658 definitions for more logical presentation.
8659 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8660 entries for V1TI variants of __builtin_altivec_ld builtin.
8661 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8662 handling of V1TI variant of LVX icode pattern.
8663 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8664 (rs6000_gimple_fold_builtin): Likewise.
8665 (altivec_init_builtins): Add code to define
8666 __builtin_altivec_lvx_v1ti function.
8668 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8671 * function.c (match_asm_constraints_1): Don't do the optimization
8672 if input isn't a REG, SUBREG, MEM or constant.
8674 2018-03-22 Tom de Vries <tom@codesourcery.com>
8676 PR tree-optimization/84956
8677 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8678 bb_has_abnormal_pred.
8680 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8683 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8684 DECL_INITIAL (decl) to decl at the end.
8685 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8688 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8690 * doc/extend.texi (__builtin_tgmath): Document when complex
8691 integer types are treated as _Complex _Float64.
8693 2018-03-21 Tom de Vries <tom@codesourcery.com>
8695 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8697 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8699 PR tree-optimization/84960
8700 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8701 if it is ENTRY block, move them into single succ of ENTRY in that case.
8703 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8705 PR tree-optimization/84811
8706 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8707 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8708 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8709 (known_subrange_p): Update accordingly. Cast each value involved
8710 in the size comparison, rather than casting the result of the
8713 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8715 PR tree-optimization/84982
8716 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8717 by flipping the least significant bit rather than all bits from
8718 bitpos to bitpos + bitsize - 1.
8720 2018-03-21 Nathan Sidwell <nathan@acm.org>
8722 * doc/extend.texi (Deprecated Features): Remove mention of
8723 long-deleted deprecations.
8725 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8728 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8729 * configure: Regenerate.
8731 2018-03-21 Tom de Vries <tom@codesourcery.com>
8733 PR tree-optimization/83126
8734 * tree-parloops.c (num_phis): New function.
8735 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8737 2018-03-21 Nathan Sidwell <nathan@acm.org>
8739 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8740 mention anon-struct/union members and trailing attributes.
8742 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8744 PR tree-optimization/84969
8745 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8746 builtin memset partitions if they set different rhs values.
8748 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8750 PR rtl-optimization/84989
8751 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8752 VEC_DUPLICATE with scalar result mode.
8754 2018-03-21 Martin Liska <mliska@suse.cz>
8757 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8758 not intended return statement.
8760 2018-03-21 Martin Liska <mliska@suse.cz>
8763 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8764 (chkp_find_bound_slots_1): Limit number of iterations.
8766 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8769 * Minor grammar fixes for x86 options.
8771 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8774 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8775 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8777 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8780 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8781 (altivec_lvx_<mode>_1op): Likewise.
8782 (altivec_stvx_<mode>_2op): Likewise.
8783 (altivec_stvx_<mode>_1op): Likewise.
8784 (altivec_lvx_<VM2:mode>): New define_expand.
8785 (altivec_stvx_<VM2:mode>): Likewise.
8786 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8787 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8788 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8789 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8790 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8791 (rs6000_gen_lvx): Likewise.
8792 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8793 (altivec_expand_stv_builtin): Likewise.
8794 (altivec_expand_builtin): Likewise.
8795 * config/rs6000/vector.md: Likewise.
8797 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8800 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8803 2018-03-20 Richard Biener <rguenther@suse.de>
8806 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8807 sign-conversions as zero, fall back to standard scalar_stmt
8810 2018-03-20 Martin Liska <mliska@suse.cz>
8813 * predict.c (rebuild_frequencies): Handle case when we have
8814 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8816 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8819 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8820 flag_section_anchors.
8821 * varasm.c (use_blocks_for_decl_p): Remove hack for
8822 dw2_force_const_mem.
8825 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8827 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8828 be created, use lowpart_subreg of operands[0] rather than operands[0]
8830 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8831 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8832 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8833 and n constraint instead of aarch64_shift_imm_di and Usd.
8834 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8835 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8837 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8840 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8841 to favor GPR over NEON registers.
8842 (<shift>di3_neon): Likewise.
8844 2018-03-20 Tom de Vries <tom@codesourcery.com>
8847 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8848 (nvptx_process_pars): Emit bar.sync asap and alap.
8850 2018-03-20 Tom de Vries <tom@codesourcery.com>
8853 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8854 seen_label if seen_label is already set.
8856 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8859 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8860 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8861 Use 1U instead of 1. Formatting fixes.
8864 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8865 instead of TREE_TYPE (s1) for the return value.
8867 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8869 PR tree-optimization/84946
8870 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8871 bitsize + bitsize in poly_uint64 rather than poly_int64.
8874 * dwarf2asm.c: Include fold-const.c.
8875 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8876 of decl rather than decl itself.
8878 PR rtl-optimization/84643
8879 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8881 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8884 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8885 calling assemble_variable.
8887 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8890 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8891 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8893 2018-03-19 Jim Wilson <jimw@sifive.com>
8896 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8897 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8898 (riscv_first_stack_step): Likewise.
8899 (riscv_option_override): Use STACK_BOUNDARY instead of
8901 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8903 (BIGGEST_ALIGNMENT): Set to 128.
8904 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8905 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8908 2018-03-19 Richard Biener <rguenther@suse.de>
8910 PR tree-optimization/84933
8911 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8912 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8914 2018-03-19 Richard Biener <rguenther@suse.de>
8916 PR tree-optimization/84859
8917 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8918 (cond_if_else_store_replacement): Perform sinking operation on
8919 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8920 Generalize what a BB with a single eligible store is.
8922 2018-03-19 Richard Biener <rguenther@suse.de>
8924 PR tree-optimization/84929
8925 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8926 chrec_is_positive against non-chrec arg.
8928 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8931 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8933 2018-03-18 Martin Liska <mliska@suse.cz>
8935 PR rtl-optimization/84635
8936 * regrename.c (build_def_use): Use matches_mode only when
8939 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8941 PR tree-optimization/84913
8942 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8943 vectorize chains of COND_EXPRs.
8945 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8947 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8949 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8951 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8953 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8955 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8957 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8958 Kito Cheng <kito.cheng@gmail.com>
8960 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8961 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8962 (nds32_adjust_reg_alloc_order): New function.
8963 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8965 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8967 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8968 nds32_print_operand, nds32_print_operand_address): Use
8969 HOST_WIDE_INT_PRINT_DEC instead.
8971 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8973 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8975 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8978 * config/i386/i386.c (initial_ix86_tune_features,
8979 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8981 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8982 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8983 rather than 1u << ix86_tune. Formatting fix.
8984 (ix86_option_override_internal): Change ix86_arch_mask from
8985 unsigned int to unsigned HOST_WIDE_INT, initialize to
8986 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8987 (ix86_function_specific_restore): Likewise.
8989 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8992 * postreload.c (reload_combine_recognize_pattern): Perform
8993 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8994 truncate_int_for_mode the result for the destination's mode.
8997 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8998 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
9000 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9003 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
9006 * lra-assigns.c (lra_split_hard_reg_for): Don't use
9007 regno_allocno_class_array and sorted_pseudos.
9008 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
9009 insns where regno is used.
9011 2018-03-16 Martin Liska <mliska@suse.cz>
9014 * multiple_target.c (create_dispatcher_calls): Redirect
9015 reference in the symbol table.
9017 2018-03-16 Martin Liska <mliska@suse.cz>
9020 * multiple_target.c (create_dispatcher_calls): Redirect also
9023 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9027 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
9028 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
9029 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
9031 2018-03-16 Julia Koval <julia.koval@intel.com>
9033 * doc/invoke.texi (Skylake Server): Add CLWB.
9034 Cannonlake): Remove CLWB.
9036 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9038 PR tree-optimization/84841
9039 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
9041 (FLOAT_ONE_CONST_TYPE): Define.
9042 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
9043 (sort_by_operand_rank): Put entries with higher constant_type last
9044 rather than first to match comments.
9046 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
9048 * config/nios2/nios2.md (movsi_internal): Fix thinko in
9051 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9054 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
9055 check and use address of target always.
9057 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
9060 * config/i386/i386.c (indirect_thunk_needed): Update comments.
9061 (indirect_thunk_bnd_needed): Likewise.
9062 (indirect_thunks_used): Likewise.
9063 (indirect_thunks_bnd_used): Likewise.
9064 (indirect_return_needed): New.
9065 (indirect_return_bnd_needed): Likewise.
9066 (output_indirect_thunk_function): Add a bool argument for
9068 (output_indirect_thunk_function): Don't generate alias for
9069 function return thunk.
9070 (ix86_code_end): Call output_indirect_thunk_function to generate
9071 function return thunks.
9072 (ix86_output_function_return): Set indirect_return_bnd_needed
9073 and indirect_return_needed instead of indirect_thunk_bnd_needed
9074 and indirect_thunk_needed.
9076 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
9078 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
9079 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
9080 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
9082 2018-03-15 David Malcolm <dmalcolm@redhat.com>
9083 Paul Hua <paul.hua.gm@gmail.com>
9086 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
9088 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
9090 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
9091 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
9094 2018-03-15 Tamar Christina <tamar.christina@arm.com>
9097 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9098 instead of GET_MODE_SIZE when comparing Units.
9100 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9103 * varasm.c (hash_section): Return an unchangeble hash value
9104 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
9105 Return !aarch64_can_use_per_function_literal_pools_p ().
9107 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9110 * optabs.c (emit_conditional_move): Pass address of cmode's copy
9111 rather than address of cmode as last argument to prepare_cmp_insn.
9113 2018-03-15 Julia Koval <julia.koval@intel.com>
9115 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
9116 F_AVX512VNNI, F_AVX512BITALG): New.
9118 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
9121 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
9122 insn for floating-point loads and stores.
9124 2018-03-14 Carl Love <cel@us.ibm.com>
9126 * config/rs6000/rs6000-c.c: Add macro definitions for
9127 ALTIVEC_BUILTIN_VEC_PERMXOR.
9128 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
9129 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
9130 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
9131 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
9133 * config/doc/extend.texi: Add prototypes for vec_permxor.
9135 2018-03-14 David Malcolm <dmalcolm@redhat.com>
9138 * diagnostic-show-locus.c (class layout_point): Convert m_line
9139 from int to linenum_type.
9140 (line_span::comparator): Use linenum "compare" function when
9141 comparing line numbers.
9142 (test_line_span): New function.
9143 (layout_range::contains_point): Convert param "row" from int to
9145 (layout_range::intersects_line_p): Likewise.
9146 (layout::will_show_line_p): Likewise.
9147 (layout::print_source_line): Likewise.
9148 (layout::should_print_annotation_line_p): Likewise.
9149 (layout::print_annotation_line): Likewise.
9150 (layout::print_leading_fixits): Likewise.
9151 (layout::annotation_line_showed_range_p): Likewise.
9152 (struct line_corrections): Likewise for field m_row.
9153 (line_corrections::line_corrections): Likewise for param "row".
9154 (layout::print_trailing_fixits): Likewise.
9155 (layout::get_state_at_point): Likewise.
9156 (layout::get_x_bound_for_row): Likewise.
9157 (layout::print_line): Likewise.
9158 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
9159 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
9160 * input.c (selftest::test_linenum_comparisons): New function.
9161 (selftest::input_c_tests): Call it.
9162 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
9163 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
9164 * selftest.h (ASSERT_GT): New macro.
9165 (ASSERT_GT_AT): New macro.
9166 (ASSERT_LT): New macro.
9167 (ASSERT_LT_AT): New macro.
9169 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
9171 PR rtl-optimization/84780
9172 * combine.c (distribute_links): Don't make a link based on pc_rtx.
9174 2018-03-14 Martin Liska <mliska@suse.cz>
9176 * tree.c (record_node_allocation_statistics): Use
9177 get_stats_node_kind.
9178 (get_stats_node_kind): New function extracted from
9179 record_node_allocation_statistics.
9180 (free_node): Use get_stats_node_kind.
9182 2018-03-14 Richard Biener <rguenther@suse.de>
9184 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
9185 that the value-set of ANTIC_IN doesn't grow.
9188 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9190 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9191 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9193 2018-03-14 Julia Koval <julia.koval@intel.com>
9195 * config.gcc (icelake-client, icelake-server): New.
9197 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
9198 (initial_ix86_arch_features): Ditto.
9199 (PTA_SKYLAKE): Add SGX.
9200 (PTA_ICELAKE): Remove.
9201 (PTA_ICELAKE_CLIENT): New.
9202 (PTA_ICELAKE_SERVER): New.
9203 (ix86_option_override_internal): Split up icelake on icelake client and
9205 (get_builtin_code_for_version): Ditto.
9206 (fold_builtin_cpu): Ditto.
9207 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
9208 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
9209 * config/i386/i386.h (processor_type): Ditto.
9210 * doc/invoke.texi: Ditto.
9212 2018-03-14 Jakub Jelinek <jakub@redhat.com>
9215 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
9216 INTEGER_CST offset, add it together with bitpos / 8 and
9217 sign extend based on POINTER_SIZE.
9221 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
9224 * config/i386/constraints.md (Yc): New register constraint.
9225 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
9226 Use Yc constraint for alternative 2 of operand 0. Remove
9227 preferred_for_speed attribute.
9229 2018-03-14 Richard Biener <rguenther@suse.de>
9231 PR tree-optimization/84830
9232 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
9233 with the old one to avoid oscillations.
9235 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
9238 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
9240 (assign_by_spills): Return a flag of reload assignment failure.
9241 Do not process the reload assignment failures. Do not spill other
9242 reload pseudos if they has the same reg class. Update n if
9244 (lra_assign): Add a return arg. Set up from the result of
9245 assign_by_spills call.
9246 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9247 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9248 usage_insns if it is not NULL.
9249 (spill_hard_reg_in_range): New function.
9250 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9251 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9252 function prototypes.
9253 (lra_assign): Change prototype.
9254 * lra.c (lra): Add code to deal with fails by splitting hard reg
9257 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
9259 * config/riscv/riscv.opt (mrelax): New option.
9260 * config/riscv/riscv.c (riscv_file_start): Emit ".option
9261 "norelax" when riscv_mrelax is disabled.
9262 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
9264 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9267 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
9268 reassociation for int modes.
9270 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9272 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
9273 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
9275 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
9276 * config/aarch64/aarch64-sve.md
9277 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
9278 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
9279 (*extend<mode><Vwide>2): Rename to...
9280 (aarch64_sve_extend<mode><Vwide>2): ...this.
9281 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
9282 renaming the old pattern to...
9283 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
9285 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
9286 define_expand, renaming the old pattern to...
9287 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
9288 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
9289 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
9290 account when deciding which SVE instruction the optab should use.
9291 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
9293 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9295 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
9296 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
9297 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
9298 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
9299 (tlsdesc_small_<mode>): Turn a define_expand and use
9300 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
9301 (tlsdesc_small_advsimd_<mode>): ...this.
9302 (tlsdesc_small_sve_<mode>): New pattern.
9304 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9306 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
9307 (UNSPEC_UMUL_HIGHPART): New constants.
9308 (MUL_HIGHPART): New int iteraor.
9309 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
9310 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
9312 (*<su>mul<mode>3_highpart): New define_insn.
9314 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9317 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9320 2018-03-13 Martin Liska <mliska@suse.cz>
9323 * (sem_item_optimizer::sem_item_optimizer): Initialize new
9325 (sem_item_optimizer::~sem_item_optimizer): Release it.
9326 (sem_item_optimizer::merge_classes): Register variable aliases.
9327 (sem_item_optimizer::fixup_pt_set): New function.
9328 (sem_item_optimizer::fixup_points_to_sets): Likewise.
9329 * ipa-icf.h: Declare new variables and functions.
9331 2018-03-13 Jakub Jelinek <jakub@redhat.com>
9334 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
9335 integer_pow2p@2 and test integer_pow2p in condition.
9336 (A < 0 ? C : 0): Similarly for @1.
9339 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
9340 characters starting at p contain '\0' character, don't look beyond
9344 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
9345 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
9348 * reg-stack.c (change_stack): Change update_end var from int to
9349 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
9350 also call set_block_for_insn on the newly added insns and rescan.
9353 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
9354 on the last operand.
9357 * tree.c (stabilize_reference_1): Return save_expr (e) for
9358 STATEMENT_LIST even if it doesn't have side-effects.
9360 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
9362 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
9364 2018-03-12 Renlin Li <renlin.li@arm.com>
9366 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
9367 aarch64_output_scalar_simd_mov_immediate.
9369 2018-03-12 Martin Sebor <msebor@redhat.com>
9371 PR tree-optimization/83456
9372 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
9373 for perfectly overlapping calls to memcpy.
9374 (gimple_fold_builtin_memory_chk): Same.
9375 (gimple_fold_builtin_strcpy): Handle no-warning.
9376 (gimple_fold_builtin_stxcpy_chk): Same.
9377 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
9379 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9381 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
9382 parameter. Use it for SFmode.
9383 (rs6000_function_arg_advance_1): Adjust.
9384 (rs6000_function_arg): Adjust.
9385 (rs6000_gimplify_va_arg): Pass false for that new parameter.
9387 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9389 PR rtl-optimization/84169
9390 PR rtl-optimization/84780
9391 * combine.c (can_combine_p): Check for a 2-insn combination whether
9392 the destination register is used between the two insns, too.
9394 2018-03-12 Richard Biener <rguenther@suse.de>
9396 PR tree-optimization/84803
9397 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
9398 for refs DR analysis didn't process.
9400 2018-03-12 Richard Biener <rguenther@suse.de>
9402 PR tree-optimization/84777
9403 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
9404 force-vectorize loops ignore whether we are optimizing for size.
9406 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
9408 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
9409 (TARGET_MD_ASM_ADJUST): Define.
9411 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
9412 Kito Cheng <kito.cheng@gmail.com>
9413 Chung-Ju Wu <jasonwucj@gmail.com>
9415 * config/nds32/nds32.c (nds32_compute_stack_frame,
9416 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
9417 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
9418 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
9419 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
9420 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
9421 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
9422 * config/nds32/nds32.md (prologue, epilogue): Use macro
9423 NDS32_V3PUSH_AVAILABLE_P to do checking.
9425 2018-03-11 Jakub Jelinek <jakub@redhat.com>
9428 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
9429 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
9430 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
9431 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
9432 addition of most attributes on !orig_type_die or the attribute not
9433 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
9435 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9436 Chung-Ju Wu <jasonwucj@gmail.com>
9438 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
9440 * config/nds32/nds32.opt (mvh): New option.
9442 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9443 Chung-Ju Wu <jasonwucj@gmail.com>
9445 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
9447 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
9448 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
9451 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9452 Chung-Ju Wu <jasonwucj@gmail.com>
9454 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
9456 * config/nds32/nds32-multiple.md (strlensi): New pattern.
9457 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
9459 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
9460 Kito Cheng <kito.cheng@gmail.com>
9461 Chung-Ju Wu <jasonwucj@gmail.com>
9463 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
9464 UNSPEC_FFMISM and UNSPEC_FLMISM.
9465 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
9466 for ffb, ffmism and flmism.
9467 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
9468 (unspec_ffmism): Ditto.
9469 (unspec_flmism): Ditto.
9470 (nds32_expand_builtin_impl): Check if string extension is available.
9471 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
9472 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
9474 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
9477 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9480 * lra-assigns.c (assign_by_spills): Return a flag of reload
9481 assignment failure. Do not process the reload assignment
9482 failures. Do not spill other reload pseudos if they has the same
9484 (lra_assign): Add a return arg. Set up from the result of
9485 assign_by_spills call.
9486 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9487 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9488 usage_insns if it is not NULL.
9489 (spill_hard_reg_in_range): New function.
9490 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9491 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9492 function prototypes.
9493 (lra_assign): Change prototype.
9494 * lra.c (lra): Add code to deal with fails by splitting hard reg
9497 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9500 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9502 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9505 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9506 (dw_val_node): Add val_symbolic_view.
9507 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9508 (symview_upper_bound): New.
9509 (new_line_info_table): Initialize symviews_since_reset.
9510 (dwarf2out_source_line): Count symviews_since_reset and set
9511 symview_upper_bound.
9512 (dw_val_equal_p): Handle symview.
9513 (add_AT_symview): New.
9514 (print_dw_val): Handle symview.
9515 (attr_checksum, attr_checksum_ordered): Likewise.
9516 (same_dw_val_p, size_of_die): Likewise.
9517 (value_format, output_die): Likewise.
9518 (add_high_low_attributes): Use add_AT_symview for entry_view.
9519 (dwarf2out_finish): Reset symview_upper_bound, clear
9522 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
9525 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
9526 Add strict argument and use it.
9527 (rs6000_split_multireg_move): Update for new strict argument.
9528 (mem_operand_gpr): Disallow all non-offsettable addresses.
9529 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9531 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9534 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9535 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9536 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9539 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9540 decl, use remap_type if we want to use the type.
9542 2018-03-09 Martin Sebor <msebor@redhat.com>
9544 PR tree-optimization/84526
9545 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9547 (builtin_access::generic_overlap): Be prepared to handle non-array
9550 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9552 PR rtl-optimization/84682
9553 * lra-constraints.c (process_address_1): Check is_address flag
9554 for address constraints.
9555 (process_alt_operands): Likewise.
9556 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9557 preprocess_constraints.
9558 * recog.h (preprocess_constraints): Add oploc parameter.
9560 * recog.c (preprocess_constraints): Test address_operand for
9561 CT_ADDRESS constraints.
9563 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9566 * lra-assigns.c (assign_by_spills): Return a flag of reload
9567 assignment failure. Do not process the reload assignment
9568 failures. Do not spill other reload pseudos if they has the same
9570 (lra_assign): Add a return arg. Set up from the result of
9571 assign_by_spills call.
9572 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9573 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9574 usage_insns if it is not NULL.
9575 (spill_hard_reg_in_range): New function.
9576 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9577 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9578 function prototypes.
9579 (lra_assign): Change prototype.
9580 * lra.c (lra): Add code to deal with fails by splitting hard reg
9583 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9586 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9587 Accept complain bool parameter. Only emit errors if it is true.
9588 (arm_parse_cpu_option_name): Likewise.
9589 (arm_target_thumb_only): Adjust callers of the above.
9590 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9591 prototype to take a default true bool parameter.
9592 (arm_parse_arch_option_name): Likewise.
9594 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9595 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9599 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9600 * configure: Regenerate.
9601 * configure.ac ("linker --version-script option"): New.
9602 ("linker soname option"): New.
9604 2018-03-09 Richard Biener <rguenther@suse.de>
9606 PR tree-optimization/84775
9607 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9608 immediate uses of predicate stmts and mark them modified.
9611 PR tree-optimization/84178
9612 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9614 (version_loop_for_if_conversion): Delay update_ssa call.
9615 (tree_if_conversion): Delay update_ssa until after predicate
9618 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9621 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9622 when the function accesses prior frames.
9624 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9627 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9628 gen_llsym, otherwise call maybe_gen_llsym.
9631 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9632 has ',' character inside of it.
9634 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9637 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9638 as clobbering CC_REGNUM.
9640 2018-03-08 Richard Biener <rguenther@suse.de>
9643 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9644 (follow_copies_to_constant): Do not follow SSA names registered
9647 2018-03-08 Richard Biener <rguenther@suse.de>
9649 PR tree-optimization/84178
9650 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9652 (version_loop_for_if_conversion): Delay update_ssa call.
9653 (tree_if_conversion): Delay update_ssa until after predicate
9656 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9658 PR tree-optimization/84178
9659 * tree-if-conv.c (release_bb_predicate): Remove the
9660 the assertion that the stmts have NULL use_ops.
9661 Discard the statements, asserting that they haven't
9662 yet been added to a BB.
9664 2018-03-08 Richard Biener <rguenther@suse.de>
9666 PR tree-optimization/84746
9667 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9668 (phi_translate): Pass in destination ANTIC_OUT set.
9669 (phi_translate_1): Likewise. For a simplified result lookup
9670 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9671 (phi_translate_set): Adjust.
9672 (do_pre_regular_insertion): Likewise.
9673 (do_pre_partial_partial_insertion): Likewise.
9675 2018-03-08 Martin Liska <mliska@suse.cz>
9677 PR gcov-profile/84735
9678 * doc/gcov.texi: Document usage of profile files.
9679 * gcov-io.h: Document changes in the format.
9681 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9685 * dwarf2out.c (struct dw_line_info_table): Update comments for
9687 (FORCE_RESET_NEXT_VIEW): New.
9688 (FORCE_RESETTING_VIEW_P): New.
9689 (RESETTING_VIEW_P): Check for -1 too.
9690 (ZERO_VIEW_P): Likewise.
9691 (new_line_info_table): Force-reset next view.
9692 (dwarf2out_begin_function): Likewise.
9693 (dwarf2out_source_line): Simplify zero_view_p initialization.
9694 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9695 view directly. Omit view when omitting .loc at line 0.
9697 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9699 PR tree-optimization/84740
9700 * tree-switch-conversion.c (process_switch): Call build_constructors
9701 only if info.phi_count is non-zero.
9703 PR tree-optimization/84739
9704 * tree-tailcall.c (find_tail_calls): Check call arguments against
9705 DECL_ARGUMENTS (current_function_decl) rather than
9706 DECL_ARGUMENTS (func) when checking for tail recursion.
9708 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9710 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9711 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9712 Volker Reichelt's entry and add entries for people that perform
9713 GCC fuzzy testing and report numerous bugs.
9715 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9718 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9719 readonly data in sdata, if that is disabled.
9720 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9721 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9722 -mreadonly-in-sdata option.
9724 2018-03-07 Martin Sebor <msebor@redhat.com>
9726 PR tree-optimization/84468
9727 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9728 basic block when looking for nul assignment.
9730 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9733 * except.h (output_function_exception_table): Adjust prototype.
9734 * except.c (output_function_exception_table): Remove FNNAME parameter
9735 and add SECTION parameter. Ouput one part of the table at a time.
9736 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9737 the first part of the exception table and emit unwind directives.
9738 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9739 (i386_pe_seh_cold_init): Likewise.
9740 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9741 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9742 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9743 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9744 * config/i386/winnt.c: Include except.h.
9745 (struct seh_frame_state): Add reg_offset, after_prologue and
9746 in_cold_section fields.
9747 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9748 (i386_pe_seh_cold_init): New function.
9749 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9750 to seh->in_cold_section.
9751 (seh_emit_push): Record the offset of the push.
9752 (seh_emit_save): Record the offet of the save.
9753 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9754 Test seh->after_prologue to disregard the epilogue.
9755 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9756 (i386_pe_end_cold_function): New function.
9758 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9761 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9762 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9765 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9767 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9768 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9771 * multiple_target.c: Include tree-inline.h and intl.h.
9772 (expand_target_clones): Diagnose and fail if node->definition and
9773 !tree_versionable_function_p (node->decl).
9775 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9777 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9779 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9780 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9781 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9783 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9786 * combine.c (try_combine): Use reg_or_subregno instead of handling
9787 just paradoxical SUBREGs and REGs.
9789 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9791 * config/arc/arc.c (arc_finalize_pic): Remove function.
9792 (arc_must_save_register): We use single base PIC register, remove
9793 checks to save/restore the PIC register.
9794 (arc_expand_prologue): Likewise.
9795 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9797 (arc_verify_short): Likewise.
9798 (arc_attr_type): Likewise.
9799 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9800 (walk_stores): Likewise.
9801 (arc_address_cost): Make it static.
9802 (arc_verify_short): Likewise.
9803 (branch_dest): Likewise.
9804 (arc_attr_type): Likewise.
9805 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9806 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9807 (arc_final_prescan_insn): Remove inserting the nops due to
9808 hardware hazards. It is done in reorg step.
9809 (insn_length_variant_t): Remove.
9810 (insn_length_parameters_t): Likewise.
9811 (arc_insn_length_parameters): Likewise.
9812 (arc_get_insn_variants): Likewise.
9813 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9815 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9818 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9821 PR tree-optimization/84687
9822 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9824 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9826 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9828 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9829 Rename to ppc_speculation_barrier.
9830 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9831 __builtin_ppc_speculation_barrier.
9833 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9836 * combine.c (combine_simplify_rtx): Don't try to simplify if
9837 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9840 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9842 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9843 to 32 bytes when compiling for POWER9.
9845 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9848 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9849 regparm >= 3 with no arg reg available also for calls with
9850 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9853 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9855 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9857 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9860 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9862 2018-03-05 Richard Biener <rguenther@suse.de>
9864 PR tree-optimization/84486
9865 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9866 When inserting a __builtin_assume_aligned call set the LHS
9867 SSA name alignment info accordingly.
9869 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9871 PR tree-optimization/84114
9872 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9873 Avoid reassociation of FLOAT_MODE addition.
9875 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9877 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9878 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9879 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9880 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9881 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9882 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9883 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9885 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9886 __builtin_ia32_wbinvd): New builtins.
9887 (SPECIAL_ARGS2): New.
9888 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9889 (SPECIAL_ARGS2): New.
9890 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9891 (ix86_valid_target_attribute_inner_p): Ditto.
9892 (ix86_init_mmx_sse_builtins): Add special_args2.
9893 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9894 TARGET_WBNOINVD_P): New.
9895 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9896 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9897 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9898 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9899 * config/i386/pconfigintrin.h: New file.
9900 * config/i386/wbnoinvdintrin.h: Ditto.
9901 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9903 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9905 2018-03-05 Richard Biener <rguenther@suse.de>
9907 PR tree-optimization/84670
9908 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9910 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9911 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9912 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9913 doesn't grow if all successors (recursively) were visited at least
9916 2018-03-05 Richard Biener <rguenther@suse.de>
9918 PR tree-optimization/84650
9919 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9920 if executed in the loop pipeline.
9922 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9924 * doc/configfiles.texi (Configuration Files): Move info about
9925 conditionalizing $target-protos.h to...
9926 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9927 differs from $target-protos.h.
9929 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9930 Chung-Ju Wu <jasonwucj@gmail.com>
9932 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9933 * config/nds32/nds32-multiple.md (setmemsi): Define.
9934 * config/nds32/nds32-memory-manipulation.c
9935 (nds32_gen_dup_4_byte_to_word_value): New.
9936 (emit_setmem_word_loop): New.
9937 (emit_setmem_byte_loop): New.
9938 (nds32_expand_setmem_loop): New.
9939 (nds32_expand_setmem_loop_v3m): New.
9940 (nds32_expand_setmem_unroll): New.
9941 (nds32_expand_setmem): New.
9943 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9944 Chung-Ju Wu <jasonwucj@gmail.com>
9946 * config/nds32/nds32-memory-manipulation.c
9947 (nds32_emit_load_store): New.
9948 (nds32_emit_post_inc_load_store): New.
9949 (nds32_emit_mem_move): New.
9950 (nds32_emit_mem_move_block): New.
9951 (nds32_expand_movmemsi_loop_unknown_size): New.
9952 (nds32_expand_movmemsi_loop_known_size): New.
9953 (nds32_expand_movmemsi_loop): New.
9954 (nds32_expand_movmemsi_unroll): New.
9955 (nds32_expand_movmemqi): Rename ...
9956 (nds32_expand_movmemsi): ... to this.
9957 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9958 (movmemsi): ... to this.
9959 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9960 (nds32_expand_movmemsi): ... to this.
9962 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9963 Monk Chiang <sh.chiang04@gmail.com>
9964 Chung-Ju Wu <jasonwucj@gmail.com>
9966 * config/nds32/nds32-protos.h
9967 (nds32_expand_load_multiple): New arguments.
9968 (nds32_expand_store_multiple): Ditto.
9969 (nds32_valid_multiple_load_store): Rename ...
9970 (nds32_valid_multiple_load_store_p): ... to this.
9971 * config/nds32/nds32-memory-manipulation.c
9972 (nds32_expand_load_multiple): Refine implementation.
9973 (nds32_expand_store_multiple): Ditto.
9974 * config/nds32/nds32-multiple.md
9975 (load_multiple): Update nds32_expand_load_multiple interface.
9976 (store_multiple): Update nds32_expand_store_multiple interface.
9977 * config/nds32/nds32-predicates.c
9978 (nds32_valid_multiple_load_store): Rename ...
9979 (nds32_valid_multiple_load_store_p): ... to this and refine
9981 * config/nds32/predicates.md
9982 (nds32_load_multiple_and_update_address_operation): New predicate.
9983 (nds32_store_multiple_and_update_address_operation): New predicate.
9985 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9986 Chung-Ju Wu <jasonwucj@gmail.com>
9988 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9989 (combo): New attribute.
9990 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9992 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9994 * config/nds32/nds32.opt: Change -mcmodel= default value.
9996 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9997 Monk Chiang <sh.chiang04@gmail.com>
9998 Chung-Ju Wu <jasonwucj@gmail.com>
10000 * config/nds32/constants.md (unspec_element): New enum.
10001 * config/nds32/constraints.md (Umw): New constraint.
10002 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
10003 * config/nds32/nds32-intrinsic.md: Likewise.
10004 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
10005 (nds32_valid_smw_lwm_base_p): New.
10006 (nds32_output_smw_single_word): New.
10007 (nds32_output_lmw_single_word): New.
10008 (nds32_expand_unaligned_load): New.
10009 (nds32_expand_unaligned_store): New.
10010 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
10011 (nds32_output_smw_single_word): Declare.
10012 (nds32_output_lmw_single_word): Declare.
10013 (nds32_expand_unaligned_load): Declare.
10014 (nds32_expand_unaligned_store): Declare.
10015 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
10016 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
10017 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
10018 NDS32_BUILTIN_UASTORE_DW.
10019 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
10022 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
10023 Kito Cheng <kito.cheng@gmail.com>
10024 Chung-Ju Wu <jasonwucj@gmail.com>
10026 * config/nds32/nds32-intrinsic.c
10027 (nds32_expand_builtin_null_ftype_reg): Delete.
10028 (nds32_expand_builtin_reg_ftype_imm): Ditto.
10029 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
10030 (nds32_read_argument): New.
10031 (nds32_legitimize_target): Ditto.
10032 (nds32_legitimize_argument): Ditto.
10033 (nds32_check_constant_argument): Ditto.
10034 (nds32_expand_unop_builtin): Ditto.
10035 (nds32_expand_unopimm_builtin): Ditto.
10036 (nds32_expand_binop_builtin): Ditto.
10037 (nds32_builtin_decl_impl): Ditto.
10038 (builtin_description): Ditto.
10039 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
10040 (nds32_init_builtins_impl): Ditto.
10041 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
10042 (nds32_builtin_decl): New.
10043 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
10044 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
10046 2018-03-02 Jeff Law <law@redhat.com>
10048 * reorg.c (stop_search_p): Handle DEBUG_INSN.
10049 (redundant_insn, fill_simple_delay_slots): Likewise.
10050 (fill_slots_from_thread): Likewise.
10051 * resource.c (mark_referenced_resources): Likewise.
10052 (mark_set_resources, find_dead_or_set_registers): Likewise.
10054 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10056 * substring-locations.h (format_warning_va): Formatting fix for
10057 ATTRIBUTE_GCC_DIAG.
10058 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
10060 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
10061 * substring-locations.c: Include intl.h.
10062 (format_warning_va): Turned into small wrapper around
10063 format_warning_n_va, renamed to ...
10064 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
10065 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
10067 (format_warning_at_substring_n): New function.
10068 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
10069 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
10070 format_warning_at_substring with just a shorter name instead of
10071 const function pointer.
10072 (fmtwarn_n): New function.
10073 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
10074 appropriate, get rid of all the fmtstr temporaries, move conditionals
10075 with G_() wrapped string literals directly into fmtwarn arguments,
10076 cast dir.len to (int), formatting fixes.
10078 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
10080 * doc/invoke.texi: Remove "Cilk Plus" references.
10082 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10083 Richard Biener <rguenther@suse.de>
10086 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
10087 for error or warning attributes if CALL_FROM_THUNK_P is set.
10090 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10093 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
10094 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
10097 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
10098 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
10100 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
10101 instead of -1U in last predictors element's probability member.
10103 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10106 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10107 arguments if they are comparable.
10109 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10111 PR tree-optimization/84634
10112 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
10113 masks and masked_loop_p with a single loop_masks, making sure it's
10114 null for bb vectorization.
10116 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10118 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10119 (vect_analyze_data_ref_access): Use loop->safe_len rather than
10120 loop->force_vectorize to check whether there is no alias.
10122 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10125 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
10127 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
10129 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
10130 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
10131 instead of a loop around prev_real_insn.
10132 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
10135 PR inline-asm/84625
10136 * config/i386/i386.c (ix86_print_operand): Use conditional
10137 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
10140 2018-03-02 Richard Biener <rguenther@suse.de>
10142 PR tree-optimization/84427
10143 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
10144 (bitmap_set_subtract_values): Rewrite to handle multiple
10147 (prune_clobbered_mems): Likewise.
10148 (phi_translate): Take edge instead of pred/phiblock.
10149 (phi_translate_1): Likewise.
10150 (phi_translate_set): Likewise. Insert all translated
10151 exprs for a value into the set, keeping possibly multiple
10152 expressions per value.
10153 (compute_antic_aux): Adjust for phi_translate changes.
10154 When intersecting union the expressions and prune those
10155 not in the final value set, keeping possibly multiple
10156 expressions per value. Do not use value-insertion
10157 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
10158 all expressions. Add verification that the value-sets
10159 only shrink during iteration.
10160 (compute_partial_antic_aux): Adjust for the phi_translate changes.
10161 (do_pre_regular_insertion): Likewise.
10162 (do_pre_partial_partial_insertion): Likewise.
10164 2018-03-02 Richard Biener <rguenther@suse.de>
10167 * config/darwin.c (saved_debug_info_level): New static global.
10168 (darwin_asm_lto_start): Disable debug info generation for LTO out.
10169 (darwin_asm_lto_end): Restore debug info generation settings.
10171 2018-03-01 Martin Liska <mliska@suse.cz>
10174 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
10175 volatile arguments.
10177 2018-03-01 Richard Biener <rguenther@suse.de>
10180 * dwarf2out.c (gen_variable_die): Properly handle late VLA
10181 type annotation with LTO when debug was disabled at compile-time.
10183 2018-03-01 Matthew Fortune <mfortune@gmail.com>
10185 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
10187 (mips_final_postscan_insn): Likewise.
10189 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
10191 PR rtl-optimization/84528
10192 * alias.c (init_alias_target): Add commentary.
10193 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
10194 a unique base value if the frame pointer is not eliminated
10195 to the stack pointer.
10197 2018-03-01 Tom de Vries <tom@codesourcery.com>
10199 PR rtl-optimization/83327
10200 * lra-int.h (hard_regs_spilled_into): Declare.
10201 * lra.c (hard_regs_spilled_into): Define.
10202 (init_reg_info): Init hard_regs_spilled_into.
10203 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
10204 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
10205 (process_bb_lives): Handle hard_regs_spilled_into.
10206 (lra_create_live_ranges_1): Before doing liveness propagation, clear
10207 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
10209 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
10211 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
10212 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
10213 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
10214 * config/rs6000/aix72.h: New file.
10216 2018-02-28 Jakub Jelinek <jakub@redhat.com>
10218 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
10219 instead of warning_at with conditional singular and plural messages
10223 * stor-layout.c (update_alignment_for_field): For
10224 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
10225 && !DECL_PACKED (field), do the alignment update, just use
10226 only desired_align instead of MAX (type_align, desired_align)
10228 (place_field): Don't do known_align < desired_align handling
10229 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
10230 is non-NULL, instead do it after rli->prev_field handling and
10231 only if not within a bitfield word. For DECL_PACKED (field)
10232 use type_align of BITS_PER_UNIT.
10234 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
10236 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
10237 superfluous parentheses and trailing spaces.
10239 2018-02-28 Richard Biener <rguenther@suse.de>
10241 PR tree-optimization/84584
10242 * graphite-scop-detection.c (scop_detection::add_scop): Discard
10243 SCoPs with fake exit edge.
10245 2018-02-28 Martin Liska <mliska@suse.cz>
10248 * timevar.c (timer::print): Fix format to properly print 100%
10251 2018-02-28 Richard Biener <rguenther@suse.de>
10253 PR middle-end/84607
10254 * genmatch.c (capture_info::walk_match): Do not mark
10255 captured expressions without operands as expr_p given
10256 they act more like predicates and should be subject to
10257 "lost tail" side-effect preserving.
10259 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
10261 PR rtl-optimization/81611
10262 * auto-inc-dec.c (attempt_change): Move dead note from
10263 mem_insn if it's the next use of regno
10264 (find_address): Take address use of reg holding
10265 non-incremented value. Add parm to limit search to the named
10267 (merge_in_block): Attempt to use a mem insn that is the next
10268 use of the original regno.
10270 2018-02-27 Martin Sebor <msebor@redhat.com>
10273 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
10274 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
10276 2018-02-27 Martin Sebor <msebor@redhat.com>
10278 PR translation/84207
10279 * diagnostic-core.h (warning_n, error_n, inform_n): Change
10280 n argument to unsigned HOST_WIDE_INT.
10281 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
10282 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
10283 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
10284 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
10286 2018-02-27 Richard Biener <rguenther@suse.de>
10288 PR tree-optimization/84512
10289 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
10290 Do not use the estimate returned from record_stmt_cost for
10291 the scalar iteration cost but sum properly using add_stmt_cost.
10293 2018-02-27 Richard Biener <rguenther@suse.de>
10295 PR tree-optimization/84466
10296 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10297 Adjust last change to less strictly validate use operands.
10299 2018-02-27 Martin Liska <mliska@suse.cz>
10301 PR gcov-profile/84548
10302 * gcov.c (process_file): Allow partial overlap and consider it
10303 also as group functions.
10304 (output_lines): Properly calculate range of lines for a group.
10306 2018-02-27 Martin Liska <mliska@suse.cz>
10308 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
10309 'ggc' suffixes. Change first column width.
10310 (timer::print): Fix formatting of the column.
10312 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
10314 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
10315 preserve inline entry blocks for the sake of debug inline
10316 entry point markers alone.
10317 (remove_unused_locals): Suggest in comments a better place to
10318 force the preservation of inline entry blocks that are
10319 otherwise unused, but do not preserve them.
10321 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10323 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
10325 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10328 * config/i386/constraints.md (Bs): Replace
10329 ix86_indirect_branch_register with
10330 TARGET_INDIRECT_BRANCH_REGISTER.
10332 * config/i386/i386.md (indirect_jump): Likewise.
10333 (tablejump): Likewise.
10334 (*sibcall_memory): Likewise.
10335 (*sibcall_value_memory): Likewise.
10336 Peepholes of indirect call and jump via memory: Likewise.
10337 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
10338 (*sibcall_value_GOT_32): Likewise.
10339 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10340 (GOT_memory_operand): Likewise.
10341 (call_insn_operand): Likewise.
10342 (sibcall_insn_operand): Likewise.
10343 (GOT32_symbol_operand): Likewise.
10344 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
10346 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10348 PR rtl-optimization/83496
10349 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
10350 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
10351 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
10352 redundant insn, if any.
10353 (relax_delay_slots): Likewise.
10354 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
10356 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
10358 PR tree-optimization/83965
10359 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
10360 that grouped statements are part of a reduction chain. Return
10361 true if the statement is not marked as a reduction itself but
10362 is part of a group.
10363 (vect_recog_dot_prod_pattern): Don't check whether the statement
10364 is part of a group here.
10365 (vect_recog_sad_pattern): Likewise.
10366 (vect_recog_widen_sum_pattern): Likewise.
10368 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10371 * final.c (rest_of_clean_state): Also look for calls inside sequences.
10373 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10376 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
10378 (ix86_output_indirect_function_return): New prototype.
10379 (ix86_split_simple_return_pop_internal): Likewise.
10380 * config/i386/i386.c (indirect_return_via_cx): New.
10381 (indirect_return_via_cx_bnd): Likewise.
10382 (indirect_thunk_name): Handle return va CX_REG.
10383 (output_indirect_thunk_function): Create alias for
10384 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
10385 (ix86_output_indirect_jmp): Remove the bool argument.
10386 (ix86_output_indirect_function_return): New function.
10387 (ix86_split_simple_return_pop_internal): Likewise.
10388 * config/i386/i386.md (*indirect_jump): Don't pass false
10389 to ix86_output_indirect_jmp.
10390 (*tablejump_1): Likewise.
10391 (simple_return_pop_internal): Change it to define_insn_and_split.
10392 Call ix86_split_simple_return_pop_internal to split it for
10393 -mfunction-return=.
10394 (simple_return_indirect_internal): Call
10395 ix86_output_indirect_function_return instead of
10396 ix86_output_indirect_jmp.
10398 2018-02-26 Jakub Jelinek <jakub@redhat.com>
10401 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
10402 memset and value initialization afterwards.
10404 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
10406 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
10408 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10411 * common/config/aarch64/aarch64-common.c
10412 (aarch_option_optimization_table[]): Switch
10413 off fomit-frame-pointer
10415 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10416 Chung-Ju Wu <jasonwucj@gmail.com>
10418 * config/nds32/nds32-multiple.md (load_multiple): Disallow
10420 (store_multiple): Ditto.
10422 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10424 * config.gcc: Add --with-cpu support for nds32 target.
10425 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
10426 * config/nds32/nds32.opt: Add -mcpu= option.
10428 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
10430 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
10431 isel=yes): Warn for these deprecated options.
10433 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
10435 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
10436 ISA_2_5_MASKS_EMBEDDED.
10438 2018-02-23 Jakub Jelinek <jakub@redhat.com>
10440 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
10441 p->max as pointers rather than using iterative_hash_expr.
10443 2018-02-23 Carl Love <cel@us.ibm.com>
10445 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
10446 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
10448 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
10449 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
10450 P8V_BUILTIN_VEC_VUNSIGNED2.
10452 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
10455 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
10456 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
10458 * lra-constraints.c (curr_insn_transform): Set up
10459 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
10461 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
10462 * lra-eliminations.c (spill_pseudos): Ditto.
10463 (process_insn_for_elimination): Ditto.
10464 * lra-lives.c (reg_early_clobber_p): Use the new macros.
10465 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
10466 LRA_NON_CLOBBERED_ALT.
10468 2018-02-22 Martin Sebor <msebor@redhat.com>
10470 PR tree-optimization/84480
10471 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
10472 to maybe_diag_stxncpy_trunc. Call it.
10473 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
10474 from gimple_fold_builtin_strcpy. Print inlining stack.
10475 (handle_builtin_stxncpy): Print inlining stack.
10476 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
10478 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
10481 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
10482 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
10483 and -fcheck-pointer-bounds are used together.
10484 (indirect_thunk_prefix): New enum.
10485 (indirect_thunk_need_prefix): New function.
10486 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
10487 "_nt" instead of "_bnd" for NOTRACK prefix.
10488 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
10489 (output_indirect_thunk_function): Likewise.
10491 (ix86_code_end): Update output_indirect_thunk_function calls.
10492 (ix86_output_indirect_branch_via_reg): Replace
10493 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
10494 (ix86_output_indirect_branch_via_push): Likewise.
10495 (ix86_output_function_return): Likewise.
10496 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
10497 incompatible with -fcf-protection=branch and
10498 -fcheck-pointer-bounds.
10500 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10503 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10504 Change gcc_assert call to output_operand_lossage.
10506 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10508 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10510 2018-02-22 DJ Delorie <dj@redhat.com>
10511 Sebastian Perta <sebastian.perta@renesas.com>
10512 Oleg Endo <olegendo@gcc.gnu.org>
10514 * config/rx/rx.c (rx_rtx_costs): New function.
10515 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
10517 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
10519 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
10521 2018-02-22 Martin Liska <mliska@suse.cz>
10524 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
10525 Add "native" as a possible value.
10527 2018-02-22 Martin Liska <mliska@suse.cz>
10530 * config/i386/i386.c (ix86_option_override_internal):
10531 Add "native" as a possible value for -march and -mtune.
10533 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10536 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10537 to all type variants.
10539 PR tree-optimization/84503
10540 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10541 width as info->bitpos + info->bitsize - start.
10542 (merged_store_group::merge_overlapping): Simplify width computation.
10543 (check_no_overlap): New function.
10544 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10545 start + width and last_order of the group, fail if check_no_overlap
10547 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10548 to group if check_no_overlap fails.
10550 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10552 * config/rs6000/altivec.md: Delete contraint arguments to
10553 define_expand, define_split, and define_peephole2, and in
10554 define_insn_and_split if always unused.
10555 * config/rs6000/darwin.md: Ditto.
10556 * config/rs6000/dfp.md: Ditto.
10557 * config/rs6000/rs6000.md: Ditto.
10558 * config/rs6000/sync.md: Ditto.
10559 * config/rs6000/vector.md: Ditto.
10560 * config/rs6000/vsx.md: Ditto.
10562 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10564 * config/rs6000/altivec.md: Write output control strings as braced
10565 blocks instead of double-quoted strings.
10566 * config/rs6000/darwin.md: Ditto.
10567 * config/rs6000/rs6000.md: Ditto.
10568 * config/rs6000/vector.md: Ditto.
10569 * config/rs6000/vsx.md: Ditto.
10571 2018-02-21 Jason Merrill <jason@redhat.com>
10573 PR c++/84314 - ICE with templates and fastcall attribute.
10574 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10576 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10578 * ipa-cp.c (determine_versionability): Fix comment typos.
10580 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10583 * ipa-cp.c (determine_versionability): Do not version functions caling
10586 2018-02-21 Martin Liska <mliska@suse.cz>
10589 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10590 Add "native" as a possible value.
10591 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10592 the macro when native cpu detection is available.
10594 2018-02-21 Martin Liska <mliska@suse.cz>
10597 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10598 Add "native" as a possible value.
10599 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10600 when native cpu detection is available.
10602 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10603 Martin Sebor <msebor@redhat.com>
10605 PR tree-optimization/84478
10606 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10608 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10609 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10610 support which is conservatively correct, for 2 only stay conservative
10611 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10612 argument to the 2 argument get_range_strlen, adjust 6 arg
10613 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10615 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10616 (gimple_fold_builtin_strlen): Pass true as last argument to
10619 2018-02-20 Martin Sebor <msebor@redhat.com>
10621 PR middle-end/84095
10622 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10623 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10624 (builtin_memref::builtin_memref): Factor out parts into
10625 set_base_and_offset and call it.
10627 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10629 PR middle-end/84406
10630 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10631 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10632 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10633 search at the associated MODE_INT.
10635 2018-02-20 Jeff Law <law@redhat.com>
10637 PR middle-end/82123
10638 PR tree-optimization/81592
10639 PR middle-end/79257
10640 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10641 for range data rather than using global data.
10642 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10643 range data rather than using global data.
10644 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10645 pass it to children as needed.
10646 (struct directive::fmtresult): Similarly.
10647 (struct directive::set_width): Similarly.
10648 (struct directive::set_precision): Similarly.
10649 (format_integer, format_directive, parse_directive): Similarly.
10650 (format_none): Accept unnamed vr_values parameter.
10651 (format_percent, format_floating, format_character): Similarly.
10652 (format_string, format_plain): Similarly.
10653 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10654 the EVRP range analyzer for range data rather than using global data.
10655 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10656 gimple-ssa-evrp-analyze.h
10657 (class sprintf_dom_walker): Add after_dom_children member function.
10658 Add evrp_range_analyzer member.
10659 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10660 range analyzer as needed.
10661 (sprintf_dom_walker::after_dom_children): New member function.
10662 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10664 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10665 (evrp_range_analyzer::pop_to_marker): Likewise.
10667 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10669 PR tree-optimization/84419
10670 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10671 with the required type if its current type is compatible but
10674 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10676 PR middle-end/82004
10677 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10678 after vectorization.
10680 2018-02-20 Martin Liska <mliska@suse.cz>
10683 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10684 possible values if we don't have a hint.
10686 2018-02-20 Martin Liska <mliska@suse.cz>
10690 * final.c (shorten_branches): Build align_tab array with one
10692 * opts.c (finish_options): Add alignment option limit check.
10693 (MAX_CODE_ALIGN): Likewise.
10694 (MAX_CODE_ALIGN_VALUE): Likewise.
10695 * doc/invoke.texi: Document maximum allowed option value for
10696 all -falign-* options.
10698 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10701 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10702 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10703 * var-tracking.c (emit_note_insn_var_location): Remove all references
10704 to NOTE_INSN_CALL_ARG_LOCATION.
10705 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10706 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10707 Use copy_rtx_if_shared.
10708 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10709 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10710 (dwarf2out_var_location): Remove handling of
10711 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10713 * final.c (final_scan_insn): Remove all references to
10714 NOTE_INSN_CALL_ARG_LOCATION.
10715 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10716 before dumping final insns.
10717 * except.c (emit_note_eh_region_end): Remove all references to
10718 NOTE_INSN_CALL_ARG_LOCATION.
10719 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10720 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10721 * config/arc/arc.c (hwloop_optimize): Likewise.
10722 * config/arm/arm.c (create_fix_barrier): Likewise.
10723 * config/s390/s390.c (s390_chunkify_start): Likewise.
10724 * config/sh/sh.c (find_barrier): Likewise.
10725 * config/i386/i386.c (rest_of_insert_endbranch,
10726 ix86_seh_fixup_eh_fallthru): Likewise.
10727 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10728 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10729 * config/frv/frv.c (frv_function_prologue): Likewise.
10730 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10732 (note_outside_basic_block_p): Remove all references to
10733 NOTE_INSN_CALL_ARG_LOCATION.
10734 * gengtype.c (adjust_field_rtx_def): Likewise.
10735 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10737 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10738 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10741 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10744 PR tree-optimization/84452
10745 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10746 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10749 2018-02-19 Martin Liska <mliska@suse.cz>
10752 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10754 2018-02-19 Martin Liska <mliska@suse.cz>
10755 Richard Sandiford <richard.sandiford@linaro.org>
10757 PR tree-optimization/82491
10758 * gimple-fold.c (get_base_constructor): Make earlier bail out
10761 2018-02-19 Carl Love <cel@us.ibm.com>
10763 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10764 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10766 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10767 P8V_BUILTIN_VEC_NEG.
10769 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10771 * config/rl78/rl78.md (movdf): New define expand.
10773 2018-02-19 Martin Liska <mliska@suse.cz>
10776 * doc/invoke.texi: Fix typo.
10777 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10779 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10781 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10782 handle rs6000_single_float and rs6000_double_float specially for
10785 2018-02-16 Jeff Law <law@redhat.com>
10787 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10788 (pop_regs): Use it.
10790 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10793 * ipa-inline.c (inline_small_functions): Fix a typo.
10795 2018-02-16 Nathan Sidwell <nathan@acm.org>
10797 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10799 2018-02-16 Carl Love <cel@us.ibm.com>
10801 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10802 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10803 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10804 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10805 expansion to P8V_BUILTIN_VEC_FLOAT2.
10807 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10809 PR rtl-optimization/70023
10810 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10811 src_regno into account.
10813 2018-02-16 Carl Love <cel@us.ibm.com>
10815 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10816 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10817 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10818 * config/rs6000/rs6000.c: Remove case statements for
10819 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10820 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10821 and P9V_BUILTIN_VEC_VINSERT4B.
10822 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10823 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10824 * config/rs6000/vsx.md:
10825 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10828 2018-02-16 Carl Love <cel@us.ibm.com>
10830 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10832 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10834 * config/rs6000/rs6000-c.c: Add the definitions for
10835 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10836 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10837 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10838 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10839 definition for insert4b and define insn *insert3b_internal.
10840 * doc/extend.texi: Add documentation for vec_extract4b.
10842 2018-02-16 Nathan Sidwell <nathan@acm.org>
10844 * doc/extend.texi (Backwards Compatibility): Mention friend
10845 injection. Note for-scope is deprecated.
10846 * doc/invoke.texi (-ffriend-injection): Deprecate.
10848 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10850 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10851 that moved to I2, also allow destinations that are a paradoxical
10852 subreg (instead of a normal reg).
10854 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10857 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10860 2018-02-16 Richard Biener <rguenther@suse.de>
10862 PR tree-optimization/84037
10863 PR tree-optimization/84016
10865 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10866 Adjust vec_construct for the fact we need additional higher latency
10867 128bit inserts for AVX256 and AVX512 vector builds.
10868 (ix86_add_stmt_cost): Scale vector construction cost for
10871 2018-02-16 Richard Biener <rguenther@suse.de>
10873 PR tree-optimization/84417
10874 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10875 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10876 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10878 2018-02-16 Martin Liska <mliska@suse.cz>
10881 * internal-fn.def (ASAN_CHECK): Set proper flags.
10882 (ASAN_MARK): Likewise.
10884 2018-02-16 Julia Koval <julia.koval@intel.com>
10886 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10887 from PTA_CANNONLAKE.
10889 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10892 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10893 Use ++iter rather than iter++ for std::list iterators.
10894 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10895 defer deleting them until all nodes in the forest are processed. Do
10896 free even leaf nodes. Change to_process into auto_vec.
10899 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10900 * vec.h (vec_default_construct): Use memset instead of placement new
10901 if BROKEN_VALUE_INITIALIZATION is defined.
10902 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10903 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10906 PR rtl-optimization/83723
10907 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10908 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10909 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10911 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10913 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10915 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10917 PR rtl-optimization/81443
10918 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10919 from inner REGs to paradoxical SUBREGs.
10921 2018-02-16 Richard Biener <rguenther@suse.de>
10923 PR tree-optimization/84399
10924 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10925 For operands we can analyze at their definition make sure we can
10926 analyze them at each use as well.
10928 2018-02-16 Richard Biener <rguenther@suse.de>
10930 PR tree-optimization/84190
10931 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10932 volatile accesses if the decl isn't volatile.
10934 2018-02-15 Jason Merrill <jason@redhat.com>
10936 PR c++/84314 - ICE with templates and fastcall attribute.
10937 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10938 TYPE_CANONICAL on an existing type.
10940 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10942 PR tree-optimization/84383
10943 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10944 dstoff nor call operand_equal_p if dstbase is NULL.
10946 PR tree-optimization/84334
10947 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10948 also a CONSTANT_CLASS_P, punt.
10950 2018-02-14 Jim Wilson <jimw@sifive.com>
10952 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10953 first SMALL_OPERAND check. New local min_second_step. Move assert
10954 to where locals are set. Add TARGET_RVC support.
10955 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10957 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10959 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10961 2018-02-14 Martin Sebor <msebor@redhat.com>
10963 PR tree-optimization/83698
10964 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10965 arrays constrain the offset range to their bounds.
10966 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10967 (builtin_access::overlap): Avoid setting the size of overlap if it's
10969 (maybe_diag_overlap): Also consider arrays when deciding what values
10970 of offsets to include in diagnostics.
10972 2018-02-14 Martin Sebor <msebor@redhat.com>
10975 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10976 that correspond to the kind of a declaration.
10978 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10981 * config/pa/pa.md: Load address of PIC label using the linkage table
10982 if the label is nonlocal.
10984 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10986 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10987 warning message if user requests -maltivec=be.
10988 * doc/invoke.texi: Document deprecation of -maltivec=be.
10990 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10993 * config/rs6000/rs6000-c.c: Update definitions for
10994 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10995 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10997 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11000 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
11001 add _get_ssp intrinsics. Remove argument from
11002 __builtin_ia32_rdssp[d|q].
11003 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
11004 * config/i386/i386-builtin.def: Remove argument from
11005 __builtin_ia32_rdssp[d|q].
11006 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
11007 ix86_expand_special_args_builtin for _rdssp[d|q].
11008 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
11009 Clear register before usage.
11010 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
11011 Add documentation for new _get_ssp and _inc_ssp intrinsics.
11013 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
11015 PR tree-optimization/84357
11016 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
11017 operand 1 of an ARRAY_REF too.
11019 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
11022 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
11023 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11025 (set_of_reg): New struct.
11026 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
11027 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
11028 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11030 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
11031 Split into bitclr, bitset, bitinvert patterns if appropriate.
11032 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
11033 use rx_fuse_in_memory_bitop.
11034 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
11035 to named insn, correct maximum insn length.
11037 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
11040 * machmode.def: Define a complex mode for PARTIAL_INT.
11041 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
11043 * doc/rtl.texi: Document CSPImode.
11044 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
11046 (msp430_hard_regno_nregs_with_padding): Likewise.
11048 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
11051 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
11053 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
11055 PR rtl-optimization/84169
11056 * combine.c (try_combine): New variable split_i2i3. Set it to true if
11057 we generated a parallel as new i3 and we split that to new i2 and i3
11058 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
11059 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
11060 those to i2, not i1. Partially rewrite this scan code.
11062 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11065 * stor-layout.c (place_field): For variable length fields, adjust
11066 offset_align afterwards not just based on the field's alignment,
11067 but also on the size.
11069 PR middle-end/84309
11070 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
11071 of exps and logs in the use_exp2 case.
11073 2018-02-13 Jeff Law <law@redhat.com>
11075 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
11076 entry for "vector".
11078 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
11081 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
11085 * common.opt (gas-loc-support, gas-locview-support): New.
11086 (ginline-points, ginternal-reset-location-views): New.
11087 * doc/invoke.texi: Document them. Use @itemx where intended.
11088 (gvariable-location-views): Adjust.
11089 * target.def (reset_location_view): New.
11090 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
11091 (TARGET_RESET_LOCATION_VIEW): New.
11092 * doc/tm.texi: Rebuilt.
11093 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
11094 (dwarf2out_default_as_locview_support): New.
11095 (output_asm_line_debug_info): Use option variables.
11096 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
11097 (output_loc_list): Likewise.
11098 (add_high_low_attributes): Check option variables.
11099 Don't output entry view attribute in strict mode.
11100 (gen_inlined_subroutine_die): Check option variables.
11101 (dwarf2out_inline_entry): Likewise.
11102 (init_sections_and_labels): Likewise.
11103 (dwarf2out_early_finish): Likewise.
11104 (maybe_reset_location_view): New, from...
11105 (dwarf2out_var_location): ... here. Call it.
11106 * debug.h (dwarf2out_default_as_loc_support): Declare.
11107 (dwarf2out_default_as_locview_support): Declare.
11108 * hooks.c (hook_int_rtx_insn_0): New.
11109 * hooks.h (hook_int_rtx_insn_0): Declare.
11110 * toplev.c (process_options): Take -gas-loc-support and
11111 -gas-locview-support from dwarf2out. Enable
11112 -gvariable-location-views by default only with locview
11113 assembler support. Enable -ginternal-reset-location-views by
11114 default only if the target defines the corresponding hook.
11115 Enable -ginline-points by default if location views are
11116 enabled; force it disabled if statement frontiers are
11118 * tree-inline.c (expand_call_inline): Check option variables.
11119 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
11121 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
11123 PR tree-optimization/84321
11124 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
11125 handling. Also check whether the anti-range contains any values
11126 that satisfy the mask; switch to a VR_RANGE if not.
11128 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
11131 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
11133 2018-02-13 Martin Jambor <mjambor@suse.cz>
11136 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
11137 of call statements, also set location of a load to a temporary.
11139 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
11141 * config/rl78/rl78.c (add_vector_labels): New function.
11142 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
11143 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
11144 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
11145 which checks that no arguments are passed.
11146 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
11147 * doc/extend.texi: Documentation for the new attribute.
11149 2018-02-13 Andreas Schwab <schwab@suse.de>
11151 * config/riscv/linux.h (CPP_SPEC): Define.
11153 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11156 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
11157 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
11158 OPTION_MASK_ISA_AES as first argument to def_builtin_const
11159 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
11160 instead of OPTION_MASK_ISA_PCLMUL as first argument to
11161 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
11162 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
11163 temporarily for AES and PCLMUL builtins.
11165 PR tree-optimization/84339
11166 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
11167 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
11170 PR middle-end/84309
11171 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
11172 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
11173 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
11175 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
11177 * omp-simd-clone.h: New file.
11178 * omp-simd-clone.c: Include omp-simd-clone.h.
11179 (expand_simd_clones): No longer static.
11180 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
11181 cgraph.h and omp-simd-clone.h.
11182 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
11183 (vect_recog_widen_shift_pattern): Formatting fix.
11184 (vect_pattern_recog_1): Don't check optab for calls.
11187 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
11188 operands[2] into a REG before using gen_lowpart on it.
11190 2018-02-12 Jeff Law <law@redhat.com>
11193 * config/sh/sh.c (find_barrier): Consider a sibling call
11196 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
11197 successfully back substituting a reg.
11199 2018-02-12 Richard Biener <rguenther@suse.de>
11201 PR tree-optimization/84037
11202 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
11203 parameter, move visited init to caller.
11204 (vect_slp_analyze_operations): Separate cost from validity
11205 check, initialize visited once for all instances.
11206 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
11208 * tree-vect-stmts.c (vect_model_simple_cost): Make early
11210 (vect_model_promotion_demotion_cost): Likewise.
11211 (vectorizable_bswap): Guard cost modeling with !slp_node
11212 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
11214 (vectorizable_call): Likewise.
11215 (vectorizable_conversion): Likewise.
11216 (vectorizable_assignment): Likewise.
11217 (vectorizable_shift): Likewise.
11218 (vectorizable_operation): Likewise.
11219 (vectorizable_store): Likewise.
11220 (vectorizable_load): Likewise.
11221 (vectorizable_condition): Likewise.
11222 (vectorizable_comparison): Likewise.
11224 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
11227 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
11228 (ASAN_MARK): Fix fnspec to account for return value, change pointer
11229 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
11231 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11233 PR middle-end/83665
11234 * params.def (inline-min-speedup): Increase from 8 to 15.
11235 (max-inline-insns-auto): Decrease from 40 to 30.
11236 * ipa-split.c (consider_split): Add some buffer for function to
11237 be considered inlining candidate.
11238 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
11241 2018-02-12 Richard Biener <rguenther@suse.de>
11243 PR tree-optimization/84037
11244 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
11245 matched stmts if we cannot swap the non-matched ones.
11247 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11249 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
11250 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
11251 _mm_maskz_scalef_round_ss): New intrinsics.
11252 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
11253 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
11254 __builtin_ia32_scalefss_round): Remove.
11255 (__builtin_ia32_scalefsd_mask_round,
11256 __builtin_ia32_scalefss_mask_round): New intrinsics.
11257 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
11258 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
11259 ((match_operand:VF_128 2 "<round_nimm_predicate>"
11260 "<round_constraint>")): Changed to ...
11261 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
11262 "<round_scalar_constraint>")): ... this.
11263 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
11264 %0, %1, %2<round_op3>}"): Changed to ...
11265 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
11266 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
11267 %2<round_scalar_mask_op3>}"): ... this.
11268 * config/i386/subst.md (round_scalar_nimm_predicate): New.
11270 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11272 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
11273 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
11274 (_mm_maskz_sqrt_round_ss): New intrinsics.
11275 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
11276 (__builtin_ia32_sqrtsd_mask_round)
11277 (__builtin_ia32_sqrtss_mask_round): New builtins.
11278 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
11279 (__builtin_ia32_sqrtss_round): Remove.
11280 (__builtin_ia32_sqrtsd_mask_round)
11281 (__builtin_ia32_sqrtss_mask_round): New builtins.
11282 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
11283 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
11284 ((match_operand:VF_128 1 "vector_operand"
11285 "xBm,<round_constraint>")): Changed to ...
11286 ((match_operand:VF_128 1 "vector_operand"
11287 "xBm,<round_scalar_constraint>")): ... this.
11288 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
11289 %0, %2, %<iptr>1<round_op3>}): Changed to ...
11290 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
11291 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
11292 %<iptr>1<round_scalar_mask_op3>}): ... this.
11293 ((set_attr "prefix" "<round_prefix>")): Changed to ...
11294 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
11296 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
11299 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
11300 Cast vec_cmpeq result to correct type.
11301 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
11302 Cast vec_cmpgt result to correct type.
11304 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
11306 * final.c (final_scan_insn_1): Renamed from...
11307 (final_scan_insn): ... this. New wrapper, to recover
11308 seen from the outermost call in recursive ones.
11309 * config/sparc/sparc.c (output_return): Drop seen from call.
11310 (output_sibcall): Likewise.
11311 * config/visium/visium.c (output_branch): Likewise.
11313 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
11315 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
11318 2018-02-10 Alan Modra <amodra@gmail.com>
11321 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
11322 Specify LR as an input.
11324 2018-02-10 Jakub Jelinek <jakub@redhat.com>
11327 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
11328 remove_member_access_dummy_vars): New functions.
11329 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
11330 lower_omp_1, execute_lower_omp): Use them.
11332 PR rtl-optimization/84308
11333 * shrink-wrap.c (spread_components): Release todo vector.
11335 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
11337 PR rtl-optimization/57193
11338 * ira-color.c (struct allocno_color_data): Add member
11339 conflict_allocno_hard_prefs.
11340 (update_conflict_allocno_hard_prefs): New.
11341 (bucket_allocno_compare_func): Add a preference based on
11342 conflict_allocno_hard_prefs.
11343 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
11344 (color_allocnos): Remove a dead code. Initiate
11345 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
11347 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11350 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
11351 constraint from =wa to wa. Avoid a subreg on the output operand,
11352 instead use a pseudo and subreg it in a move.
11353 (p9_xxbrd_<mode>): Changed to ...
11354 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
11355 (p9_xxbrd_v2df): New expander.
11356 (p9_xxbrw_<mode>): Changed to ...
11357 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
11358 (p9_xxbrw_v4sf): New expander.
11360 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11362 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
11364 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
11367 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
11368 multiply in 32-bit mode.
11369 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
11370 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
11373 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11375 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
11376 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
11377 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
11378 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
11380 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
11383 * dwarf2out.c (is_trivial_indirect_ref): New function.
11384 (dwarf2out_late_global_decl): Do not generate a location
11385 attribute for variables that have a non-trivial DECL_VALUE_EXPR
11386 and that are not defined in the current unit.
11388 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11390 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
11391 instead of a libcall for UNORDERED.
11393 2018-02-09 Tamar Christina <tamar.christina@arm.com>
11396 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
11397 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
11399 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11402 * config/s390/s390.c (s390_set_current_function): Invoke
11403 s390_indirect_branch_settings also if fndecl didn't change.
11405 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11407 * config/rs6000/rs6000.md (blockage): Set length to zero.
11409 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11411 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
11413 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11416 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
11417 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
11421 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
11422 PARALLEL incoming that failed vt_get_decl_and_offset check.
11424 PR middle-end/84237
11425 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
11426 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
11428 (get_variable_section): For decls in named .bss* sections pass true as
11429 second argument to bss_initializer_p.
11431 2018-02-09 Marek Polacek <polacek@redhat.com>
11432 Jakub Jelinek <jakub@redhat.com>
11435 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
11436 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
11437 Sync some changes from cxx_fold_indirect_ref.
11439 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11441 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
11443 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
11444 (BLOCK_INLINE_ENTRY_LABEL): New.
11445 (dwarf2out_var_location): Disregard inline entry markers.
11446 (inline_entry_data): New struct.
11447 (inline_entry_data_hasher): New hashtable type.
11448 (inline_entry_data_hasher::hash): New.
11449 (inline_entry_data_hasher::equal): New.
11450 (inline_entry_data_table): New variable.
11451 (add_high_low_attributes): Add DW_AT_entry_pc and
11452 DW_AT_GNU_entry_view attributes if a pending entry is found
11453 in inline_entry_data_table. Add old entry_pc attribute only
11454 if debug nonbinding markers are disabled.
11455 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
11456 markers are enabled.
11457 (block_within_block_p, dwarf2out_inline_entry): New.
11458 (dwarf2out_finish): Check that no entries remained in
11459 inline_entry_data_table.
11460 * final.c (reemit_insn_block_notes): Handle inline entry notes.
11461 (final_scan_insn, notice_source_line): Likewise.
11462 (rest_of_clean_state): Skip inline entry markers.
11463 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
11465 * gimple.c (gimple_build_debug_inline_entry): New.
11466 * gimple.h (enum gimple_debug_subcode): Add
11467 GIMPLE_DEBUG_INLINE_ENTRY.
11468 (gimple_build_debug_inline_entry): Declare.
11469 (gimple_debug_inline_entry_p): New.
11470 (gimple_debug_nonbind_marker_p): Adjust.
11471 * insn-notes.def (INLINE_ENTRY): New.
11472 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
11473 inline entry marker notes.
11474 (print_insn): Likewise.
11475 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
11476 (INSN_DEBUG_MARKER_KIND): Likewise.
11477 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
11478 * tree-inline.c (expand_call_inline): Build and insert
11479 debug_inline_entry stmt.
11480 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
11481 inline entry blocks early, if nonbind markers are enabled.
11482 (dump_scope_block): Dump fragment info.
11483 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
11484 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
11485 (gimple_build_debug_inline_entry): New.
11486 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
11487 Enable/disable inline entry points too.
11488 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
11489 (DEBUG_INSN): Describe inline entry markers.
11491 * common.opt (gvariable-location-views): New.
11492 (gvariable-location-views=incompat5): New.
11493 * config.in: Rebuilt.
11494 * configure: Rebuilt.
11495 * configure.ac: Test assembler for view support.
11496 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
11497 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
11498 * dwarf2out.c (var_loc_view): New typedef.
11499 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11500 (dwarf2out_locviews_in_attribute): New.
11501 (dwarf2out_locviews_in_loclist): New.
11502 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11503 (enum dw_line_info_opcode): Add LI_adv_address.
11504 (struct dw_line_info_table): Add view.
11505 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11506 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11507 (zero_view_p): New variable.
11508 (ZERO_VIEW_P): New macro.
11509 (output_asm_line_debug_info): New.
11510 (struct var_loc_node): Add view.
11511 (add_AT_view_list, AT_loc_list): New.
11512 (add_var_loc_to_decl): Add view param. Test it against last.
11513 (new_loc_list): Add view params. Record them.
11514 (AT_loc_list_ptr): Handle loc and view lists.
11515 (view_list_to_loc_list_val_node): New.
11516 (print_dw_val): Handle dw_val_class_view_list.
11517 (size_of_die): Likewise.
11518 (value_format): Likewise.
11519 (loc_list_has_views): New.
11520 (gen_llsym): Set vl_symbol too.
11521 (maybe_gen_llsym, skip_loc_list_entry): New.
11522 (dwarf2out_maybe_output_loclist_view_pair): New.
11523 (output_loc_list): Output view list or entries too.
11524 (output_view_list_offset): New.
11525 (output_die): Handle dw_val_class_view_list.
11526 (output_dwarf_version): New.
11527 (output_compilation_unit_header): Use it.
11528 (output_skeleton_debug_sections): Likewise.
11529 (output_rnglists, output_line_info): Likewise.
11530 (output_pubnames, output_aranges): Update version comments.
11531 (output_one_line_info_table): Output view numbers in asm comments.
11532 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11533 Call maybe_gen_llsym.
11534 (loc_list_from_tree_1): Adjust.
11535 (add_AT_location_description): Create view list attribute if
11536 needed, check it's absent otherwise.
11537 (convert_cfa_to_fb_loc_list): Adjust.
11538 (maybe_emit_file): Call output_asm_line_debug_info for test.
11539 (dwarf2out_var_location): Reset views as needed. Precompute
11540 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11541 attribute. Set view.
11542 (new_line_info_table): Reset next view.
11543 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11544 (dwarf2out_source_line): Likewise. Output view resets and labels to
11545 the assembler, or select appropriate line info opcodes.
11546 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11547 (optimize_string_length): Catch it. Adjust.
11548 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11549 dw_val_class_view_list, and remove it if no longer needed.
11550 (hash_loc_list): Hash view numbers.
11551 (loc_list_hasher::equal): Compare them.
11552 (optimize_location_lists): Check whether a view list symbol is
11553 needed, and whether the locview attribute is present, and
11554 whether they match. Remove the locview attribute if no longer
11556 (index_location_lists): Call skip_loc_list_entry for test.
11557 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11558 Use output_dwarf_version.
11559 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11560 (struct dw_val_node): Add val_view_list.
11561 * final.c (SEEN_NEXT_VIEW): New.
11562 (set_next_view_needed): New.
11563 (clear_next_view_needed): New.
11564 (maybe_output_next_view): New.
11565 (final_start_function): Rename to...
11566 (final_start_function_1): ... this. Take pointer to FIRST,
11567 add SEEN parameter. Emit param bindings in the initial view.
11568 (final_start_function): Reintroduce SEEN-less interface.
11569 (final): Rename to...
11570 (final_1): ... this. Take SEEN parameter. Output final pending
11571 next view at the end.
11572 (final): Reintroduce seen-less interface.
11573 (final_scan_insn): Output pending next view before switching
11574 sections or ending a block. Mark the next view as needed when
11575 outputting variable locations. Notify debug backend of section
11576 changes, and of location view changes.
11577 (rest_of_handle_final): Adjust.
11578 * toplev.c (process_options): Autodetect value for debug variable
11579 location views option. Warn on incompat5 without -gdwarf-5.
11580 * doc/invoke.texi (gvariable-location-views): New.
11581 (gvariable-location-views=incompat5): New.
11582 (gno-variable-location-views): New.
11584 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11586 PR tree-optimization/84136
11587 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11588 that the result of find_edge is non-NULL.
11590 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11593 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11594 storing integer register in SImode. Fix cost of 256 and 512
11595 byte aligned SSE register store.
11597 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11599 * config/i386/i386.c (ix86_multiplication_cost): Fix
11600 multiplication cost for TARGET_AVX512DQ.
11602 2018-02-08 Marek Polacek <polacek@redhat.com>
11604 PR tree-optimization/84238
11605 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11608 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11610 PR tree-optimization/84265
11611 * tree-vect-stmts.c (vectorizable_store): Don't treat
11612 VMAT_CONTIGUOUS accesses as grouped.
11613 (vectorizable_load): Likewise.
11615 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11617 PR tree-optimization/81635
11618 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11619 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11620 (test_round_for_mask): New functions.
11621 (wide_int_cc_tests): Call test_round_for_mask.
11622 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11623 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11624 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11625 range returned by get_range_info.
11627 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11630 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11631 * symtab.c: Include builtins.h
11632 (symtab_node::output_to_lto_symbol_table_p): Move here
11633 from lto-streamer-out.c:output_symbol_p.
11634 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11635 (output_symbol_p): Move all logic to symtab.c
11636 (produce_symtab): Update.
11638 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11640 * config/s390/s390-opts.h (enum indirect_branch): Define.
11641 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11642 (s390_indirect_branch_via_thunk)
11643 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11644 (enum s390_indirect_branch_type): Define.
11645 * config/s390/s390.c (struct s390_frame_layout, struct
11646 machine_function): Remove.
11647 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11648 (indirect_branch_table_label_no, indirect_branch_table_name):
11650 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11651 (enum s390_indirect_branch_option): Define.
11652 (s390_return_addr_from_memory): New function.
11653 (s390_handle_string_attribute): New function.
11654 (s390_attribute_table): Add new attribute handler.
11655 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11656 (s390_indirect_branch_via_thunk): New function.
11657 (s390_indirect_branch_via_inline_thunk): New function.
11658 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11659 sibling call optimization for non z10 compiles.
11660 (s390_emit_call): Force indirect branch target to be a single
11661 register. Add r1 clobber for non-z10 compiles.
11662 (s390_emit_epilogue): Emit return jump via return_use expander.
11663 (s390_reorg): Handle JUMP_INSNs as execute targets.
11664 (s390_option_override_internal): Perform validity checks for the
11665 new command line options.
11666 (s390_indirect_branch_attrvalue): New function.
11667 (s390_indirect_branch_settings): New function.
11668 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11669 (s390_output_indirect_thunk_function): New function.
11670 (s390_code_end): Implement target hook.
11671 (s390_case_values_threshold): Implement target hook.
11672 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11674 * config/s390/s390.h (struct s390_frame_layout)
11675 (struct machine_function): Move here from s390.c.
11676 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11677 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11678 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11679 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11680 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11681 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11682 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11683 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11684 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11685 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11686 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11687 (mnemonic attribute): Add values which aren't recognized
11689 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11690 pattern for branch conversion. Fix mnemonic attribute.
11691 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11692 indirect branch via thunk if requested.
11693 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11694 ("*indirect_jump"): Disable for branch conversion using out of
11696 ("indirect_jump_via_thunk<mode>_z10")
11697 ("indirect_jump_via_thunk<mode>")
11698 ("indirect_jump_via_inlinethunk<mode>_z10")
11699 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11700 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11701 ("casesi_jump_via_inlinethunk<mode>_z10")
11702 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11703 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11704 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11705 ("*indirect2_jump"): Disable for branch conversion.
11706 ("casesi_jump"): Turn into expander and expand patterns for branch
11708 ("return_use"): New expander.
11709 ("*return"): Emit return via thunk and rename it to ...
11710 ("*return<mode>"): ... this one.
11711 * config/s390/s390.opt: Add new options and and enum for the
11714 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11716 * lra-constraints.c (match_reload): Unconditionally use
11717 gen_lowpart_SUBREG, rather than selecting between that
11718 and equivalent gen_rtx_SUBREG code.
11720 2018-02-08 Richard Biener <rguenther@suse.de>
11722 PR tree-optimization/84233
11723 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11724 changed flag instead of boguously re-using phi_inserted.
11726 2018-02-08 Martin Jambor <mjambor@suse.cz>
11728 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11729 static local variables.
11731 2018-02-08 Richard Biener <rguenther@suse.de>
11733 PR tree-optimization/84278
11734 * tree-vect-stmts.c (vectorizable_store): When looking for
11735 smaller vector types to perform grouped strided loads/stores
11736 make sure the mode is supported by the target.
11737 (vectorizable_load): Likewise.
11739 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11741 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11742 Increase LDP/STP opportunities by adding adjacent callee-saves.
11744 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11746 PR rtl-optimization/84068
11747 PR rtl-optimization/83459
11748 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11750 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11752 PR tree-optimization/84224
11753 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11754 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11755 non-zero arguments.
11757 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11760 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11761 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11762 count, remove one USE.
11764 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11766 * doc/install.texi (Configuration): Document the
11767 --with-long-double-format={ibm,ieee} PowerPC configuration
11771 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11772 Convert from define_expand to be define_insn_and_split. Rework
11773 float/double/_Float128 conversions to QI/HI/SImode to work with
11774 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11775 conversions to QI/HImode types did a store and then a load to
11776 truncate the value. For conversions to VSX registers, don't split
11777 the insn, instead emit the code directly. Use the code iterator
11778 any_fix to combine signed and unsigned conversions.
11779 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11780 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11781 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11782 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11783 (fix_<mode>di2_hw): Likewise.
11784 (fixuns_<mode>di2_hw): Likewise.
11785 (fix_<mode>si2_hw): Likewise.
11786 (fixuns_<mode>si2_hw): Likewise.
11787 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11788 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11789 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11790 fix<uns>_trunc<SFDF:mode>si2_p8.
11791 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11792 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11793 (fix<uns>_<mode>_mem): Likewise.
11794 (fctiw<u>z_<mode>_mem): Likewise.
11795 (fix<uns>_<mode>_mem): Likewise.
11796 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11797 the register allocator from doing a direct move to the GPRs to do
11798 a store, and instead use the ISA 3.0 store byte/half-word from
11799 vector register instruction. For IEEE 128-bit floating point,
11800 also optimize stores of 32-bit ints.
11801 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11803 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11805 * genextract.c (push_pathstr_operand): New function to support
11807 (walk_rtx): Call push_pathstr_operand.
11808 (print_path): Support [a-zA-Z].
11810 2018-02-07 Richard Biener <rguenther@suse.de>
11812 PR tree-optimization/84037
11813 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11814 (cse_and_gimplify_to_preheader): Declare.
11815 (vect_get_place_in_interleaving_chain): Likewise.
11816 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11818 (_loop_vec_info::~_loop_vec_info): Delete it.
11819 (cse_and_gimplify_to_preheader): New function.
11820 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11821 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11822 (vectorizable_load): Likewise. For grouped stores always base
11823 the IV on the first element.
11824 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11825 condition before gimplifying.
11827 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11829 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11830 *DIV_EXPR and *MOD_EXPR.
11832 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11835 * config/i386/i386.c (ix86_option_override_internal): Mask out
11836 the CF_SET bit when checking -fcf-protection.
11838 2018-02-07 Tom de Vries <tom@codesourcery.com>
11841 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11844 2018-02-07 Richard Biener <rguenther@suse.de>
11846 PR tree-optimization/84204
11847 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11850 PR tree-optimization/84205
11851 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11852 special-case isl_ast_op_zdiv_r.
11854 PR tree-optimization/84223
11855 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11856 Only add conditions from within the region.
11857 (gather_bbs::after_dom_children): Adjust.
11859 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11862 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11863 * config/avr/avr.md: Only post-reload split REG-REG moves if
11864 either register is GENERAL_REG_P.
11866 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11868 PR tree-optimization/84235
11869 * tree-ssa-scopedtables.c
11870 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11871 if the subtraction is performed in floating point type where NaNs are
11872 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11873 build 1. Formatting fix.
11875 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11878 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11879 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11880 and skip it regardless of bb boundaries. Use CALL_P macro,
11881 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11882 unnecessarily, formatting fix.
11884 2018-02-06 Michael Collison <michael.collison@arm.com>
11886 * config/arm/thumb2.md:
11887 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11888 (*thumb_mov_notscc): Ditto.
11890 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11893 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11894 unsigned_fix, not "s".
11896 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11898 * configure.ac (gcc_fn_eh_frame_ro): New function.
11899 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11900 correct .eh_frame permissions.
11901 * configure: Regenerate.
11903 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11905 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11906 irrelevant options.
11908 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11910 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11911 Display warning message for -mno-speculate-indirect-jumps.
11913 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11915 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11917 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11919 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11921 PR tree-optimization/84225
11922 * tree-eh.c (find_trapping_overflow): Only call
11923 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11925 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11928 * config/i386/i386.c: Reimplement the check of possible options
11929 -mibt/-mshstk conbination. Change error messages.
11930 * doc/invoke.texi: Fix a typo: remove extra '='.
11932 2018-02-06 Marek Polacek <polacek@redhat.com>
11934 PR tree-optimization/84228
11935 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11937 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11940 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11941 emitted arch directives.
11942 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11943 __ARM_FEATURE_COPROC before changing architectures.
11945 2018-02-06 Richard Biener <rguenther@suse.de>
11947 * config/i386/i386.c (print_reg): Fix typo.
11948 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11950 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11952 * configure: Regenerate.
11954 2018-02-05 Martin Sebor <msebor@redhat.com>
11956 PR tree-optimization/83369
11957 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11960 2018-02-05 Martin Liska <mliska@suse.cz>
11962 * doc/invoke.texi: Cherry-pick upstream r323995.
11964 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11966 * ira.c (ira_init_register_move_cost): Adjust comment.
11968 2018-02-05 Martin Liska <mliska@suse.cz>
11970 PR gcov-profile/84137
11971 * doc/gcov.texi: Fix typo in documentation.
11973 2018-02-05 Martin Liska <mliska@suse.cz>
11975 PR gcov-profile/83879
11976 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11977 using dlopen functionality.
11979 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11981 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11982 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11983 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11984 _mm_maskz_range_round_ss): New intrinsics.
11985 (__builtin_ia32_rangesd128_round)
11986 (__builtin_ia32_rangess128_round): Remove.
11987 (__builtin_ia32_rangesd128_mask_round,
11988 __builtin_ia32_rangess128_mask_round): New builtins.
11989 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11990 __builtin_ia32_rangess128_round): Remove.
11991 (__builtin_ia32_rangesd128_mask_round,
11992 __builtin_ia32_rangess128_mask_round): New builtins.
11993 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11994 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11995 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11996 "<round_saeonly_constraint>")): Changed to ...
11997 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11998 "<round_saeonly_scalar_constraint>")): ... this.
11999 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
12000 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
12001 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
12002 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
12003 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
12005 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
12007 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
12009 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
12010 Remove all values except native, 8540 and 8548.
12012 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
12014 * config/i386/i386.c (ix86_output_function_return): Pass
12015 INVALID_REGNUM, instead of -1, as invalid register number to
12016 indirect_thunk_name and output_indirect_thunk.
12018 2018-02-02 Julia Koval <julia.koval@intel.com>
12020 * config.gcc: Add -march=icelake.
12021 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
12022 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
12023 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
12024 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
12025 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
12026 (processor_target_table): Add icelake.
12027 (ix86_option_override_internal): Handle new PTAs.
12028 (get_builtin_code_for_version): Handle icelake.
12029 (M_INTEL_COREI7_ICELAKE): New.
12030 (fold_builtin_cpu): Handle icelake.
12031 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
12032 * doc/invoke.texi: Add -march=icelake.
12034 2018-02-02 Julia Koval <julia.koval@intel.com>
12036 * config/i386/i386.c (ix86_option_override_internal): Change flags type
12037 to wide_int_bitmask.
12038 * wide-int-bitmask.h: New.
12040 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12043 * config/i386/i386.md: Replace Pmode with word_mode in
12044 builtin_setjmp_setup and builtin_longjmp to support x32.
12046 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
12050 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
12052 (rs6000_supported_cpu_names): New static variable.
12053 (linux_cpu_translation_table): Likewise.
12054 (elf_platform) <cpu>: Define new static variable and use it.
12055 Translate kernel AT_PLATFORM name to canonical name if needed.
12056 Error if platform name is unknown.
12058 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
12061 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
12063 2018-02-01 Jeff Law <law@redhat.com>
12066 * config/i386/i386.c (release_scratch_register_on_entry): Add new
12067 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
12068 the scratch if RELEASE_VIA_POP is false.
12069 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
12070 If we have to save a temporary register, decrement SIZE appropriately.
12071 Pass new arguments to release_scratch_register_on_entry.
12072 (ix86_adjust_stack_and_probe): Likewise.
12073 (ix86_emit_probe_stack_range): Pass new arguments to
12074 release_scratch_register_on_entry.
12076 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
12078 PR rtl-optimization/84157
12079 * combine.c (change_zero_ext): Use REG_P predicate in
12080 front of HARD_REGISTER_P predicate.
12082 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
12084 * config/avr/avr.c (avr_option_override): Move disabling of
12085 -fdelete-null-pointer-checks to...
12086 * common/config/avr/avr-common.c (avr_option_optimization_table):
12089 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12091 PR tree-optimization/81635
12092 * tree-data-ref.c (split_constant_offset_1): For types that
12093 wrap on overflow, try to use range info to prove that wrapping
12096 2018-02-01 Renlin Li <renlin.li@arm.com>
12099 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
12100 TAILCALL_ADDR_REGS.
12101 (aarch64_register_move_cost): Likewise.
12102 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
12103 TAILCALL_ADDR_REGS.
12104 (REG_CLASS_NAMES): Likewise.
12105 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
12106 TAILCALL_ADDR_REGS. Remove IP registers.
12107 * config/aarch64/aarch64.md (Ucs): Update register constraint.
12109 2018-02-01 Richard Biener <rguenther@suse.de>
12111 * domwalk.h (dom_walker::dom_walker): Add additional constructor
12112 for specifying RPO order and allow NULL for that.
12113 * domwalk.c (dom_walker::dom_walker): Likewise.
12114 (dom_walker::walk): Handle NULL RPO order.
12115 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
12117 (rewrite_update_dom_walker): Likewise.
12118 (mark_def_dom_walker): Likewise.
12120 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12122 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
12123 (aarch64_maybe_expand_sve_subreg_move): Declare.
12124 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
12125 * config/aarch64/predicates.md (aarch64_any_register_operand): New
12127 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
12128 that are semantically a reverse operation.
12129 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
12130 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
12131 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
12133 (aarch64_can_change_mode_class): For big-endian, forbid changes
12134 between two SVE modes if they have different element sizes.
12136 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12138 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
12139 the TImode handling for big-endian targets.
12141 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12143 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
12144 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
12146 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
12147 Remove BSWAP handing for big-endian targets and use the form of
12148 LD1RQ appropariate for the mode.
12150 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12152 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
12153 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
12154 duplicated element.
12156 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12159 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
12160 check for operands that need to go through aarch64_sve_reload_be.
12162 2018-02-01 Jakub Jelinek <jakub@redhat.com>
12164 PR tree-optimization/81661
12165 PR tree-optimization/84117
12166 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
12167 * tree-eh.c: Include gimplify.h.
12168 (find_trapping_overflow, replace_trapping_overflow,
12169 rewrite_to_non_trapping_overflow): New functions.
12170 * tree-vect-loop.c: Include tree-eh.h.
12171 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
12172 * tree-data-ref.c: Include tree-eh.h.
12173 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
12175 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
12177 PR rtl-optimization/84123
12178 * combine.c (change_zero_ext): Check if hard register satisfies
12179 can_change_dest_mode before calling gen_lowpart_SUBREG.
12181 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
12184 * ira.c (ira_init_register_move_cost): Remove assert.
12186 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12188 PR rtl-optimization/84071
12189 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
12190 * doc/tm.texi: Regenerate.
12192 2018-01-31 Richard Biener <rguenther@suse.de>
12194 PR tree-optimization/84132
12195 * tree-data-ref.c (analyze_miv_subscript): Properly
12196 check whether evolution_function_is_affine_multivariate_p
12197 before calling gcd_of_steps_may_divide_p.
12199 2018-01-31 Julia Koval <julia.koval@intel.com>
12202 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
12203 * config/i386/i386.md (rdpid_rex64) New.
12204 (rdpid): Make 32bit only.
12206 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
12209 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
12210 an IDENTIFIER_NODE for FUNCTION_TYPE's.
12212 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12215 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12217 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12219 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12221 PR rtl-optimization/84071
12222 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
12223 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
12225 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12227 * config/arc/arc.c (arc_handle_aux_attribute): New function.
12228 (arc_attribute_table): Add 'aux' attribute.
12229 (arc_in_small_data_p): Consider aux like variables.
12230 (arc_is_aux_reg_p): New function.
12231 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
12232 (arc_get_aux_arg): New function.
12233 (prepare_move_operands): Handle aux-register access.
12234 (arc_handle_aux_attribute): New function.
12235 * doc/extend.texi (ARC Variable attributes): Add subsection.
12237 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12239 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
12240 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
12241 (arc_attribute_table): Add 'uncached' attribute.
12242 (arc_print_operand): Print '.di' flag for uncached memory
12244 (arc_in_small_data_p): Do not consider for small data the uncached
12246 (arc_is_uncached_mem_p): New function.
12247 * config/arc/predicates.md (compact_store_memory_operand): Check
12248 for uncached memory accesses.
12249 (nonvol_nonimm_operand): Likewise.
12250 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
12252 2018-01-31 Jakub Jelinek <jakub@redhat.com>
12255 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
12256 falign-loops=): Add Optimization flag.
12258 2018-01-30 Jeff Law <law@redhat.com>
12261 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
12262 INT_REGISTERS_SAVED. Check it prior to calling
12263 get_scratch_register_on_entry.
12264 (ix86_adjust_stack_and_probe): Similarly.
12265 (ix86_emit_probe_stack_range): Similarly.
12266 (ix86_expand_prologue): Corresponding changes.
12268 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12271 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
12272 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
12274 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
12277 * lra-constraints.c (curr_insn_transform): Process AND in the
12280 2018-01-30 Jakub Jelinek <jakub@redhat.com>
12282 PR rtl-optimization/83986
12283 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
12284 dependence against last_pending_memory_flush in addition to
12285 pending_jump_insns.
12287 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
12289 PR tree-optimization/81611
12290 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
12293 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12296 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
12299 2018-01-30 Richard Biener <rguenther@suse.de>
12300 Jakub Jelinek <jakub@redhat.com>
12302 PR tree-optimization/84111
12303 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
12304 inner loops added during recursion, as they don't have up-to-date
12307 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12310 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
12311 (can_inline_edge_by_limits_p): ... here.
12312 (can_early_inline_edge_p, check_callers,
12313 update_caller_keys, update_callee_keys, recursive_inlining,
12314 add_new_edges_to_heap, speculation_useful_p,
12315 inline_small_functions,
12316 inline_small_functions, flatten_function,
12317 inline_to_all_callers_1): Update.
12319 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12321 * profile-count.c (profile_count::combine_with_ipa_count): Handle
12324 2018-01-30 Richard Biener <rguenther@suse.de>
12326 PR tree-optimization/83008
12327 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
12328 invariant and constant vector uses in stmts when they need
12329 more than one stmt.
12331 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12334 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
12335 * configure: Regenerate.
12337 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12339 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
12341 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
12342 Use gen_rtx_REG rather than gen_lowpart.
12344 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12346 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
12347 rather than 0 when creating partial subregs.
12349 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12351 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
12354 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
12357 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
12358 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
12359 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
12360 flags. This restores the settings used before the 2017-07-24.
12361 Turning off pre increment/decrement/modify allows IVOPTS to
12362 optimize DF/SF loops where the index is an int.
12364 2018-01-29 Richard Biener <rguenther@suse.de>
12365 Kelvin Nilsen <kelvin@gcc.gnu.org>
12368 * tree-vect-stmts.c (vectorizable_call): Don't call
12369 targetm.vectorize_builtin_md_vectorized_function if callee is
12372 2018-01-22 Carl Love <cel@us.ibm.com>
12374 * doc/extend.tex: Fix typo in second arg in
12375 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
12377 2018-01-29 Richard Biener <rguenther@suse.de>
12379 PR tree-optimization/84086
12380 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
12381 (flush_ssaname_freelist): When SSA names were released reset
12382 the SCEV hash table.
12384 2018-01-29 Richard Biener <rguenther@suse.de>
12386 PR tree-optimization/84057
12387 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
12388 removed paths when removing edges.
12390 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
12392 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
12393 -mfunction-return=@var{choice}.
12395 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
12397 PR diagnostic/84034
12398 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
12399 Handle CR like TAB.
12400 (layout::print_source_line): Likewise.
12401 (test_get_line_width_without_trailing_whitespace): Add test cases.
12403 2018-01-27 Jakub Jelinek <jakub@redhat.com>
12405 PR middle-end/84040
12406 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
12409 2018-01-26 Jim Wilson <jimw@sifive.com>
12411 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
12413 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
12416 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12418 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
12419 and CMP + SUB-immediate -> SUBS.
12421 2018-01-26 Martin Sebor <msebor@redhat.com>
12423 PR tree-optimization/83896
12424 * tree-ssa-strlen.c (get_string_len): Rename...
12425 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
12426 Avoid assuming length is constant.
12427 (handle_char_store): Use HOST_WIDE_INT for string length.
12429 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
12432 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
12433 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
12435 2018-01-26 Richard Biener <rguenther@suse.de>
12437 PR rtl-optimization/84003
12438 * dse.c (record_store): Only record redundant stores when
12439 the earlier store aliases at least all accesses the later one does.
12441 2018-01-26 Jakub Jelinek <jakub@redhat.com>
12443 PR rtl-optimization/83985
12444 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
12445 REG_CFA_RESTORE insns.
12446 (delete_unmarked_insns): Don't ignore separate shrink wrapping
12447 REG_CFA_RESTORE insns here.
12450 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
12451 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
12453 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12455 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
12456 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
12457 (arc_init): Likewise.
12458 (arc_override_options): Likewise.
12459 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
12461 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
12463 * config/arc/arc.h (TARGET_DBNZ): Define.
12464 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
12465 properly set the tune attribute.
12466 (dbnz): Use TARGET_DBNZ guard.
12467 * config/arc/arc.opt (mtune): Add core3 option.
12469 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12471 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
12472 recognize new pic like addresses.
12473 (arc_delegitimize_address): Clean up.
12475 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12477 * config/arc/arc-arches.def: Option mrf16 valid for all
12479 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
12480 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
12481 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
12482 * config/arc/arc-tables.opt: Regenerate.
12483 * config/arc/arc.c (arc_conditional_register_usage): Handle
12484 reduced register file case.
12485 (arc_file_start): Set must have build attributes.
12486 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
12487 mrf16 option value.
12488 * config/arc/arc.opt (mrf16): Add new option.
12489 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
12490 * config/arc/genmultilib.awk: Handle new mrf16 option.
12491 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
12492 * config/arc/t-multilib: Regenerate.
12493 * doc/invoke.texi (ARC Options): Document mrf16 option.
12495 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12497 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
12498 * config/arc/arc.c (arc_handle_secure_attribute): New function.
12499 (arc_attribute_table): Add 'secure_call' attribute.
12500 (arc_print_operand): Print secure call operand.
12501 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12503 (arc_is_secure_call_p): New function. * config/arc/arc.md
12504 (call_i): Add support for sjli instruction.
12505 (call_value_i): Likewise.
12506 * config/arc/constraints.md (Csc): New constraint.
12508 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12509 John Eric Martin <John.Martin@emmicro-us.com>
12511 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12512 * config/arc/arc.c (_arc_jli_section): New struct.
12513 (arc_jli_section): New type.
12514 (rc_jli_sections): New static variable.
12515 (arc_handle_jli_attribute): New function.
12516 (arc_attribute_table): Add jli_always and jli_fixed attribute.
12517 (arc_file_end): New function.
12518 (TARGET_ASM_FILE_END): Define.
12519 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
12520 (arc_add_jli_section): New function.
12521 (jli_call_scan): Likewise.
12522 (arc_reorg): Call jli_call_scan.
12523 (arc_output_addsi): Remove 'S' from printing asm operand.
12524 (arc_is_jli_call_p): New function.
12525 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
12527 (movhi_insn): Likewise.
12528 (movsi_insn): Likewise.
12529 (movsi_set_cc_insn): Likewise.
12530 (loadqi_update): Likewise.
12531 (load_zeroextendqisi_update): Likewise.
12532 (load_signextendqisi_update): Likewise.
12533 (loadhi_update): Likewise.
12534 (load_zeroextendhisi_update): Likewise.
12535 (load_signextendhisi_update): Likewise.
12536 (loadsi_update): Likewise.
12537 (loadsf_update): Likewise.
12538 (movsicc_insn): Likewise.
12539 (bset_insn): Likewise.
12540 (bxor_insn): Likewise.
12541 (bclr_insn): Likewise.
12542 (bmsk_insn): Likewise.
12543 (bicsi3_insn): Likewise.
12544 (cmpsi_cc_c_insn): Likewise.
12545 (movsi_ne): Likewise.
12546 (movsi_cond_exec): Likewise.
12547 (clrsbsi2): Likewise.
12548 (norm_f): Likewise.
12558 (call_i): Remove 'S' asm letter, add jli instruction.
12559 (call_value_i): Likewise.
12560 * config/arc/arc.op (mjli-always): New option.
12561 * config/arc/constraints.md (Cji): New constraint.
12562 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12564 (subsf3_fpx): Likewise.
12565 (mulsf3_fpx): Likewise.
12566 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12568 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12569 function attrbutes.
12570 * doc/invoke.texi (ARC): Document mjli-always option.
12572 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12574 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12575 avoid addition with 0 and use incw and decw where possible.
12577 2018-01-26 Richard Biener <rguenther@suse.de>
12579 PR tree-optimization/81082
12580 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12581 association if it requires casting to unsigned.
12582 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12583 from fold_plusminus_mult_expr to catch important cases late when
12584 range info is available.
12586 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12588 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12589 * configure.ac (hidden_linkonce): New test.
12590 * configure: Regenerate.
12591 * config.in: Regenerate.
12593 2018-01-26 Julia Koval <julia.koval@intel.com>
12595 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12596 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12597 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12598 _mm_mask_bitshuffle_epi64_mask): Fix type.
12599 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12600 USI_FTYPE_V4DI_V4DI_USI): Remove.
12601 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12602 __builtin_ia32_vpshufbitqmb256_mask,
12603 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12604 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12605 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12607 2018-01-26 Alan Modra <amodra@gmail.com>
12610 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12611 UNSPEC_VBPERMQ. Sort other unspecs.
12613 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12615 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12617 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12619 PR middle-end/83055
12620 * predict.c (drop_profile): Do not push/pop cfun; update also
12622 (handle_missing_profiles): Fix logic looking for zero profiles.
12624 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12626 PR middle-end/83977
12627 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12628 on functions with #pragma omp declare simd or functions with simd
12630 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12631 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12632 Remove trailing \n from warning_at calls.
12634 2018-01-25 Tom de Vries <tom@codesourcery.com>
12637 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12638 for neutered workers.
12640 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12643 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12644 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12646 2018-01-24 Jeff Law <law@redhat.com>
12649 * i386.c (get_probe_interval): Move to earlier point.
12650 (ix86_compute_frame_layout): If -fstack-clash-protection and
12651 the frame is larger than the probe interval, then use pushes
12652 to save registers rather than reg->mem moves.
12653 (ix86_expand_prologue): Remove conditional for int_registers_saved
12656 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12659 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12660 min/max for never referenced object.
12662 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12664 PR middle-end/83977
12665 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12667 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12668 attributes from DECL_ATTRIBUTES (decl) without affecting
12669 DECL_ATTRIBUTES (current_function_decl).
12670 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12671 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12673 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12675 PR tree-optimization/83979
12676 * fold-const.c (fold_comparison): Use constant_boolean_node
12677 instead of boolean_{true,false}_node.
12679 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12681 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12684 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12686 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12687 Simplify the clause that sets the length attribute.
12688 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12689 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12690 clause that sets the length attribute.
12691 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12693 2018-01-24 Tom de Vries <tom@codesourcery.com>
12696 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12697 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12698 Add strict parameter.
12699 (prevent_branch_around_nothing): Insert dummy insn between branch to
12700 label and label with no ptx insn inbetween.
12701 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12703 2018-01-24 Tom de Vries <tom@codesourcery.com>
12706 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12707 for neutered threads in warp.
12708 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12710 2018-01-24 Richard Biener <rguenther@suse.de>
12712 PR tree-optimization/83176
12713 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12716 2018-01-24 Richard Biener <rguenther@suse.de>
12718 PR tree-optimization/82819
12719 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12720 code generating pluses that are no-ops in the target precision.
12722 2018-01-24 Richard Biener <rguenther@suse.de>
12724 PR middle-end/84000
12725 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12727 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12729 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12730 to merge probabilities.
12731 * predict.c (probably_never_executed): Also mark as cold functions
12732 with global 0 profile and guessed local profile.
12733 * profile-count.c (profile_probability::combine_with_count): New
12735 * profile-count.h (profile_probability::operator*,
12736 profile_probability::operator*=, profile_probability::operator/,
12737 profile_probability::operator/=): Reduce precision to adjusted
12738 and set value to guessed on contradictory divisions.
12739 (profile_probability::combine_with_freq): Remove.
12740 (profile_probability::combine_wiht_count): Declare.
12741 (profile_count::force_nonzero):: Set to adjusted.
12742 (profile_count::probability_in):: Set quality to adjusted.
12743 * tree-ssa-tail-merge.c (replace_block_by): Use
12744 combine_with_count.
12746 2018-01-23 Andrew Waterman <andrew@sifive.com>
12747 Jim Wilson <jimw@sifive.com>
12749 * config/riscv/riscv.c (riscv_stack_boundary): New.
12750 (riscv_option_override): Set riscv_stack_boundary. Handle
12751 riscv_preferred_stack_boundary_arg.
12752 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12753 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12754 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12755 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12756 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12757 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12759 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12762 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12763 of struct ix86_frame.
12764 (ix86_expand_epilogue): Likewise. Add a local variable for
12765 the reg_save_offset field in struct ix86_frame.
12767 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12769 PR tree-optimization/82604
12770 * tree-loop-distribution.c (enum partition_kind): New enum item
12771 PKIND_PARTIAL_MEMSET.
12772 (partition_builtin_p): Support above new enum item.
12773 (generate_code_for_partition): Ditto.
12774 (compute_access_range): Differentiate cases that equality can be
12775 proven at all loops, the innermost loops or no loops.
12776 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12777 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12778 (finalize_partitions, distribute_loop): Don't fuse partition of
12779 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12780 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12781 parloop is enabled.
12783 2018-01-23 Martin Liska <mliska@suse.cz>
12785 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12786 order to ignore the predictor.
12787 (PRED_POLYMORPHIC_CALL): Likewise.
12788 (PRED_RECURSIVE_CALL): Likewise.
12790 2018-01-23 Martin Liska <mliska@suse.cz>
12792 * tree-profile.c (tree_profiling): Print function header to
12793 aware reader which function we are working on.
12794 * value-prof.c (gimple_find_values_to_profile): Do not print
12795 not interesting value histograms.
12797 2018-01-23 Martin Liska <mliska@suse.cz>
12799 * profile-count.h (enum profile_quality): Add
12800 profile_uninitialized as the first value. Do not number values
12801 as they are zero based.
12802 (profile_count::verify): Update sanity check.
12803 (profile_probability::verify): Likewise.
12805 2018-01-23 Nathan Sidwell <nathan@acm.org>
12807 * doc/invoke.texi (ffor-scope): Deprecate.
12809 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12811 PR tree-optimization/83510
12812 * domwalk.c (set_all_edges_as_executable): New function.
12813 (dom_walker::dom_walker): Convert bool param
12814 "skip_unreachable_blocks" to enum reachability. Move setup of
12815 edge flags to set_all_edges_as_executable and only do it when
12816 reachability is REACHABLE_BLOCKS.
12817 * domwalk.h (enum dom_walker::reachability): New enum.
12818 (dom_walker::dom_walker): Convert bool param
12819 "skip_unreachable_blocks" to enum reachability.
12820 (set_all_edges_as_executable): New decl.
12821 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12822 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12824 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12825 but converting true to REACHABLE_BLOCKS.
12826 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12828 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12829 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12830 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12832 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12833 if check_all_array_refs will be called.
12835 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12837 * tree.c (selftest::test_location_wrappers): Add more test
12840 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12842 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12843 (selftest::test_bit_in_range): Likewise.
12845 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12848 * doc/sourcebuild.texi (vect_float): Say that the selector
12849 only describes the situation when -funsafe-math-optimizations is on.
12850 (vect_float_strict): Document.
12852 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12854 PR tree-optimization/83965
12855 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12856 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12857 instead of checking only for a reduction.
12858 (vect_recog_widen_sum_pattern): Likewise.
12860 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12862 * predict.c (probably_never_executed): Only use precise profile info.
12863 (compute_function_frequency): Skip after inlining hack since we now
12864 have quality checking.
12866 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12868 * profile-count.h (profile_probability::very_unlikely,
12869 profile_probability::unlikely, profile_probability::even): Set
12870 precision to guessed.
12872 2018-01-23 Richard Biener <rguenther@suse.de>
12874 PR tree-optimization/83963
12875 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12876 Properly terminate dominator walk when crossing the exit edge not
12877 when visiting its source block.
12879 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12882 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12883 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12885 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12887 PR tree-optimization/83957
12888 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12889 semicolon after for body surrounded by braces.
12891 PR tree-optimization/83081
12892 * profile-count.h (profile_probability::split): New method.
12893 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12894 Use profile_probability::split.
12895 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12896 when splitting a single conditional jump into 2.
12898 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12900 PR tree-optimization/69452
12901 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12904 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12906 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12907 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12908 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12910 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12912 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12914 * config/rl78/rl78.md (movdi): New define_expand.
12915 * config/rl78/rl78.c (rl78_split_movdi): New function.
12917 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12920 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12922 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12923 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12924 128-bit to produce an UNSPEC move to get the double word with the
12925 signbit and then a shift directly to do signbit.
12926 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12927 implementation with a new version that just does either a direct
12928 move or a regular move. Move memory interface to separate insns.
12929 Move insns so they are next to the expander.
12930 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12931 with signbit move. Split big and little endian case.
12932 (signbit<mode>2_dm_mem_le): Likewise.
12933 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12934 (signbit<mode>2_dm2): Likewise.
12936 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12938 * config/rl78/rl78.md (anddi3): New define_expand.
12940 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12942 * config/rl78/rl78.md (umindi3): New define_expand.
12944 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12946 * config/rl78/rl78.md (smindi3): New define_expand.
12948 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12950 * config/rl78/rl78.md (smaxdi3): New define_expand.
12952 2018-01-22 Carl Love <cel@us.ibm.com>
12954 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12955 LVX_V1TI): Add macro expansion.
12956 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12957 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12958 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12959 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12960 Change check to determine if the instruction is a byte reversing
12961 entry. Fix typo in comment.
12962 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12963 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12964 Add def_builtin calls for new builtins.
12965 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12966 Add define_insn expansion.
12968 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12970 * config/rl78/rl78.md (umaxdi3): New define_expand.
12972 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12974 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12975 for non-QImode registers.
12977 2018-01-22 Richard Biener <rguenther@suse.de>
12979 PR tree-optimization/83963
12980 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12981 including the loop exit block.
12982 (scop_detection::merge_sese): Likewise.
12983 (scop_detection::add_scop): Do it here instead.
12985 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12987 * doc/sourcebuild.texi (arm_softfloat): Document.
12989 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12992 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12993 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12994 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12996 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12997 David Edelsohn <dje.gcc@gmail.com>
13000 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13001 Change "crset eq" to "crset 2".
13002 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13003 (*call_indirect_aix<mode>_nospec): Likewise.
13004 (*call_value_indirect_aix<mode>_nospec): Likewise.
13005 (*call_indirect_elfv2<mode>_nospec): Likewise.
13006 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
13007 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
13008 change assembly output from . to $.
13009 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13010 (indirect_jump<mode>_nospec): Change assembly output from . to $.
13011 (*tablejump<mode>_internal1_nospec): Likewise.
13013 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
13016 * config/sh/sh_optimize_sett_clrt.cc:
13017 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
13019 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
13021 PR tree-optimization/83940
13022 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
13023 offset_dt to vect_constant_def rather than vect_unknown_def_type.
13024 (vect_check_load_store_mask): Add a mask_dt_out parameter and
13025 use it to pass back the definition type.
13026 (vect_check_store_rhs): Likewise rhs_dt_out.
13027 (vect_build_gather_load_calls): Add a mask_dt argument and use
13028 it instead of a call to vect_is_simple_use.
13029 (vectorizable_store): Update calls to vect_check_load_store_mask
13030 and vect_check_store_rhs. Use the dt returned by the latter instead
13031 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
13032 instead of calls to vect_is_simple_use. Pass the scalar rather
13033 than the vector operand to vect_is_simple_use when handling
13034 second and subsequent copies of an rhs value.
13035 (vectorizable_load): Update calls to vect_check_load_store_mask
13036 and vect_build_gather_load_calls. Use the cached mask_dt and
13037 gs_info.offset_dt instead of calls to vect_is_simple_use.
13039 2018-01-20 Jakub Jelinek <jakub@redhat.com>
13041 PR middle-end/83945
13042 * tree-emutls.c: Include gimplify.h.
13043 (lower_emutls_2): New function.
13044 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
13045 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
13046 it before further processing.
13049 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
13050 UINTVAL (trueop1) instead of INTVAL (op1).
13052 2018-01-19 Jakub Jelinek <jakub@redhat.com>
13056 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
13057 INCOMING_FRAME_SP_OFFSET if not defined.
13058 (scan_trace): Add ENTRY argument. If true and
13059 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
13060 emit a note to adjust the CFA offset.
13061 (create_cfi_notes): Adjust scan_trace callers.
13062 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
13063 INCOMING_FRAME_SP_OFFSET in the CIE.
13064 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
13065 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
13067 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
13068 * doc/tm.texi: Regenerated.
13070 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13072 PR rtl-optimization/83147
13073 * lra-constraints.c (remove_inheritance_pseudos): Use
13074 lra_substitute_pseudo_within_insn.
13076 2018-01-19 Tom de Vries <tom@codesourcery.com>
13077 Cesar Philippidis <cesar@codesourcery.com>
13080 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
13082 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
13085 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
13086 spaces for function labels.
13088 2018-01-19 Martin Liska <mliska@suse.cz>
13090 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
13091 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
13092 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
13093 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
13094 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
13095 (PRED_CONST_RETURN): Change from 69 to 65.
13096 (PRED_NULL_RETURN): Change from 91 to 71.
13097 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
13098 (PRED_LOOP_GUARD): Change from 66 to 73.
13100 2018-01-19 Martin Liska <mliska@suse.cz>
13102 * predict.c (predict_insn_def): Add new assert.
13103 (struct branch_predictor): Change type to signed integer.
13104 (test_prediction_value_range): Amend test to cover
13105 PROB_UNINITIALIZED.
13106 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
13107 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
13108 (PRED_LOOP_ITERATIONS_MAX): Likewise.
13109 (PRED_LOOP_IV_COMPARE): Likewise.
13110 * predict.h (PROB_UNINITIALIZED): Define new constant.
13112 2018-01-19 Martin Liska <mliska@suse.cz>
13114 * predict.c (dump_prediction): Add new format for
13115 analyze_brprob.py script which is enabled with -details
13117 * profile-count.h (precise_p): New function.
13119 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13121 PR tree-optimization/83922
13122 * tree-vect-loop.c (vect_verify_full_masking): Return false if
13123 there are no statements that need masking.
13124 (vect_active_double_reduction_p): New function.
13125 (vect_analyze_loop_operations): Use it when handling phis that
13126 are not in the loop header.
13128 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13130 PR tree-optimization/83914
13131 * tree-vect-loop.c (vectorizable_induction): Don't convert
13132 init_expr or apply the peeling adjustment for inductions
13133 that are nested within the vectorized loop.
13135 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13137 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
13140 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13144 * function.h (gimplify_parameters): Add gimple_seq * argument.
13145 * function.c: Include gimple.h and options.h.
13146 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
13147 for the added local temporaries if needed.
13148 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
13149 if there are any parameter cleanups, wrap whole body into a
13150 try/finally with the cleanups.
13152 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
13155 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
13156 Use GET_MODE_CLASS for scalar floating point.
13158 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13162 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
13163 Fix call of call_cgraph_insertion_hooks.
13165 2018-01-18 Martin Sebor <msebor@redhat.com>
13167 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
13169 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13172 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
13175 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
13178 * common.opt: (-ffile-prefix-map): New option.
13179 * opts.c (common_handle_option): Defer it.
13180 * opts-global.c (handle_common_deferred_options): Handle it.
13181 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
13182 * file-prefix-map.h: New file.
13183 (remap_debug_filename, add_debug_prefix_map): ...here.
13184 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
13185 * final.c (debug_prefix_map, add_debug_prefix_map
13186 remap_debug_filename): Move to...
13187 * file-prefix-map.c: New file.
13188 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
13189 generalize, get rid of alloca(), use strrchr() instead of strchr().
13190 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
13191 Implement in terms of add_prefix_map().
13192 (remap_macro_filename, remap_debug_filename): Implement in term of
13194 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
13195 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
13196 * dbxout.c: Include file-prefix-map.h.
13197 * varasm.c: Likewise.
13198 * vmsdbgout.c: Likewise.
13199 * xcoffout.c: Likewise.
13200 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
13201 * doc/cppopts.texi (-fmacro-prefix-map): Document.
13202 * doc/invoke.texi (-ffile-prefix-map): Document.
13203 (-fdebug-prefix-map): Update description.
13205 2018-01-18 Martin Liska <mliska@suse.cz>
13207 * config/i386/i386.c (indirect_thunk_name): Document that also
13209 (output_indirect_thunk): Document why both instructions
13210 (pause and lfence) are generated.
13212 2018-01-18 Richard Biener <rguenther@suse.de>
13214 PR tree-optimization/83887
13215 * graphite-scop-detection.c
13216 (scop_detection::get_nearest_dom_with_single_entry): Remove.
13217 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
13218 (scop_detection::merge_sese): Re-implement with a flood-fill
13219 algorithm that properly finds a SESE region if it exists.
13221 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13224 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
13225 pointer_diff optimizations use view_convert instead of convert.
13227 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13229 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13230 Generate different code for -mno-speculate-indirect-jumps.
13231 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13232 (*call_indirect_aix<mode>): Disable for
13233 -mno-speculate-indirect-jumps.
13234 (*call_indirect_aix<mode>_nospec): New define_insn.
13235 (*call_value_indirect_aix<mode>): Disable for
13236 -mno-speculate-indirect-jumps.
13237 (*call_value_indirect_aix<mode>_nospec): New define_insn.
13238 (*sibcall_nonlocal_sysv<mode>): Generate different code for
13239 -mno-speculate-indirect-jumps.
13240 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13242 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
13244 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
13245 long double type, set the flags for noting the default long double
13246 type, even if we don't pass or return a long double type.
13248 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
13251 * ipa-inline.c (flatten_function): Do not overwrite final inlining
13254 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
13256 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
13257 support for merge[hl].
13258 (fold_mergehl_helper): New helper function.
13259 (tree-vector-builder.h): New #include for tree_vector_builder usage.
13260 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
13261 (altivec_vmrglw_direct): Add xxmrglw insn.
13263 2018-01-17 Andrew Waterman <andrew@sifive.com>
13265 * config/riscv/riscv.c (riscv_conditional_register_usage): If
13266 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
13268 2018-01-17 David Malcolm <dmalcolm@redhat.com>
13271 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
13272 call the lto_location_cache before reading the
13273 DECL_SOURCE_LOCATION of the types.
13275 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
13276 Richard Sandiford <richard.sandiford@linaro.org>
13278 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
13279 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
13280 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
13281 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
13282 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
13284 * config/aarch64/constraints.md (aarch64_movti_operand):
13286 * config/aarch64/predicates.md (Uti): Add new constraint.
13288 2018-01-17 Carl Love <cel@us.ibm.com>
13290 * config/rs6000/vsx.md (define_expand xl_len_r,
13291 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
13292 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
13294 (define_expand, define_insn): Move the shift left from the
13295 define_insn to the define_expand for lxvl and stxvl instructions.
13296 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
13297 and XL_LEN_R definitions to PURE.
13299 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
13301 * config/i386/i386.c (indirect_thunk_name): Declare regno
13302 as unsigned int. Compare regno with INVALID_REGNUM.
13303 (output_indirect_thunk): Ditto.
13304 (output_indirect_thunk_function): Ditto.
13305 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
13306 in the call to output_indirect_thunk_function.
13308 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
13310 PR middle-end/83884
13311 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
13312 rather than the size of inner_type to determine the stack slot size
13313 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
13315 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
13318 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
13321 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
13323 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
13324 endian Linux systems to optionally enable multilibs for selecting
13325 the long double type if the user configured an explicit type.
13326 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
13327 have no long double multilibs if not defined.
13328 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
13329 warn if the user used -mabi={ieee,ibm}longdouble and we built
13330 multilibs for long double.
13331 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
13332 appropriate multilib option.
13333 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
13335 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
13336 for building long double multilibs.
13337 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
13339 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
13341 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
13344 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
13346 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
13349 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
13352 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
13355 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13357 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
13358 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
13360 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
13362 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
13363 different rtl trees depending on TARGET_64BIT.
13364 (rs6000_gen_lvx): Likewise.
13366 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13368 * config/visium/visium.md (nop): Tweak comment.
13369 (hazard_nop): Likewise.
13371 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13373 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
13374 -mspeculate-indirect-jumps.
13375 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
13376 for -mno-speculate-indirect-jumps.
13377 (*call_indirect_elfv2<mode>_nospec): New define_insn.
13378 (*call_value_indirect_elfv2<mode>): Disable for
13379 -mno-speculate-indirect-jumps.
13380 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
13381 (indirect_jump): Emit different RTL for
13382 -mno-speculate-indirect-jumps.
13383 (*indirect_jump<mode>): Disable for
13384 -mno-speculate-indirect-jumps.
13385 (*indirect_jump<mode>_nospec): New define_insn.
13386 (tablejump): Emit different RTL for
13387 -mno-speculate-indirect-jumps.
13388 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
13389 (tablejumpsi_nospec): New define_expand.
13390 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
13391 (tablejumpdi_nospec): New define_expand.
13392 (*tablejump<mode>_internal1): Disable for
13393 -mno-speculate-indirect-jumps.
13394 (*tablejump<mode>_internal1_nospec): New define_insn.
13395 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
13398 2018-01-16 Artyom Skrobov tyomitch@gmail.com
13400 * caller-save.c (insert_save): Drop unnecessary parameter. All
13403 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13404 Richard Biener <rguenth@suse.de>
13407 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
13408 return early, inline manually is_gimple_sizepos. Make sure if we
13409 call gimplify_expr we don't end up with a gimple constant.
13410 * tree.c (variably_modified_type_p): Don't return true for
13411 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
13412 * gimplify.h (is_gimple_sizepos): Remove.
13414 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13416 PR tree-optimization/83857
13417 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
13418 vectorizable_live_operation for pure SLP statements.
13419 (vectorizable_live_operation): Handle PHIs.
13421 2018-01-16 Richard Biener <rguenther@suse.de>
13423 PR tree-optimization/83867
13424 * tree-vect-stmts.c (vect_transform_stmt): Precompute
13425 nested_in_vect_loop_p since the scalar stmt may get invalidated.
13427 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13430 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
13431 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
13432 If off is not INTEGER_CST, issue a may not be aligned warning
13433 rather than isn't aligned. Use isn%'t rather than isn't.
13434 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
13436 <case MULT_EXPR>: Improve the case when bottom and one of the
13437 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
13438 operand, in that case check if the other operand is multiple of
13439 bottom divided by the INTEGER_CST operand.
13441 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13444 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
13445 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
13446 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
13447 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
13448 * config/pa/pa.c (pa_function_arg_advance): Likewise.
13449 (pa_function_arg, pa_arg_partial_bytes): Likewise.
13450 (pa_function_arg_size): New function.
13452 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13454 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
13455 in a separate statement.
13457 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13459 PR tree-optimization/83847
13460 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
13461 group gathers and scatters.
13463 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13465 PR rtl-optimization/86620
13466 * params.def (max-sched-ready-insns): Bump minimum value to 1.
13468 PR rtl-optimization/83213
13469 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
13470 to last if both are JUMP_INSNs.
13472 PR tree-optimization/83843
13473 * gimple-ssa-store-merging.c
13474 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
13475 store_immediate_info for bswap/nop orig_stores.
13477 2018-01-15 Andrew Waterman <andrew@sifive.com>
13479 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
13481 <UDIV>: Increase cost if !TARGET_DIV.
13483 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
13485 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
13486 (define_attr "cr_logical_3op"): New.
13487 (cceq_ior_compare): Adjust.
13488 (cceq_ior_compare_complement): Adjust.
13489 (*cceq_rev_compare): Adjust.
13490 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
13491 (is_cracked_insn): Adjust.
13492 (insn_must_be_first_in_group): Adjust.
13493 * config/rs6000/40x.md: Adjust.
13494 * config/rs6000/440.md: Adjust.
13495 * config/rs6000/476.md: Adjust.
13496 * config/rs6000/601.md: Adjust.
13497 * config/rs6000/603.md: Adjust.
13498 * config/rs6000/6xx.md: Adjust.
13499 * config/rs6000/7450.md: Adjust.
13500 * config/rs6000/7xx.md: Adjust.
13501 * config/rs6000/8540.md: Adjust.
13502 * config/rs6000/cell.md: Adjust.
13503 * config/rs6000/e300c2c3.md: Adjust.
13504 * config/rs6000/e500mc.md: Adjust.
13505 * config/rs6000/e500mc64.md: Adjust.
13506 * config/rs6000/e5500.md: Adjust.
13507 * config/rs6000/e6500.md: Adjust.
13508 * config/rs6000/mpc.md: Adjust.
13509 * config/rs6000/power4.md: Adjust.
13510 * config/rs6000/power5.md: Adjust.
13511 * config/rs6000/power6.md: Adjust.
13512 * config/rs6000/power7.md: Adjust.
13513 * config/rs6000/power8.md: Adjust.
13514 * config/rs6000/power9.md: Adjust.
13515 * config/rs6000/rs64.md: Adjust.
13516 * config/rs6000/titan.md: Adjust.
13518 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13520 * config/i386/predicates.md (indirect_branch_operand): Rewrite
13521 ix86_indirect_branch_register logic.
13523 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13525 * config/i386/constraints.md (Bs): Update
13526 ix86_indirect_branch_register check. Don't check
13527 ix86_indirect_branch_register with GOT_memory_operand.
13529 * config/i386/predicates.md (GOT_memory_operand): Don't check
13530 ix86_indirect_branch_register here.
13531 (GOT32_symbol_operand): Likewise.
13533 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13535 * config/i386/predicates.md (constant_call_address_operand):
13536 Rewrite ix86_indirect_branch_register logic.
13537 (sibcall_insn_operand): Likewise.
13539 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13541 * config/i386/constraints.md (Bs): Replace
13542 ix86_indirect_branch_thunk_register with
13543 ix86_indirect_branch_register.
13545 * config/i386/i386.md (indirect_jump): Likewise.
13546 (tablejump): Likewise.
13547 (*sibcall_memory): Likewise.
13548 (*sibcall_value_memory): Likewise.
13549 Peepholes of indirect call and jump via memory: Likewise.
13550 * config/i386/i386.opt: Likewise.
13551 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13552 (GOT_memory_operand): Likewise.
13553 (call_insn_operand): Likewise.
13554 (sibcall_insn_operand): Likewise.
13555 (GOT32_symbol_operand): Likewise.
13557 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13559 PR middle-end/83837
13560 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13561 type rather than type addr's type points to.
13562 (expand_omp_atomic_mutex): Likewise.
13563 (expand_omp_atomic): Likewise.
13565 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13568 * config/i386/i386.c (output_indirect_thunk_function): Use
13569 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13570 for __x86_return_thunk.
13572 2018-01-15 Richard Biener <rguenther@suse.de>
13574 PR middle-end/83850
13575 * expmed.c (extract_bit_field_1): Fix typo.
13577 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13580 * config/arm/iterators.md (VF): New mode iterator.
13581 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13582 Remove integer-related logic from pattern.
13583 (neon_vabd<mode>_3): Likewise.
13585 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13587 PR middle-end/82694
13588 * common.opt (fstrict-overflow): No longer an alias.
13589 (fwrapv-pointer): New option.
13590 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13591 also for pointer types based on flag_wrapv_pointer.
13592 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13593 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13594 opts->x_flag_wrapv got set.
13595 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13596 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13597 POINTER_TYPE_OVERFLOW_UNDEFINED.
13598 * match.pd: Likewise in address comparison pattern.
13599 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13601 2018-01-15 Richard Biener <rguenther@suse.de>
13604 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13605 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13606 Reset type names to their identifier if their TYPE_DECL doesn't
13607 have linkage (and thus is used for ODR and devirt).
13608 (save_debug_info_for_decl): Remove.
13609 (save_debug_info_for_type): Likewise.
13610 (add_tree_to_fld_list): Adjust.
13611 * tree-pretty-print.c (dump_generic_node): Make dumping of
13612 type names more robust.
13614 2018-01-15 Richard Biener <rguenther@suse.de>
13616 * BASE-VER: Bump to 8.0.1.
13618 2018-01-14 Martin Sebor <msebor@redhat.com>
13621 * builtins.c (check_access): Avoid warning when the no-warning bit
13624 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13626 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13627 * ira-color (allocno_hard_regs_compare): Likewise.
13629 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13632 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13633 Use .pushsection/.popsection.
13635 2018-01-14 Martin Sebor <msebor@redhat.com>
13638 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13640 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13642 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13643 entry from extra_headers.
13644 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13645 extra_headers, make the list bitwise identical to the i?86-*-* one.
13647 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13649 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13650 -mcmodel=large with -mindirect-branch=thunk,
13651 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13652 -mfunction-return=thunk-extern.
13653 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13654 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13655 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13657 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13659 * config/i386/i386.c (print_reg): Print the name of the full
13660 integer register without '%'.
13661 (ix86_print_operand): Handle 'V'.
13662 * doc/extend.texi: Document 'V' modifier.
13664 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13666 * config/i386/constraints.md (Bs): Disallow memory operand for
13667 -mindirect-branch-register.
13669 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13670 (GOT_memory_operand): Likewise.
13671 (call_insn_operand): Likewise.
13672 (sibcall_insn_operand): Likewise.
13673 (GOT32_symbol_operand): Likewise.
13674 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13675 for -mindirect-branch-register.
13676 (tablejump): Likewise.
13677 (*sibcall_memory): Likewise.
13678 (*sibcall_value_memory): Likewise.
13679 Disallow peepholes of indirect call and jump via memory for
13680 -mindirect-branch-register.
13681 (*call_pop): Replace m with Bw.
13682 (*call_value_pop): Likewise.
13683 (*sibcall_pop_memory): Replace m with Bs.
13684 * config/i386/i386.opt (mindirect-branch-register): New option.
13685 * doc/invoke.texi: Document -mindirect-branch-register option.
13687 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13689 * config/i386/i386-protos.h (ix86_output_function_return): New.
13690 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13691 set function_return_type.
13692 (indirect_thunk_name): Add ret_p to indicate thunk for function
13694 (output_indirect_thunk_function): Pass false to
13695 indirect_thunk_name.
13696 (ix86_output_indirect_branch_via_reg): Likewise.
13697 (ix86_output_indirect_branch_via_push): Likewise.
13698 (output_indirect_thunk_function): Create alias for function
13699 return thunk if regno < 0.
13700 (ix86_output_function_return): New function.
13701 (ix86_handle_fndecl_attribute): Handle function_return.
13702 (ix86_attribute_table): Add function_return.
13703 * config/i386/i386.h (machine_function): Add
13704 function_return_type.
13705 * config/i386/i386.md (simple_return_internal): Use
13706 ix86_output_function_return.
13707 (simple_return_internal_long): Likewise.
13708 * config/i386/i386.opt (mfunction-return=): New option.
13709 (indirect_branch): Mention -mfunction-return=.
13710 * doc/extend.texi: Document function_return function attribute.
13711 * doc/invoke.texi: Document -mfunction-return= option.
13713 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13715 * config/i386/i386-opts.h (indirect_branch): New.
13716 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13717 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13718 with local indirect jump when converting indirect call and jump.
13719 (ix86_set_indirect_branch_type): New.
13720 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13721 (indirectlabelno): New.
13722 (indirect_thunk_needed): Likewise.
13723 (indirect_thunk_bnd_needed): Likewise.
13724 (indirect_thunks_used): Likewise.
13725 (indirect_thunks_bnd_used): Likewise.
13726 (INDIRECT_LABEL): Likewise.
13727 (indirect_thunk_name): Likewise.
13728 (output_indirect_thunk): Likewise.
13729 (output_indirect_thunk_function): Likewise.
13730 (ix86_output_indirect_branch_via_reg): Likewise.
13731 (ix86_output_indirect_branch_via_push): Likewise.
13732 (ix86_output_indirect_branch): Likewise.
13733 (ix86_output_indirect_jmp): Likewise.
13734 (ix86_code_end): Call output_indirect_thunk_function if needed.
13735 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13737 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13738 (ix86_attribute_table): Add indirect_branch.
13739 * config/i386/i386.h (machine_function): Add indirect_branch_type
13740 and has_local_indirect_jump.
13741 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13743 (tablejump): Likewise.
13744 (*indirect_jump): Use ix86_output_indirect_jmp.
13745 (*tablejump_1): Likewise.
13746 (simple_return_indirect_internal): Likewise.
13747 * config/i386/i386.opt (mindirect-branch=): New option.
13748 (indirect_branch): New.
13751 (thunk-inline): Likewise.
13752 (thunk-extern): Likewise.
13753 * doc/extend.texi: Document indirect_branch function attribute.
13754 * doc/invoke.texi: Document -mindirect-branch= option.
13756 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13759 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13761 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13763 * ipa-inline.c (want_inline_small_function_p): Return false if
13764 inlining has already failed with CIF_FINAL_ERROR.
13765 (update_caller_keys): Call want_inline_small_function_p before
13767 (update_callee_keys): Likewise.
13769 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13771 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13773 (rs6000_quadword_masked_address_p): Likewise.
13774 (quad_aligned_load_p): Likewise.
13775 (quad_aligned_store_p): Likewise.
13776 (const_load_sequence_p): Add comment to describe the outer-most loop.
13777 (mimic_memory_attributes_and_flags): New function.
13778 (rs6000_gen_stvx): Likewise.
13779 (replace_swapped_aligned_store): Likewise.
13780 (rs6000_gen_lvx): Likewise.
13781 (replace_swapped_aligned_load): Likewise.
13782 (replace_swapped_load_constant): Capitalize argument name in
13783 comment describing this function.
13784 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13785 and stores that access quad-word aligned addresses and replace
13786 with stvx or lvx instructions when appropriate.
13787 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13788 New function prototype.
13789 (rs6000_quadword_masked_address_p): Likewise.
13790 (rs6000_gen_lvx): Likewise.
13791 (rs6000_gen_stvx): Likewise.
13792 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13793 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13794 when memory address is aligned.
13795 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13796 this split to select lvx instruction when memory address is aligned.
13797 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13798 instruction when memory address is aligned.
13799 (*vsx_le_perm_load_v16qi): Likewise.
13800 (four unnamed splitters): Modify to select the stvx instruction
13801 when memory is aligned.
13803 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13805 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13806 which appears in the queue multiple times.
13808 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13809 Alan Hayward <alan.hayward@arm.com>
13810 David Sherwood <david.sherwood@arm.com>
13812 * tree-vectorizer.h (vec_lower_bound): New structure.
13813 (_loop_vec_info): Add check_nonzero and lower_bounds.
13814 (LOOP_VINFO_CHECK_NONZERO): New macro.
13815 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13816 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13817 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13818 fields. Make seg_len the distance travelled, not including the
13820 (dr_direction_indicator): Declare.
13821 (dr_zero_step_indicator): Likewise.
13822 (dr_known_forward_stride_p): Likewise.
13823 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13825 (runtime_alias_check_p): Allow runtime alias checks with
13827 (operator ==): Compare access_size and align.
13828 (prune_runtime_alias_test_list): Rework for new distinction between
13829 the access_size and seg_len.
13830 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13832 (get_segment_min_max): New function.
13833 (create_intersect_range_checks): Use it.
13834 (dr_step_indicator): New function.
13835 (dr_direction_indicator): Likewise.
13836 (dr_zero_step_indicator): Likewise.
13837 (dr_known_forward_stride_p): Likewise.
13838 * tree-loop-distribution.c (data_ref_segment_size): Return
13839 DR_STEP * (niters - 1).
13840 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13842 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13843 (vect_preserves_scalar_order_p): New function, split out from...
13844 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13845 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13846 (vect_vfa_access_size): New function.
13847 (vect_vfa_align): Likewise.
13848 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13849 (dump_lower_bound): New function.
13850 (vect_check_lower_bound): Likewise.
13851 (vect_small_gap_p): Likewise.
13852 (vectorizable_with_step_bound_p): Likewise.
13853 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13854 depencies if the vectorization factor is 1. Convert the checks
13855 for nonzero steps into checks on the bounds of DR_STEP. Try using
13856 a bunds check for variable steps if the minimum required step is
13857 relatively small. Update calls to the dr_with_seg_len
13858 constructor and to vect_compile_time_alias.
13859 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13861 (vect_loop_versioning): Call it.
13862 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13864 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13866 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13867 Alan Hayward <alan.hayward@arm.com>
13868 David Sherwood <david.sherwood@arm.com>
13870 * doc/sourcebuild.texi (vect_scatter_store): Document.
13871 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13873 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13875 * genopinit.c (main): Add supports_vec_scatter_store and
13876 supports_vec_scatter_store_cached to target_optabs.
13877 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13878 IFN_MASK_SCATTER_STORE.
13879 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13881 * internal-fn.h (internal_store_fn_p): Declare.
13882 (internal_fn_stored_value_index): Likewise.
13883 * internal-fn.c (scatter_store_direct): New macro.
13884 (expand_scatter_store_optab_fn): New function.
13885 (direct_scatter_store_optab_supported_p): New macro.
13886 (internal_store_fn_p): New function.
13887 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13888 IFN_MASK_SCATTER_STORE.
13889 (internal_fn_mask_index): Likewise.
13890 (internal_fn_stored_value_index): New function.
13891 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13892 for scatter stores.
13893 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13894 * optabs-query.c (supports_vec_scatter_store_p): New function.
13895 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13896 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13897 true for scatter stores.
13898 (vect_gather_scatter_fn_p): Handle scatter stores too.
13899 (vect_check_gather_scatter): Consider using scatter stores if
13900 supports_vec_scatter_store_p.
13901 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13902 scatter stores too.
13903 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13904 internal_fn_stored_value_index.
13905 (check_load_store_masking): Handle scatter stores too.
13906 (vect_get_store_rhs): Make public.
13907 (vectorizable_call): Use internal_store_fn_p.
13908 (vectorizable_store): Handle scatter store internal functions.
13909 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13910 when deciding whether the end of the group has been reached.
13911 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13912 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13913 (mask_scatter_store<mode>): New insns.
13915 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13916 Alan Hayward <alan.hayward@arm.com>
13917 David Sherwood <david.sherwood@arm.com>
13919 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13920 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13921 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13923 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13924 Use vect_truncate_gather_scatter_offset if we can't treat the
13925 operation as a normal gather load or scatter store.
13926 (get_group_load_store_type): Take the gather_scatter_info
13927 as argument. Try using a gather load or scatter store for
13928 single-element groups.
13929 (get_load_store_type): Update calls to get_group_load_store_type
13930 and vect_use_strided_gather_scatters_p.
13932 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13933 Alan Hayward <alan.hayward@arm.com>
13934 David Sherwood <david.sherwood@arm.com>
13936 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13937 optional tree argument.
13938 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13940 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13941 but continue to use the current value as a fallback.
13942 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13943 to compare the updates.
13944 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13945 (get_load_store_type): Use it when handling a strided access.
13946 (vect_get_strided_load_store_ops): New function.
13947 (vect_get_data_ptr_increment): Likewise.
13948 (vectorizable_load): Handle strided gather loads. Always pass
13949 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13951 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13952 Alan Hayward <alan.hayward@arm.com>
13953 David Sherwood <david.sherwood@arm.com>
13955 * doc/md.texi (gather_load@var{m}): Document.
13956 (mask_gather_load@var{m}): Likewise.
13957 * genopinit.c (main): Add supports_vec_gather_load and
13958 supports_vec_gather_load_cached to target_optabs.
13959 * optabs-tree.c (init_tree_optimization_optabs): Use
13960 ggc_cleared_alloc to allocate target_optabs.
13961 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13962 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13964 * internal-fn.h (internal_load_fn_p): Declare.
13965 (internal_gather_scatter_fn_p): Likewise.
13966 (internal_fn_mask_index): Likewise.
13967 (internal_gather_scatter_fn_supported_p): Likewise.
13968 * internal-fn.c (gather_load_direct): New macro.
13969 (expand_gather_load_optab_fn): New function.
13970 (direct_gather_load_optab_supported_p): New macro.
13971 (direct_internal_fn_optab): New function.
13972 (internal_load_fn_p): Likewise.
13973 (internal_gather_scatter_fn_p): Likewise.
13974 (internal_fn_mask_index): Likewise.
13975 (internal_gather_scatter_fn_supported_p): Likewise.
13976 * optabs-query.c (supports_at_least_one_mode_p): New function.
13977 (supports_vec_gather_load_p): Likewise.
13978 * optabs-query.h (supports_vec_gather_load_p): Declare.
13979 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13980 and memory_type field.
13981 (NUM_PATTERNS): Bump to 15.
13982 * tree-vect-data-refs.c: Include internal-fn.h.
13983 (vect_gather_scatter_fn_p): New function.
13984 (vect_describe_gather_scatter_call): Likewise.
13985 (vect_check_gather_scatter): Try using internal functions for
13986 gather loads. Recognize existing calls to a gather load function.
13987 (vect_analyze_data_refs): Consider using gather loads if
13988 supports_vec_gather_load_p.
13989 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13990 (vect_get_gather_scatter_offset_type): Likewise.
13991 (vect_convert_mask_for_vectype): Likewise.
13992 (vect_add_conversion_to_patterm): Likewise.
13993 (vect_try_gather_scatter_pattern): Likewise.
13994 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13995 (vect_vect_recog_func_ptrs): Add it.
13996 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13997 internal_fn_mask_index and internal_gather_scatter_fn_p.
13998 (check_load_store_masking): Take the gather_scatter_info as an
13999 argument and handle gather loads.
14000 (vect_get_gather_scatter_ops): New function.
14001 (vectorizable_call): Check internal_load_fn_p.
14002 (vectorizable_load): Likewise. Handle gather load internal
14004 (vectorizable_store): Update call to check_load_store_masking.
14005 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
14006 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
14007 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
14008 (aarch64_gather_scale_operand_d): New predicates.
14009 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
14010 (mask_gather_load<mode>): New insns.
14012 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14013 Alan Hayward <alan.hayward@arm.com>
14014 David Sherwood <david.sherwood@arm.com>
14016 * optabs.def (fold_left_plus_optab): New optab.
14017 * doc/md.texi (fold_left_plus_@var{m}): Document.
14018 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
14019 * internal-fn.c (fold_left_direct): Define.
14020 (expand_fold_left_optab_fn): Likewise.
14021 (direct_fold_left_optab_supported_p): Likewise.
14022 * fold-const-call.c (fold_const_fold_left): New function.
14023 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
14024 * tree-parloops.c (valid_reduction_p): New function.
14025 (gather_scalar_reductions): Use it.
14026 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
14027 (vect_finish_replace_stmt): Declare.
14028 * tree-vect-loop.c (fold_left_reduction_fn): New function.
14029 (needs_fold_left_reduction_p): New function, split out from...
14030 (vect_is_simple_reduction): ...here. Accept reductions that
14031 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
14032 (vect_force_simple_reduction): Also store the reduction type in
14033 the assignment's STMT_VINFO_REDUC_TYPE.
14034 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
14035 (merge_with_identity): New function.
14036 (vect_expand_fold_left): Likewise.
14037 (vectorize_fold_left_reduction): Likewise.
14038 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
14039 scalar phi in place for it. Check for target support and reject
14040 cases that would reassociate the operation. Defer the transform
14041 phase to vectorize_fold_left_reduction.
14042 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
14043 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
14044 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
14046 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14048 * tree-if-conv.c (predicate_mem_writes): Remove redundant
14049 call to ifc_temp_var.
14051 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14052 Alan Hayward <alan.hayward@arm.com>
14053 David Sherwood <david.sherwood@arm.com>
14055 * target.def (legitimize_address_displacement): Take the original
14056 offset as a poly_int.
14057 * targhooks.h (default_legitimize_address_displacement): Update
14059 * targhooks.c (default_legitimize_address_displacement): Likewise.
14060 * doc/tm.texi: Regenerate.
14061 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
14062 as an argument, moving assert of ad->disp == ad->disp_term to...
14063 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
14064 Try calling targetm.legitimize_address_displacement before expanding
14065 the address rather than afterwards, and adjust for the new interface.
14066 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
14067 Match the new hook interface. Handle SVE addresses.
14068 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
14069 new hook interface.
14071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14073 * Makefile.in (OBJS): Add early-remat.o.
14074 * target.def (select_early_remat_modes): New hook.
14075 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
14076 * doc/tm.texi: Regenerate.
14077 * targhooks.h (default_select_early_remat_modes): Declare.
14078 * targhooks.c (default_select_early_remat_modes): New function.
14079 * timevar.def (TV_EARLY_REMAT): New timevar.
14080 * passes.def (pass_early_remat): New pass.
14081 * tree-pass.h (make_pass_early_remat): Declare.
14082 * early-remat.c: New file.
14083 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
14085 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
14087 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14088 Alan Hayward <alan.hayward@arm.com>
14089 David Sherwood <david.sherwood@arm.com>
14091 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
14092 vfm1 with a bound_epilog parameter.
14093 (vect_do_peeling): Update calls accordingly, and move the prologue
14094 call earlier in the function. Treat the base bound_epilog as 0 for
14095 fully-masked loops and retain vf - 1 for other loops. Add 1 to
14096 this base when peeling for gaps.
14097 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
14098 with fully-masked loops.
14099 (vect_estimate_min_profitable_iters): Handle the single peeled
14100 iteration in that case.
14102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14103 Alan Hayward <alan.hayward@arm.com>
14104 David Sherwood <david.sherwood@arm.com>
14106 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
14107 single-element interleaving even if the size is not a power of 2.
14108 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
14109 accesses for single-element interleaving if the group size is
14112 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14113 Alan Hayward <alan.hayward@arm.com>
14114 David Sherwood <david.sherwood@arm.com>
14116 * doc/md.texi (fold_extract_last_@var{m}): Document.
14117 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
14118 * optabs.def (fold_extract_last_optab): New optab.
14119 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
14120 * internal-fn.c (fold_extract_direct): New macro.
14121 (expand_fold_extract_optab_fn): Likewise.
14122 (direct_fold_extract_optab_supported_p): Likewise.
14123 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
14124 * tree-vect-loop.c (vect_model_reduction_cost): Handle
14125 EXTRACT_LAST_REDUCTION.
14126 (get_initial_def_for_reduction): Do not create an initial vector
14127 for EXTRACT_LAST_REDUCTION reductions.
14128 (vectorizable_reduction): Leave the scalar phi in place for
14129 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
14130 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
14131 epilogue code for EXTRACT_LAST_REDUCTION and defer the
14132 transform phase to vectorizable_condition.
14133 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
14135 (vect_finish_stmt_generation): ...here.
14136 (vect_finish_replace_stmt): New function.
14137 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
14138 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
14140 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
14142 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14143 Alan Hayward <alan.hayward@arm.com>
14144 David Sherwood <david.sherwood@arm.com>
14146 * doc/md.texi (extract_last_@var{m}): Document.
14147 * optabs.def (extract_last_optab): New optab.
14148 * internal-fn.def (EXTRACT_LAST): New internal function.
14149 * internal-fn.c (cond_unary_direct): New macro.
14150 (expand_cond_unary_optab_fn): Likewise.
14151 (direct_cond_unary_optab_supported_p): Likewise.
14152 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
14153 loops using EXTRACT_LAST.
14154 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
14155 (extract_last_<mode>): ...this optab.
14156 (vec_extract<mode><Vel>): Update accordingly.
14158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14159 Alan Hayward <alan.hayward@arm.com>
14160 David Sherwood <david.sherwood@arm.com>
14162 * target.def (empty_mask_is_expensive): New hook.
14163 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
14164 * doc/tm.texi: Regenerate.
14165 * targhooks.h (default_empty_mask_is_expensive): Declare.
14166 * targhooks.c (default_empty_mask_is_expensive): New function.
14167 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
14168 if the target says that empty masks are expensive.
14169 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
14171 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
14173 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14174 Alan Hayward <alan.hayward@arm.com>
14175 David Sherwood <david.sherwood@arm.com>
14177 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
14178 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
14179 (vect_use_loop_mask_for_alignment_p): New function.
14180 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
14181 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
14182 niters_skip argument. Make sure that the first niters_skip elements
14183 of the first iteration are inactive.
14184 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
14185 Update call to vect_set_loop_masks_directly.
14186 (get_misalign_in_elems): New function, split out from...
14187 (vect_gen_prolog_loop_niters): ...here.
14188 (vect_update_init_of_dr): Take a code argument that specifies whether
14189 the adjustment should be added or subtracted.
14190 (vect_update_init_of_drs): Likewise.
14191 (vect_prepare_for_masked_peels): New function.
14192 (vect_do_peeling): Skip prologue peeling if we're using a mask
14193 instead. Update call to vect_update_inits_of_drs.
14194 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14196 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
14197 alignment. Do not include the number of peeled iterations in
14198 the minimum threshold in that case.
14199 (vectorizable_induction): Adjust the start value down by
14200 LOOP_VINFO_MASK_SKIP_NITERS iterations.
14201 (vect_transform_loop): Call vect_prepare_for_masked_peels.
14202 Take the number of skipped iterations into account when calculating
14204 * tree-vect-stmts.c (vect_gen_while_not): New function.
14206 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14207 Alan Hayward <alan.hayward@arm.com>
14208 David Sherwood <david.sherwood@arm.com>
14210 * doc/sourcebuild.texi (vect_fully_masked): Document.
14211 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
14212 default value to 0.
14213 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
14215 (vect_analyze_loop_2): ...here. Don't check the vectorization
14216 factor against the number of loop iterations if the loop is
14219 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14220 Alan Hayward <alan.hayward@arm.com>
14221 David Sherwood <david.sherwood@arm.com>
14223 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
14224 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
14225 (dump_groups): Update accordingly.
14226 (iv_use::mem_type): New member variable.
14227 (address_p): New function.
14228 (record_use): Add a mem_type argument and initialize the new
14230 (record_group_use): Add a mem_type argument. Use address_p.
14231 Remove obsolete null checks of base_object. Update call to record_use.
14232 (find_interesting_uses_op): Update call to record_group_use.
14233 (find_interesting_uses_cond): Likewise.
14234 (find_interesting_uses_address): Likewise.
14235 (get_mem_type_for_internal_fn): New function.
14236 (find_address_like_use): Likewise.
14237 (find_interesting_uses_stmt): Try find_address_like_use before
14238 calling find_interesting_uses_op.
14239 (addr_offset_valid_p): Use the iv mem_type field as the type
14240 of the addressed memory.
14241 (add_autoinc_candidates): Likewise.
14242 (get_address_cost): Likewise.
14243 (split_small_address_groups_p): Use address_p.
14244 (split_address_groups): Likewise.
14245 (add_iv_candidate_for_use): Likewise.
14246 (autoinc_possible_for_pair): Likewise.
14247 (rewrite_groups): Likewise.
14248 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
14249 (determine_group_iv_cost): Update after split of USE_ADDRESS.
14250 (get_alias_ptr_type_for_ptr_address): New function.
14251 (rewrite_use_address): Rewrite address uses in calls that were
14252 identified by find_address_like_use.
14254 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14255 Alan Hayward <alan.hayward@arm.com>
14256 David Sherwood <david.sherwood@arm.com>
14258 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
14260 * gimple-expr.h (is_gimple_addressable: Likewise.
14261 * gimple-expr.c (is_gimple_address): Likewise.
14262 * internal-fn.c (expand_call_mem_ref): New function.
14263 (expand_mask_load_optab_fn): Use it.
14264 (expand_mask_store_optab_fn): Likewise.
14266 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14267 Alan Hayward <alan.hayward@arm.com>
14268 David Sherwood <david.sherwood@arm.com>
14270 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
14271 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
14272 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
14273 (cond_umax@var{mode}): Document.
14274 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
14275 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
14276 (cond_umin_optab, cond_umax_optab): New optabs.
14277 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
14278 (COND_IOR, COND_XOR): New internal functions.
14279 * internal-fn.h (get_conditional_internal_fn): Declare.
14280 * internal-fn.c (cond_binary_direct): New macro.
14281 (expand_cond_binary_optab_fn): Likewise.
14282 (direct_cond_binary_optab_supported_p): Likewise.
14283 (get_conditional_internal_fn): New function.
14284 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
14285 Cope with reduction statements that are vectorized as calls rather
14287 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
14288 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
14289 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
14290 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14291 (UNSPEC_COND_EOR): New unspecs.
14292 (optab): Add mappings for them.
14293 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
14294 (sve_int_op, sve_fp_op): New int attributes.
14296 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14297 Alan Hayward <alan.hayward@arm.com>
14298 David Sherwood <david.sherwood@arm.com>
14300 * optabs.def (while_ult_optab): New optab.
14301 * doc/md.texi (while_ult@var{m}@var{n}): Document.
14302 * internal-fn.def (WHILE_ULT): New internal function.
14303 * internal-fn.h (direct_internal_fn_supported_p): New override
14304 that takes two types as argument.
14305 * internal-fn.c (while_direct): New macro.
14306 (expand_while_optab_fn): New function.
14307 (convert_optab_supported_p): Likewise.
14308 (direct_while_optab_supported_p): New macro.
14309 * wide-int.h (wi::udiv_ceil): New function.
14310 * tree-vectorizer.h (rgroup_masks): New structure.
14311 (vec_loop_masks): New typedef.
14312 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
14313 and fully_masked_p.
14314 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
14315 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
14316 (vect_max_vf): New function.
14317 (slpeel_make_loop_iterate_ntimes): Delete.
14318 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
14319 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
14320 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
14321 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
14322 internal-fn.h, stor-layout.h and optabs-query.h.
14323 (vect_set_loop_mask): New function.
14324 (add_preheader_seq): Likewise.
14325 (add_header_seq): Likewise.
14326 (interleave_supported_p): Likewise.
14327 (vect_maybe_permute_loop_masks): Likewise.
14328 (vect_set_loop_masks_directly): Likewise.
14329 (vect_set_loop_condition_masked): Likewise.
14330 (vect_set_loop_condition_unmasked): New function, split out from
14331 slpeel_make_loop_iterate_ntimes.
14332 (slpeel_make_loop_iterate_ntimes): Rename to..
14333 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
14334 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
14335 (vect_do_peeling): Update call accordingly.
14336 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
14338 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14339 mask_compare_type, can_fully_mask_p and fully_masked_p.
14340 (release_vec_loop_masks): New function.
14341 (_loop_vec_info): Use it to free the loop masks.
14342 (can_produce_all_loop_masks_p): New function.
14343 (vect_get_max_nscalars_per_iter): Likewise.
14344 (vect_verify_full_masking): Likewise.
14345 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
14346 retries, and free the mask rgroups before retrying. Check loop-wide
14347 reasons for disallowing fully-masked loops. Make the final decision
14348 about whether use a fully-masked loop or not.
14349 (vect_estimate_min_profitable_iters): Do not assume that peeling
14350 for the number of iterations will be needed for fully-masked loops.
14351 (vectorizable_reduction): Disable fully-masked loops.
14352 (vectorizable_live_operation): Likewise.
14353 (vect_halve_mask_nunits): New function.
14354 (vect_double_mask_nunits): Likewise.
14355 (vect_record_loop_mask): Likewise.
14356 (vect_get_loop_mask): Likewise.
14357 (vect_transform_loop): Handle the case in which the final loop
14358 iteration might handle a partial vector. Call vect_set_loop_condition
14359 instead of slpeel_make_loop_iterate_ntimes.
14360 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
14361 (check_load_store_masking): New function.
14362 (prepare_load_store_mask): Likewise.
14363 (vectorizable_store): Handle fully-masked loops.
14364 (vectorizable_load): Likewise.
14365 (supportable_widening_operation): Use vect_halve_mask_nunits for
14367 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
14368 (vect_gen_while): New function.
14369 * config/aarch64/aarch64.md (umax<mode>3): New expander.
14370 (aarch64_uqdec<mode>): New insn.
14372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14373 Alan Hayward <alan.hayward@arm.com>
14374 David Sherwood <david.sherwood@arm.com>
14376 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
14377 (reduc_xor_scal_optab): New optabs.
14378 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
14379 (reduc_xor_scal_@var{m}): Document.
14380 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
14381 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
14382 internal functions.
14383 * fold-const-call.c (fold_const_call): Handle them.
14384 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
14385 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
14386 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
14387 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
14388 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
14389 (UNSPEC_XORV): New unspecs.
14390 (optab): Add entries for them.
14391 (BITWISEV): New int iterator.
14392 (bit_reduc_op): New int attributes.
14394 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14395 Alan Hayward <alan.hayward@arm.com>
14396 David Sherwood <david.sherwood@arm.com>
14398 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
14399 * internal-fn.def (VEC_SHL_INSERT): New internal function.
14400 * optabs.def (vec_shl_insert_optab): New optab.
14401 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
14402 (duplicate_and_interleave): Likewise.
14403 * tree-vect-loop.c: Include internal-fn.h.
14404 (neutral_op_for_slp_reduction): New function, split out from
14405 get_initial_defs_for_reduction.
14406 (get_initial_def_for_reduction): Handle option 2 for variable-length
14407 vectors by loading the neutral value into a vector and then shifting
14408 the initial value into element 0.
14409 (get_initial_defs_for_reduction): Replace the code argument with
14410 the neutral value calculated by neutral_op_for_slp_reduction.
14411 Use gimple_build_vector for constant-length vectors.
14412 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
14413 but the first group_size elements have a neutral value.
14414 Use duplicate_and_interleave otherwise.
14415 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
14416 Update call to get_initial_defs_for_reduction. Handle SLP
14417 reductions for variable-length vectors by creating one vector
14418 result for each scalar result, with the elements associated
14419 with other scalar results stubbed out with the neutral value.
14420 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
14421 Require IFN_VEC_SHL_INSERT for double reductions on
14422 variable-length vectors, or SLP reductions that have
14423 a neutral value. Require can_duplicate_and_interleave_p
14424 support for variable-length unchained SLP reductions if there
14425 is no neutral value, such as for MIN/MAX reductions. Also require
14426 the number of vector elements to be a multiple of the number of
14427 SLP statements when doing variable-length unchained SLP reductions.
14428 Update call to vect_create_epilog_for_reduction.
14429 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
14430 and remove initial values.
14431 (duplicate_and_interleave): Make public.
14432 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
14433 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
14435 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14436 Alan Hayward <alan.hayward@arm.com>
14437 David Sherwood <david.sherwood@arm.com>
14439 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
14440 (can_duplicate_and_interleave_p): New function.
14441 (vect_get_and_check_slp_defs): Take the vector of statements
14442 rather than just the current one. Remove excess parentheses.
14443 Restriction rejectinon of vect_constant_def and vect_external_def
14444 for variable-length vectors to boolean types, or types for which
14445 can_duplicate_and_interleave_p is false.
14446 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
14447 (duplicate_and_interleave): New function.
14448 (vect_get_constant_vectors): Use gimple_build_vector for
14449 constant-length vectors and suitable variable-length constant
14450 vectors. Use duplicate_and_interleave for other variable-length
14451 vectors. Don't defer the update when inserting new statements.
14453 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14454 Alan Hayward <alan.hayward@arm.com>
14455 David Sherwood <david.sherwood@arm.com>
14457 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
14458 min_profitable_iters doesn't go negative.
14460 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14461 Alan Hayward <alan.hayward@arm.com>
14462 David Sherwood <david.sherwood@arm.com>
14464 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
14465 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
14466 * optabs.def (vec_mask_load_lanes_optab): New optab.
14467 (vec_mask_store_lanes_optab): Likewise.
14468 * internal-fn.def (MASK_LOAD_LANES): New internal function.
14469 (MASK_STORE_LANES): Likewise.
14470 * internal-fn.c (mask_load_lanes_direct): New macro.
14471 (mask_store_lanes_direct): Likewise.
14472 (expand_mask_load_optab_fn): Handle masked operations.
14473 (expand_mask_load_lanes_optab_fn): New macro.
14474 (expand_mask_store_optab_fn): Handle masked operations.
14475 (expand_mask_store_lanes_optab_fn): New macro.
14476 (direct_mask_load_lanes_optab_supported_p): Likewise.
14477 (direct_mask_store_lanes_optab_supported_p): Likewise.
14478 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
14480 (vect_load_lanes_supported): Likewise.
14481 * tree-vect-data-refs.c (strip_conversion): New function.
14482 (can_group_stmts_p): Likewise.
14483 (vect_analyze_data_ref_accesses): Use it instead of checking
14484 for a pair of assignments.
14485 (vect_store_lanes_supported): Take a masked_p parameter.
14486 (vect_load_lanes_supported): Likewise.
14487 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
14488 vect_store_lanes_supported and vect_load_lanes_supported.
14489 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14490 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
14491 parameter. Don't allow gaps for masked accesses.
14492 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
14493 and vect_load_lanes_supported.
14494 (get_load_store_type): Take a masked_p parameter and update
14495 call to get_group_load_store_type.
14496 (vectorizable_store): Update call to get_load_store_type.
14497 Handle IFN_MASK_STORE_LANES.
14498 (vectorizable_load): Update call to get_load_store_type.
14499 Handle IFN_MASK_LOAD_LANES.
14501 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14502 Alan Hayward <alan.hayward@arm.com>
14503 David Sherwood <david.sherwood@arm.com>
14505 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14507 * config/aarch64/aarch64-protos.h
14508 (aarch64_sve_struct_memory_operand_p): Declare.
14509 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14510 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14511 (VPRED, vpred): Handle SVE structure modes.
14512 * config/aarch64/constraints.md (Utx): New constraint.
14513 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14514 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14515 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
14516 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
14517 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
14518 structure modes. Split into pieces after RA.
14519 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
14520 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
14522 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
14523 SVE structure modes.
14524 (aarch64_classify_address): Likewise.
14525 (sizetochar): Move earlier in file.
14526 (aarch64_print_operand): Handle SVE register lists.
14527 (aarch64_array_mode): New function.
14528 (aarch64_sve_struct_memory_operand_p): Likewise.
14529 (TARGET_ARRAY_MODE): Redefine.
14531 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14532 Alan Hayward <alan.hayward@arm.com>
14533 David Sherwood <david.sherwood@arm.com>
14535 * target.def (array_mode): New target hook.
14536 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14537 * doc/tm.texi: Regenerate.
14538 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14539 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14540 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14541 targetm.array_mode.
14542 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14545 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14546 Alan Hayward <alan.hayward@arm.com>
14547 David Sherwood <david.sherwood@arm.com>
14549 * fold-const.c (fold_binary_loc): Check the argument types
14550 rather than the result type when testing for a vector operation.
14552 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14554 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14555 * doc/tm.texi: Regenerate.
14557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14558 Alan Hayward <alan.hayward@arm.com>
14559 David Sherwood <david.sherwood@arm.com>
14561 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14562 (sve): Document new AArch64 extension.
14563 * doc/md.texi (w): Extend the description of the AArch64
14564 constraint to include SVE vectors.
14565 (Upl, Upa): Document new AArch64 predicate constraints.
14566 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14568 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14569 (msve-vector-bits=): New option.
14570 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14571 SVE when these are disabled.
14572 (sve): New extension.
14573 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14574 modes. Adjust their number of units based on aarch64_sve_vg.
14575 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14576 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14577 aarch64_addr_query_type.
14578 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14579 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14580 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14581 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14582 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14583 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14584 (aarch64_simd_imm_zero_p): Delete.
14585 (aarch64_check_zero_based_sve_index_immediate): Declare.
14586 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14587 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14588 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14589 (aarch64_sve_float_mul_immediate_p): Likewise.
14590 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14591 rather than an rtx.
14592 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14593 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14594 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14595 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14596 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14597 (aarch64_regmode_natural_size): Likewise.
14598 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14599 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14601 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14602 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14603 for VG and the SVE predicate registers.
14604 (V_ALIASES): Add a "z"-prefixed alias.
14605 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14606 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14607 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14608 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14609 (REG_CLASS_NAMES): Add entries for them.
14610 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14611 and the predicate registers.
14612 (aarch64_sve_vg): Declare.
14613 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14614 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14615 (REGMODE_NATURAL_SIZE): Define.
14616 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14618 * config/aarch64/aarch64.c: Include cfgrtl.h.
14619 (simd_immediate_info): Add a constructor for series vectors,
14620 and an associated step field.
14621 (aarch64_sve_vg): New variable.
14622 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14623 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14624 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14625 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14626 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14627 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14628 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14629 (aarch64_get_mask_mode): New functions.
14630 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14631 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14632 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14633 predicate modes and predicate registers. Explicitly restrict
14634 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14635 to store a vector mode if it is recognized by
14636 aarch64_classify_vector_mode.
14637 (aarch64_regmode_natural_size): New function.
14638 (aarch64_hard_regno_caller_save_mode): Return the original mode
14640 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14641 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14642 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14643 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14645 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14646 does not overlap dest if the function is frame-related. Handle
14648 (aarch64_split_add_offset): New function.
14649 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14650 them aarch64_add_offset.
14651 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14652 and update call to aarch64_sub_sp.
14653 (aarch64_add_cfa_expression): New function.
14654 (aarch64_expand_prologue): Pass extra temporary registers to the
14655 functions above. Handle the case in which we need to emit new
14656 DW_CFA_expressions for registers that were originally saved
14657 relative to the stack pointer, but now have to be expressed
14658 relative to the frame pointer.
14659 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14661 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14662 IP0 and IP1 values for SVE frames.
14663 (aarch64_expand_vec_series): New function.
14664 (aarch64_expand_sve_widened_duplicate): Likewise.
14665 (aarch64_expand_sve_const_vector): Likewise.
14666 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14667 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14668 into the register, rather than emitting a SET directly.
14669 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14670 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14671 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14672 (offset_9bit_signed_scaled_p): New functions.
14673 (aarch64_replicate_bitmask_imm): New function.
14674 (aarch64_bitmask_imm): Use it.
14675 (aarch64_cannot_force_const_mem): Reject expressions involving
14676 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14677 (aarch64_classify_index): Handle SVE indices, by requiring
14678 a plain register index with a scale that matches the element size.
14679 (aarch64_classify_address): Handle SVE addresses. Assert that
14680 the mode of the address is VOIDmode or an integer mode.
14681 Update call to aarch64_classify_symbol.
14682 (aarch64_classify_symbolic_expression): Update call to
14683 aarch64_classify_symbol.
14684 (aarch64_const_vec_all_in_range_p): New function.
14685 (aarch64_print_vector_float_operand): Likewise.
14686 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14687 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14688 and the FP immediates 1.0 and 0.5.
14689 (aarch64_print_address_internal): Handle SVE addresses.
14690 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14691 (aarch64_regno_regclass): Handle predicate registers.
14692 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14694 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14695 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14696 (aarch64_convert_sve_vector_bits): New function.
14697 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14698 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14699 rather than an rtx.
14700 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14701 Handle SVE vector and predicate modes. Accept VL-based constants
14702 that need only one temporary register, and VL offsets that require
14703 no temporary registers.
14704 (aarch64_conditional_register_usage): Mark the predicate registers
14705 as fixed if SVE isn't available.
14706 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14707 Return true for SVE vector and predicate modes.
14708 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14709 rather than an unsigned int. Handle SVE modes.
14710 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14712 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14714 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14715 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14716 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14717 (aarch64_sve_float_mul_immediate_p): New functions.
14718 (aarch64_sve_valid_immediate): New function.
14719 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14720 Explicitly reject structure modes. Check for INDEX constants.
14721 Handle PTRUE and PFALSE constants.
14722 (aarch64_check_zero_based_sve_index_immediate): New function.
14723 (aarch64_simd_imm_zero_p): Delete.
14724 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14725 vector modes. Accept constants in the range of CNT[BHWD].
14726 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14727 ask for an Advanced SIMD mode.
14728 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14729 (aarch64_simd_vector_alignment): Handle SVE predicates.
14730 (aarch64_vectorize_preferred_vector_alignment): New function.
14731 (aarch64_simd_vector_alignment_reachable): Use it instead of
14733 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14734 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14736 (MAX_VECT_LEN): Delete.
14737 (expand_vec_perm_d): Add a vec_flags field.
14738 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14739 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14740 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14742 (aarch64_evpc_rev): Rename to...
14743 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14744 (aarch64_evpc_rev_global): New function.
14745 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14746 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14748 (aarch64_evpc_sve_tbl): New function.
14749 (aarch64_expand_vec_perm_const_1): Update after rename of
14750 aarch64_evpc_rev. Handle SVE permutes too, trying
14751 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14752 than aarch64_evpc_tbl.
14753 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14754 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14755 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14756 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14757 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14758 (aarch64_expand_sve_vcond): New functions.
14759 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14760 of aarch64_vector_mode_p.
14761 (aarch64_dwarf_poly_indeterminate_value): New function.
14762 (aarch64_compute_pressure_classes): Likewise.
14763 (aarch64_can_change_mode_class): Likewise.
14764 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14765 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14766 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14767 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14768 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14769 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14770 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14771 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14773 (Dn, Dl, Dr): Accept const as well as const_vector.
14774 (Dz): Likewise. Compare against CONST0_RTX.
14775 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14776 of "vector" where appropriate.
14777 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14778 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14779 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14780 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14781 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14782 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14783 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14784 (v_int_equiv): Extend to SVE modes.
14785 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14787 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14788 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14789 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14790 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14791 (SVE_COND_FP_CMP): New int iterators.
14792 (perm_hilo): Handle the new unpack unspecs.
14793 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14795 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14796 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14797 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14798 (aarch64_equality_operator, aarch64_constant_vector_operand)
14799 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14800 (aarch64_sve_nonimmediate_operand): Likewise.
14801 (aarch64_sve_general_operand): Likewise.
14802 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14803 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14804 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14805 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14806 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14807 (aarch64_sve_float_arith_immediate): Likewise.
14808 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14809 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14810 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14811 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14812 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14813 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14814 (aarch64_sve_float_arith_operand): Likewise.
14815 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14816 (aarch64_sve_float_mul_operand): Likewise.
14817 (aarch64_sve_vec_perm_operand): Likewise.
14818 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14819 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14820 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14821 as well as const_vector.
14822 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14823 in file. Use CONST0_RTX and CONSTM1_RTX.
14824 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14825 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14826 Use aarch64_simd_imm_zero.
14827 * config/aarch64/aarch64-sve.md: New file.
14828 * config/aarch64/aarch64.md: Include it.
14829 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14830 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14831 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14832 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14833 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14834 (sve): New attribute.
14835 (enabled): Disable instructions with the sve attribute unless
14837 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14838 aarch64_expand_mov_immediate.
14839 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14840 CNT[BHSD] immediates.
14841 (movti): Split CONST_POLY_INT moves into two halves.
14842 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14843 Split additions that need a temporary here if the destination
14844 is the stack pointer.
14845 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14846 (*add<mode>3_poly_1): New instruction.
14847 (set_clobber_cc): New expander.
14849 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14851 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14852 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14853 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14854 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14855 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14856 Change innermode from fixed_mode_size to machine_mode.
14857 (simplify_subreg): Update call accordingly. Handle a constant-sized
14858 subreg of a variable-length CONST_VECTOR.
14860 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14861 Alan Hayward <alan.hayward@arm.com>
14862 David Sherwood <david.sherwood@arm.com>
14864 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14865 (add_offset_to_base): New function, split out from...
14866 (create_mem_ref): ...here. When handling a scale other than 1,
14867 check first whether the address is valid without the offset.
14868 Add it into the base if so, leaving the index and scale as-is.
14870 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14873 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14874 fold_for_warn before checking if arg2 is INTEGER_CST.
14876 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14878 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14879 (store_multiple_operation): Delete.
14880 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14881 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14882 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14883 guarded by TARGET_STRING.
14884 (rs6000_output_load_multiple): Delete.
14885 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14886 OPTION_MASK_STRING / TARGET_STRING handling.
14887 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14888 (const rs6000_opt_masks) <"string">: Change mask to 0.
14889 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14890 (MASK_STRING): Delete.
14891 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14893 (load_multiple): Delete.
14900 (store_multiple): Delete.
14907 (movmemsi_8reg): Delete.
14908 (corresponding unnamed define_insn): Delete.
14909 (movmemsi_6reg): Delete.
14910 (corresponding unnamed define_insn): Delete.
14911 (movmemsi_4reg): Delete.
14912 (corresponding unnamed define_insn): Delete.
14913 (movmemsi_2reg): Delete.
14914 (corresponding unnamed define_insn): Delete.
14915 (movmemsi_1reg): Delete.
14916 (corresponding unnamed define_insn): Delete.
14917 * config/rs6000/rs6000.opt (mno-string): New.
14918 (mstring): Replace by deprecation warning stub.
14919 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14921 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14923 * regrename.c (regrename_do_replace): If replacing the same
14924 reg multiple times, try to reuse last created gen_raw_REG.
14927 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14928 main to workaround a bug in GDB.
14930 2018-01-12 Tom de Vries <tom@codesourcery.com>
14933 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14935 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14937 PR rtl-optimization/80481
14938 * ira-color.c (get_cap_member): New function.
14939 (allocnos_conflict_by_live_ranges_p): Use it.
14940 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14941 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14943 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14946 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14947 (*saddl_se_1): Ditto.
14948 (*ssubsi_1): Ditto.
14949 (*ssubl_se_1): Ditto.
14951 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14953 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14954 rather than wi::to_widest for DR_INITs.
14955 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14956 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14957 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14959 (vect_analyze_group_access_1): Note that here.
14961 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14963 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14964 polynomial type sizes.
14966 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14968 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14969 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14970 (gimple_add_tmp_var): Likewise.
14972 2018-01-12 Martin Liska <mliska@suse.cz>
14974 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14975 (gimple_alloc_sizes): Likewise.
14976 (dump_gimple_statistics): Use PRIu64 in printf format.
14977 * gimple.h: Change uint64_t to int.
14979 2018-01-12 Martin Liska <mliska@suse.cz>
14981 * tree-core.h: Use uint64_t instead of int.
14982 * tree.c (tree_node_counts): Likewise.
14983 (tree_node_sizes): Likewise.
14984 (dump_tree_statistics): Use PRIu64 in printf format.
14986 2018-01-12 Martin Liska <mliska@suse.cz>
14988 * Makefile.in: As qsort_chk is implemented in vec.c, add
14989 vec.o to linkage of gencfn-macros.
14990 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14991 passing the info to record_node_allocation_statistics.
14992 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14994 * ggc-common.c (struct ggc_usage): Add operator== and use
14995 it in operator< and compare function.
14996 * mem-stats.h (struct mem_usage): Likewise.
14997 * vec.c (struct vec_usage): Remove operator< and compare
14998 function. Can be simply inherited.
15000 2018-01-12 Martin Jambor <mjambor@suse.cz>
15003 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
15004 * tree-ssa-math-opts.c: Include domwalk.h.
15005 (convert_mult_to_fma_1): New function.
15006 (fma_transformation_info): New type.
15007 (fma_deferring_state): Likewise.
15008 (cancel_fma_deferring): New function.
15009 (result_of_phi): Likewise.
15010 (last_fma_candidate_feeds_initial_phi): Likewise.
15011 (convert_mult_to_fma): Added deferring logic, split actual
15012 transformation to convert_mult_to_fma_1.
15013 (math_opts_dom_walker): New type.
15014 (math_opts_dom_walker::after_dom_children): New method, body moved
15015 here from pass_optimize_widening_mul::execute, added deferring logic
15017 (pass_optimize_widening_mul::execute): Moved most of code to
15018 math_opts_dom_walker::after_dom_children.
15019 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
15020 * config/i386/i386.c (ix86_option_override_internal): Added
15021 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
15023 2018-01-12 Richard Biener <rguenther@suse.de>
15026 * dwarf2out.c (gen_variable_die): Do not reset old_die for
15027 inline instance vars.
15029 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
15032 * config/rx/rx.c (rx_is_restricted_memory_address):
15033 Handle SUBREG case.
15035 2018-01-12 Richard Biener <rguenther@suse.de>
15037 PR tree-optimization/80846
15038 * target.def (split_reduction): New target hook.
15039 * targhooks.c (default_split_reduction): New function.
15040 * targhooks.h (default_split_reduction): Declare.
15041 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
15042 target requests first reduce vectors by combining low and high
15044 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
15045 (get_vectype_for_scalar_type_and_size): Export.
15046 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
15047 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
15048 * doc/tm.texi: Regenerate.
15049 * config/i386/i386.c (ix86_split_reduction): Implement
15050 TARGET_VECTORIZE_SPLIT_REDUCTION.
15052 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15055 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
15056 in PIC mode except for TARGET_VXWORKS_RTP.
15057 * config/sparc/sparc.c: Include cfgrtl.h.
15058 (TARGET_INIT_PIC_REG): Define.
15059 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
15060 (sparc_pic_register_p): New predicate.
15061 (sparc_legitimate_address_p): Use it.
15062 (sparc_legitimize_pic_address): Likewise.
15063 (sparc_delegitimize_address): Likewise.
15064 (sparc_mode_dependent_address_p): Likewise.
15065 (gen_load_pcrel_sym): Remove 4th parameter.
15066 (load_got_register): Adjust call to above. Remove obsolete stuff.
15067 (sparc_expand_prologue): Do not call load_got_register here.
15068 (sparc_flat_expand_prologue): Likewise.
15069 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
15070 (sparc_use_pseudo_pic_reg): New function.
15071 (sparc_init_pic_reg): Likewise.
15072 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15073 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
15075 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
15077 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
15078 Add item for branch_cost.
15080 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15082 PR rtl-optimization/83565
15083 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
15084 not extend the result to a larger mode for rotate operations.
15085 (num_sign_bit_copies1): Likewise.
15087 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15090 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
15092 Use values-Xc.o for -pedantic.
15093 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
15095 2018-01-12 Martin Liska <mliska@suse.cz>
15098 * ipa-devirt.c (final_warning_record::grow_type_warnings):
15100 (possible_polymorphic_call_targets): Use it.
15101 (ipa_devirt): Likewise.
15103 2018-01-12 Martin Liska <mliska@suse.cz>
15105 * profile-count.h (enum profile_quality): Use 0 as invalid
15106 enum value of profile_quality.
15108 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
15110 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
15111 -mext-string options.
15113 2018-01-12 Richard Biener <rguenther@suse.de>
15115 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
15116 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
15117 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
15119 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
15121 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15123 * configure.ac (--with-long-double-format): Add support for the
15124 configuration option to change the default long double format on
15126 * config.gcc (powerpc*-linux*-*): Likewise.
15127 * configure: Regenerate.
15128 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
15129 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
15130 used without modification.
15132 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15134 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
15135 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
15136 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
15137 MISC_BUILTIN_SPEC_BARRIER.
15138 (rs6000_init_builtins): Likewise.
15139 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
15141 (speculation_barrier): New define_insn.
15142 * doc/extend.texi: Document __builtin_speculation_barrier.
15144 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15147 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
15148 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
15149 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
15151 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
15152 integral modes instead of "ss" and "sd".
15153 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
15154 vectors with 32-bit and 64-bit elements.
15155 (vecdupssescalarmodesuffix): New mode attribute.
15156 (vec_dup<mode>): Use it.
15158 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
15161 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
15162 frame if argument is passed on stack.
15164 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15167 * ree.c (combine_reaching_defs): Optimize also
15168 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
15169 reg2=any_extend(exp); reg1=reg2;, formatting fix.
15171 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15173 PR middle-end/83189
15174 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
15176 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15178 PR middle-end/83718
15179 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
15180 after they are computed.
15182 2018-01-11 Bin Cheng <bin.cheng@arm.com>
15184 PR tree-optimization/83695
15185 * gimple-loop-linterchange.cc
15186 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
15187 reset cached scev information after interchange.
15188 (pass_linterchange::execute): Remove call to scev_reset_htab.
15190 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15192 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
15193 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
15194 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
15195 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
15196 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
15197 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
15198 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
15199 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
15200 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
15201 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
15202 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
15203 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
15204 (V_lane_reg): Likewise.
15205 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
15207 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
15208 (vfmal_lane_low<mode>_intrinsic,
15209 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
15210 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
15211 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
15212 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
15213 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
15214 vfmsl_lane_high<mode>_intrinsic): New define_insns.
15216 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15218 * config/arm/arm-cpus.in (fp16fml): New feature.
15219 (ALL_SIMD): Add fp16fml.
15220 (armv8.2-a): Add fp16fml as an option.
15221 (armv8.3-a): Likewise.
15222 (armv8.4-a): Add fp16fml as part of fp16.
15223 * config/arm/arm.h (TARGET_FP16FML): Define.
15224 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
15226 * config/arm/arm-modes.def (V2HF): Define.
15227 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
15228 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
15229 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
15230 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
15231 vfmsl_low, vfmsl_high): New set of builtins.
15232 * config/arm/iterators.md (PLUSMINUS): New code iterator.
15233 (vfml_op): New code attribute.
15234 (VFMLHALVES): New int iterator.
15235 (VFML, VFMLSEL): New mode attributes.
15236 (V_reg): Define mapping for V2HF.
15237 (V_hi, V_lo): New mode attributes.
15238 (VF_constraint): Likewise.
15239 (vfml_half, vfml_half_selector): New int attributes.
15240 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
15242 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
15243 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
15245 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
15246 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
15247 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
15248 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
15250 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
15251 Document new effective target and option set.
15253 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15255 * config/arm/arm-cpus.in (armv8_4): New feature.
15256 (ARMv8_4a): New fgroup.
15257 (armv8.4-a): New arch.
15258 * config/arm/arm-tables.opt: Regenerate.
15259 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
15260 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
15261 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
15262 Add matching rules for -march=armv8.4-a and extensions.
15263 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
15265 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
15268 * config/rx/rx.md (BW): New mode attribute.
15269 (sync_lock_test_and_setsi): Add mode suffix to insn output.
15271 2018-01-11 Richard Biener <rguenther@suse.de>
15273 PR tree-optimization/83435
15274 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
15275 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
15276 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
15278 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15279 Alan Hayward <alan.hayward@arm.com>
15280 David Sherwood <david.sherwood@arm.com>
15282 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
15284 (aarch64_classify_address): Initialize it. Track polynomial offsets.
15285 (aarch64_print_address_internal): Use it to check for a zero offset.
15287 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15288 Alan Hayward <alan.hayward@arm.com>
15289 David Sherwood <david.sherwood@arm.com>
15291 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
15292 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
15293 Return a poly_int64 rather than a HOST_WIDE_INT.
15294 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
15295 rather than a HOST_WIDE_INT.
15296 * config/aarch64/aarch64.h (aarch64_frame): Protect with
15297 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
15298 hard_fp_offset, frame_size, initial_adjust, callee_offset and
15299 final_offset from HOST_WIDE_INT to poly_int64.
15300 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
15301 to_constant when getting the number of units in an Advanced SIMD
15303 (aarch64_builtin_vectorized_function): Check for a constant number
15305 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
15307 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
15308 attribute instead of GET_MODE_NUNITS.
15309 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
15310 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
15311 GET_MODE_SIZE for fixed-size registers.
15312 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
15313 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
15314 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
15315 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
15316 (aarch64_print_operand, aarch64_print_address_internal)
15317 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
15318 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
15319 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
15320 Handle polynomial GET_MODE_SIZE.
15321 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
15322 wider than SImode without modification.
15323 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
15324 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
15325 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
15326 passing and returning SVE modes.
15327 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
15328 rather than GEN_INT.
15329 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
15330 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
15331 (aarch64_allocate_and_probe_stack_space): Likewise.
15332 (aarch64_layout_frame): Cope with polynomial offsets.
15333 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
15334 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
15335 polynomial offsets.
15336 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
15337 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
15338 poly_int64 rather than a HOST_WIDE_INT.
15339 (aarch64_get_separate_components, aarch64_process_components)
15340 (aarch64_expand_prologue, aarch64_expand_epilogue)
15341 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
15342 (aarch64_anchor_offset): New function, split out from...
15343 (aarch64_legitimize_address): ...here.
15344 (aarch64_builtin_vectorization_cost): Handle polynomial
15345 TYPE_VECTOR_SUBPARTS.
15346 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
15348 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
15349 number of elements from the PARALLEL rather than the mode.
15350 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
15351 rather than GET_MODE_BITSIZE.
15352 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
15353 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
15354 (aarch64_expand_vec_perm_const_1): Handle polynomial
15355 d->perm.length () and d->perm elements.
15356 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
15357 Apply to_constant to d->perm elements.
15358 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
15359 polynomial CONST_VECTOR_NUNITS.
15360 (aarch64_move_pointer): Take amount as a poly_int64 rather
15362 (aarch64_progress_pointer): Avoid temporary variable.
15363 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
15364 the mode attribute instead of GET_MODE.
15366 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15367 Alan Hayward <alan.hayward@arm.com>
15368 David Sherwood <david.sherwood@arm.com>
15370 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
15371 x exists before using it.
15372 (aarch64_add_constant_internal): Rename to...
15373 (aarch64_add_offset_1): ...this. Replace regnum with separate
15374 src and dest rtxes. Handle the case in which they're different,
15375 including when the offset is zero. Replace scratchreg with an rtx.
15376 Use 2 additions if there is no spare register into which we can
15377 move a 16-bit constant.
15378 (aarch64_add_constant): Delete.
15379 (aarch64_add_offset): Replace reg with separate src and dest
15380 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
15381 Use aarch64_add_offset_1.
15382 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
15383 an rtx rather than an int. Take the delta as a poly_int64
15384 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
15385 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
15386 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
15387 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
15388 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
15389 and aarch64_add_sp.
15390 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
15391 aarch64_add_constant.
15393 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15395 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
15396 Use scalar_float_mode.
15398 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15400 * config/aarch64/aarch64-simd.md
15401 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
15402 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
15403 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
15404 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
15405 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
15406 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
15407 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
15408 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
15409 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
15410 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
15412 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15415 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
15416 targ_options->x_arm_arch_string is non NULL.
15418 2018-01-11 Tamar Christina <tamar.christina@arm.com>
15420 * config/aarch64/aarch64.h
15421 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
15423 2018-01-11 Sudakshina Das <sudi.das@arm.com>
15426 * expmed.c (emit_store_flag_force): Swap if const op0
15427 and change VOIDmode to mode of op0.
15429 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15431 PR rtl-optimization/83761
15432 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
15433 than bytes to mode_for_size.
15435 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15437 PR middle-end/83189
15438 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
15439 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
15442 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15444 PR middle-end/83575
15445 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
15446 when in layout mode.
15447 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
15448 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
15451 2018-01-10 Michael Collison <michael.collison@arm.com>
15453 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
15454 * config/aarch64/aarch64-option-extension.def: Add
15455 AARCH64_OPT_EXTENSION of 'fp16fml'.
15456 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15457 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
15458 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
15459 * config/aarch64/constraints.md (Ui7): New constraint.
15460 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
15461 (VFMLA_SEL_W): Ditto.
15464 (VFMLA16_LOW): New int iterator.
15465 (VFMLA16_HIGH): Ditto.
15466 (UNSPEC_FMLAL): New unspec.
15467 (UNSPEC_FMLSL): Ditto.
15468 (UNSPEC_FMLAL2): Ditto.
15469 (UNSPEC_FMLSL2): Ditto.
15470 (f16mac): New code attribute.
15471 * config/aarch64/aarch64-simd-builtins.def
15472 (aarch64_fmlal_lowv2sf): Ditto.
15473 (aarch64_fmlsl_lowv2sf): Ditto.
15474 (aarch64_fmlalq_lowv4sf): Ditto.
15475 (aarch64_fmlslq_lowv4sf): Ditto.
15476 (aarch64_fmlal_highv2sf): Ditto.
15477 (aarch64_fmlsl_highv2sf): Ditto.
15478 (aarch64_fmlalq_highv4sf): Ditto.
15479 (aarch64_fmlslq_highv4sf): Ditto.
15480 (aarch64_fmlal_lane_lowv2sf): Ditto.
15481 (aarch64_fmlsl_lane_lowv2sf): Ditto.
15482 (aarch64_fmlal_laneq_lowv2sf): Ditto.
15483 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
15484 (aarch64_fmlalq_lane_lowv4sf): Ditto.
15485 (aarch64_fmlsl_lane_lowv4sf): Ditto.
15486 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
15487 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
15488 (aarch64_fmlal_lane_highv2sf): Ditto.
15489 (aarch64_fmlsl_lane_highv2sf): Ditto.
15490 (aarch64_fmlal_laneq_highv2sf): Ditto.
15491 (aarch64_fmlsl_laneq_highv2sf): Ditto.
15492 (aarch64_fmlalq_lane_highv4sf): Ditto.
15493 (aarch64_fmlsl_lane_highv4sf): Ditto.
15494 (aarch64_fmlalq_laneq_highv4sf): Ditto.
15495 (aarch64_fmlsl_laneq_highv4sf): Ditto.
15496 * config/aarch64/aarch64-simd.md:
15497 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
15498 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15499 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15500 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15501 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15502 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15503 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15504 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15505 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15506 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15507 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15508 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15509 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15510 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15511 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15512 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15513 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15514 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15515 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
15516 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
15517 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
15518 (vfmlsl_low_u32): Ditto.
15519 (vfmlalq_low_u32): Ditto.
15520 (vfmlslq_low_u32): Ditto.
15521 (vfmlal_high_u32): Ditto.
15522 (vfmlsl_high_u32): Ditto.
15523 (vfmlalq_high_u32): Ditto.
15524 (vfmlslq_high_u32): Ditto.
15525 (vfmlal_lane_low_u32): Ditto.
15526 (vfmlsl_lane_low_u32): Ditto.
15527 (vfmlal_laneq_low_u32): Ditto.
15528 (vfmlsl_laneq_low_u32): Ditto.
15529 (vfmlalq_lane_low_u32): Ditto.
15530 (vfmlslq_lane_low_u32): Ditto.
15531 (vfmlalq_laneq_low_u32): Ditto.
15532 (vfmlslq_laneq_low_u32): Ditto.
15533 (vfmlal_lane_high_u32): Ditto.
15534 (vfmlsl_lane_high_u32): Ditto.
15535 (vfmlal_laneq_high_u32): Ditto.
15536 (vfmlsl_laneq_high_u32): Ditto.
15537 (vfmlalq_lane_high_u32): Ditto.
15538 (vfmlslq_lane_high_u32): Ditto.
15539 (vfmlalq_laneq_high_u32): Ditto.
15540 (vfmlslq_laneq_high_u32): Ditto.
15541 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15542 (AARCH64_FL_FOR_ARCH8_4): New.
15543 (AARCH64_ISA_F16FML): New ISA flag.
15544 (TARGET_F16FML): New feature flag for fp16fml.
15545 (doc/invoke.texi): Document new fp16fml option.
15547 2018-01-10 Michael Collison <michael.collison@arm.com>
15549 * config/aarch64/aarch64-builtins.c:
15550 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15551 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15552 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15553 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15554 (AARCH64_ISA_SHA3): New ISA flag.
15555 (TARGET_SHA3): New feature flag for sha3.
15556 * config/aarch64/iterators.md (sha512_op): New int attribute.
15557 (CRYPTO_SHA512): New int iterator.
15558 (UNSPEC_SHA512H): New unspec.
15559 (UNSPEC_SHA512H2): Ditto.
15560 (UNSPEC_SHA512SU0): Ditto.
15561 (UNSPEC_SHA512SU1): Ditto.
15562 * config/aarch64/aarch64-simd-builtins.def
15563 (aarch64_crypto_sha512hqv2di): New builtin.
15564 (aarch64_crypto_sha512h2qv2di): Ditto.
15565 (aarch64_crypto_sha512su0qv2di): Ditto.
15566 (aarch64_crypto_sha512su1qv2di): Ditto.
15567 (aarch64_eor3qv8hi): Ditto.
15568 (aarch64_rax1qv2di): Ditto.
15569 (aarch64_xarqv2di): Ditto.
15570 (aarch64_bcaxqv8hi): Ditto.
15571 * config/aarch64/aarch64-simd.md:
15572 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15573 (aarch64_crypto_sha512su0qv2di): Ditto.
15574 (aarch64_crypto_sha512su1qv2di): Ditto.
15575 (aarch64_eor3qv8hi): Ditto.
15576 (aarch64_rax1qv2di): Ditto.
15577 (aarch64_xarqv2di): Ditto.
15578 (aarch64_bcaxqv8hi): Ditto.
15579 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15580 (vsha512h2q_u64): Ditto.
15581 (vsha512su0q_u64): Ditto.
15582 (vsha512su1q_u64): Ditto.
15583 (veor3q_u16): Ditto.
15584 (vrax1q_u64): Ditto.
15585 (vxarq_u64): Ditto.
15586 (vbcaxq_u16): Ditto.
15587 * config/arm/types.md (crypto_sha512): New type attribute.
15588 (crypto_sha3): Ditto.
15589 (doc/invoke.texi): Document new sha3 option.
15591 2018-01-10 Michael Collison <michael.collison@arm.com>
15593 * config/aarch64/aarch64-builtins.c:
15594 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15595 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15596 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15597 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15598 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15599 (AARCH64_ISA_SM4): New ISA flag.
15600 (TARGET_SM4): New feature flag for sm4.
15601 * config/aarch64/aarch64-simd-builtins.def
15602 (aarch64_sm3ss1qv4si): Ditto.
15603 (aarch64_sm3tt1aq4si): Ditto.
15604 (aarch64_sm3tt1bq4si): Ditto.
15605 (aarch64_sm3tt2aq4si): Ditto.
15606 (aarch64_sm3tt2bq4si): Ditto.
15607 (aarch64_sm3partw1qv4si): Ditto.
15608 (aarch64_sm3partw2qv4si): Ditto.
15609 (aarch64_sm4eqv4si): Ditto.
15610 (aarch64_sm4ekeyqv4si): Ditto.
15611 * config/aarch64/aarch64-simd.md:
15612 (aarch64_sm3ss1qv4si): Ditto.
15613 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15614 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15615 (aarch64_sm4eqv4si): Ditto.
15616 (aarch64_sm4ekeyqv4si): Ditto.
15617 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15618 (sm3part_op): Ditto.
15619 (CRYPTO_SM3TT): Ditto.
15620 (CRYPTO_SM3PART): Ditto.
15621 (UNSPEC_SM3SS1): New unspec.
15622 (UNSPEC_SM3TT1A): Ditto.
15623 (UNSPEC_SM3TT1B): Ditto.
15624 (UNSPEC_SM3TT2A): Ditto.
15625 (UNSPEC_SM3TT2B): Ditto.
15626 (UNSPEC_SM3PARTW1): Ditto.
15627 (UNSPEC_SM3PARTW2): Ditto.
15628 (UNSPEC_SM4E): Ditto.
15629 (UNSPEC_SM4EKEY): Ditto.
15630 * config/aarch64/constraints.md (Ui2): New constraint.
15631 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15632 * config/arm/types.md (crypto_sm3): New type attribute.
15633 (crypto_sm4): Ditto.
15634 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15635 (vsm3tt1aq_u32): Ditto.
15636 (vsm3tt1bq_u32): Ditto.
15637 (vsm3tt2aq_u32): Ditto.
15638 (vsm3tt2bq_u32): Ditto.
15639 (vsm3partw1q_u32): Ditto.
15640 (vsm3partw2q_u32): Ditto.
15641 (vsm4eq_u32): Ditto.
15642 (vsm4ekeyq_u32): Ditto.
15643 (doc/invoke.texi): Document new sm4 option.
15645 2018-01-10 Michael Collison <michael.collison@arm.com>
15647 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15648 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15649 (AARCH64_FL_FOR_ARCH8_4): New.
15650 (AARCH64_FL_V8_4): New flag.
15651 (doc/invoke.texi): Document new armv8.4-a option.
15653 2018-01-10 Michael Collison <michael.collison@arm.com>
15655 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15656 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15657 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15658 * config/aarch64/aarch64-option-extension.def: Add
15659 AARCH64_OPT_EXTENSION of 'sha2'.
15660 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15661 (crypto): Disable sha2 and aes if crypto disabled.
15662 (crypto): Enable aes and sha2 if enabled.
15663 (simd): Disable sha2 and aes if simd disabled.
15664 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15666 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15667 (TARGET_SHA2): New feature flag for sha2.
15668 (TARGET_AES): New feature flag for aes.
15669 * config/aarch64/aarch64-simd.md:
15670 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15671 conditional on TARGET_AES.
15672 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15673 (aarch64_crypto_sha1hsi): Make pattern conditional
15675 (aarch64_crypto_sha1hv4si): Ditto.
15676 (aarch64_be_crypto_sha1hv4si): Ditto.
15677 (aarch64_crypto_sha1su1v4si): Ditto.
15678 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15679 (aarch64_crypto_sha1su0v4si): Ditto.
15680 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15681 (aarch64_crypto_sha256su0v4si): Ditto.
15682 (aarch64_crypto_sha256su1v4si): Ditto.
15683 (doc/invoke.texi): Document new aes and sha2 options.
15685 2018-01-10 Martin Sebor <msebor@redhat.com>
15687 PR tree-optimization/83781
15688 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15691 2018-01-11 Martin Sebor <msebor@gmail.com>
15692 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15694 PR tree-optimization/83501
15695 PR tree-optimization/81703
15697 * tree-ssa-strlen.c (get_string_cst): Rename...
15698 (get_string_len): ...to this. Handle global constants.
15699 (handle_char_store): Adjust.
15701 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15702 Jim Wilson <jimw@sifive.com>
15704 * config/riscv/riscv-protos.h (riscv_output_return): New.
15705 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15706 (riscv_attribute_table, riscv_output_return),
15707 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15708 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15709 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15711 (riscv_expand_prologue): Add early return for naked function.
15712 (riscv_expand_epilogue): Likewise.
15713 (riscv_function_ok_for_sibcall): Return false for naked function.
15714 (riscv_set_current_function): New.
15715 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15716 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15717 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15718 * doc/extend.texi (RISC-V Function Attributes): New.
15720 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15722 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15723 check for 128-bit long double before checking TCmode.
15724 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15725 128-bit long doubles before checking TFmode or TCmode.
15726 (FLOAT128_IBM_P): Likewise.
15728 2018-01-10 Martin Sebor <msebor@redhat.com>
15730 PR tree-optimization/83671
15731 * builtins.c (c_strlen): Unconditionally return zero for the empty
15733 Use -Warray-bounds for warnings.
15734 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15735 for non-constant array indices with COMPONENT_REF, arrays of
15736 arrays, and pointers to arrays.
15737 (gimple_fold_builtin_strlen): Determine and set length range for
15738 non-constant character arrays.
15740 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15742 PR middle-end/81897
15743 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15746 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15748 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15750 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15753 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15754 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15755 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15756 indexed_or_indirect_operand predicate.
15757 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15758 (*vsx_le_perm_load_v8hi): Likewise.
15759 (*vsx_le_perm_load_v16qi): Likewise.
15760 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15761 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15762 (*vsx_le_perm_store_v8hi): Likewise.
15763 (*vsx_le_perm_store_v16qi): Likewise.
15764 (eight unnamed splitters): Likewise.
15766 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15768 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15769 * config/rs6000/emmintrin.h: Likewise.
15770 * config/rs6000/mmintrin.h: Likewise.
15771 * config/rs6000/xmmintrin.h: Likewise.
15773 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15776 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15778 * tree.c (tree_nop_conversion): Return true for location wrapper
15780 (maybe_wrap_with_location): New function.
15781 (selftest::check_strip_nops): New function.
15782 (selftest::test_location_wrappers): New function.
15783 (selftest::tree_c_tests): Call it.
15784 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15785 (maybe_wrap_with_location): New decl.
15786 (EXPR_LOCATION_WRAPPER_P): New macro.
15787 (location_wrapper_p): New inline function.
15788 (tree_strip_any_location_wrapper): New inline function.
15790 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15793 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15794 stack_realign_offset for the largest alignment of stack slot
15796 (ix86_find_max_used_stack_alignment): New function.
15797 (ix86_finalize_stack_frame_flags): Use it. Set
15798 max_used_stack_alignment if we don't realign stack.
15799 * config/i386/i386.h (machine_function): Add
15800 max_used_stack_alignment.
15802 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15804 * config/arm/arm.opt (-mbranch-cost): New option.
15805 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15808 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15811 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15812 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15814 2018-01-10 Richard Biener <rguenther@suse.de>
15817 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15818 early out so it also covers the case where we have a non-NULL
15821 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15823 PR tree-optimization/83753
15824 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15825 for non-strided grouped accesses if the number of elements is 1.
15827 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15830 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15831 * i386.h (TARGET_USE_GATHER): Define.
15832 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15834 2018-01-10 Martin Liska <mliska@suse.cz>
15837 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15838 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15840 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15841 CLEANUP_NO_PARTITIONING is not set.
15843 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15845 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15846 for vectors, as a partial revert of r254296.
15847 * rtl.h (const_vec_p): Delete.
15848 (const_vec_duplicate_p): Don't test for vector CONSTs.
15849 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15850 * expmed.c (make_tree): Likewise.
15853 * common.md (E, F): Use CONSTANT_P instead of checking for
15855 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15856 checking for CONST_VECTOR.
15858 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15860 PR middle-end/83575
15861 * predict.c (force_edge_cold): Handle in more sane way edges
15862 with no prediction.
15864 2018-01-09 Carl Love <cel@us.ibm.com>
15866 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15868 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15869 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15870 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15871 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15872 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15873 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15874 * config/rs6000/rs6000-protos.h: Add extern defition for
15875 rs6000_generate_float2_double_code.
15876 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15878 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15879 (float2_v2df): Add define_expand.
15881 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15884 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15885 op_mode in the force_to_mode call.
15887 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15889 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15890 instead of checking each element individually.
15891 (aarch64_evpc_uzp): Likewise.
15892 (aarch64_evpc_zip): Likewise.
15893 (aarch64_evpc_ext): Likewise.
15894 (aarch64_evpc_rev): Likewise.
15895 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15896 instead of checking each element individually. Return true without
15898 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15899 whether all selected elements come from the same input, instead of
15900 checking each element individually. Remove calls to gen_rtx_REG,
15901 start_sequence and end_sequence and instead assert that no rtl is
15904 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15906 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15907 order of HIGH and CONST checks.
15909 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15911 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15912 if the destination isn't an SSA_NAME.
15914 2018-01-09 Richard Biener <rguenther@suse.de>
15916 PR tree-optimization/83668
15917 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15919 (canonicalize_loop_form): ... here, renamed from ...
15920 (canonicalize_loop_closed_ssa_form): ... this and amended to
15921 swap successor edges for loop exit blocks to make us use
15922 the RPO order we need for initial schedule generation.
15924 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15926 PR tree-optimization/64811
15927 * match.pd: When optimizing comparisons with Inf, avoid
15928 introducing or losing exceptions from comparisons with NaN.
15930 2018-01-09 Martin Liska <mliska@suse.cz>
15933 * asan.c (shadow_mem_size): Add gcc_assert.
15935 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15937 Don't save registers in main().
15940 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15941 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15942 * config/avr/avr.c (avr_set_current_function): Don't error if
15943 naked, OS_task or OS_main are specified at the same time.
15944 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15946 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15948 * common/config/avr/avr-common.c (avr_option_optimization_table):
15949 Switch on -mmain-is-OS_task for optimizing compilations.
15951 2018-01-09 Richard Biener <rguenther@suse.de>
15953 PR tree-optimization/83572
15954 * graphite.c: Include cfganal.h.
15955 (graphite_transform_loops): Connect infinite loops to exit
15956 and remove fake edges at the end.
15958 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15960 * ipa-inline.c (edge_badness): Revert accidental checkin.
15962 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15965 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15966 symbols; not inline clones.
15968 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15971 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15972 hard registers. Formatting fixes.
15974 PR preprocessor/83722
15975 * gcc.c (try_generate_repro): Pass
15976 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15977 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15980 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15981 Kito Cheng <kito.cheng@gmail.com>
15983 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15984 (riscv_leaf_function_p): Delete.
15985 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15987 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15989 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15991 (do_ifelse): New function.
15992 (do_isel): New function.
15993 (do_sub3): New function.
15994 (do_add3): New function.
15995 (do_load_mask_compare): New function.
15996 (do_overlap_load_compare): New function.
15997 (expand_compare_loop): New function.
15998 (expand_block_compare): Call expand_compare_loop() when appropriate.
15999 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
16000 option description.
16001 (-mblock-compare-inline-loop-limit): New option.
16003 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16006 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
16007 Reverse order of second and third operands in first alternative.
16008 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
16009 of first and second elements in UNSPEC_VPERMR vector.
16010 (altivec_expand_vec_perm_le): Likewise.
16012 2018-01-08 Jeff Law <law@redhat.com>
16014 PR rtl-optimizatin/81308
16015 * tree-switch-conversion.c (cfg_altered): New file scoped static.
16016 (process_switch): If group_case_labels makes a change, then set
16018 (pass_convert_switch::execute): If a switch is converted, then
16019 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
16021 PR rtl-optimization/81308
16022 * recog.c (split_all_insns): Conditionally cleanup the CFG after
16025 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
16027 PR target/83663 - Revert r255946
16028 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
16029 generation for cases where splatting a value is not useful.
16030 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
16031 across a vec_duplicate and a paradoxical subreg forming a vector
16032 mode to a vec_concat.
16034 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16036 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
16037 -march=armv8.3-a variants.
16038 * config/arm/t-multilib: Likewise.
16039 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
16041 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16043 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
16045 (cceq_ior_compare_complement): Give it a name so I can use it, and
16046 change boolean_or_operator predicate to boolean_operator so it can
16047 be used to generate a crand.
16048 (eqne): New code iterator.
16049 (bd/bd_neg): New code_attrs.
16050 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
16051 a single define_insn.
16052 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
16053 decrement (bdnzt/bdnzf/bdzt/bdzf).
16054 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
16055 with the new names of the branch decrement patterns, and added the
16056 names of the branch decrement conditional patterns.
16058 2018-01-08 Richard Biener <rguenther@suse.de>
16060 PR tree-optimization/83563
16061 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
16064 2018-01-08 Richard Biener <rguenther@suse.de>
16066 PR middle-end/83713
16067 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
16069 2018-01-08 Richard Biener <rguenther@suse.de>
16071 PR tree-optimization/83685
16072 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
16073 references to abnormals.
16075 2018-01-08 Richard Biener <rguenther@suse.de>
16078 * dwarf2out.c (output_indirect_strings): Handle empty
16079 skeleton_debug_str_hash.
16080 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
16082 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16084 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
16085 (emit_store_direct): Likewise.
16086 (arc_trampoline_adjust_address): Likewise.
16087 (arc_asm_trampoline_template): New function.
16088 (arc_initialize_trampoline): Use asm_trampoline_template.
16089 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
16090 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
16091 * config/arc/arc.md (flush_icache): Delete pattern.
16093 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16095 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
16096 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
16099 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16102 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
16103 by not USED_FOR_TARGET.
16104 (make_pass_resolve_sw_modes): Likewise.
16106 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16108 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
16111 2018-01-08 Richard Biener <rguenther@suse.de>
16113 PR middle-end/83580
16114 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
16116 2018-01-08 Richard Biener <rguenther@suse.de>
16118 PR middle-end/83517
16119 * match.pd ((t * 2) / 2) -> t): Add missing :c.
16121 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
16123 PR middle-end/81897
16124 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
16125 basic blocks with a small number of successors.
16126 (convert_control_dep_chain_into_preds): Improve handling of
16128 (dump_predicates): Split apart into...
16129 (dump_pred_chain): ...here...
16130 (dump_pred_info): ...and here.
16131 (can_one_predicate_be_invalidated_p): Add debugging printfs.
16132 (can_chain_union_be_invalidated_p): Improve check for invalidation
16134 (uninit_uses_cannot_happen): Avoid unnecessary if
16135 convert_control_dep_chain_into_preds yielded nothing.
16137 2018-01-06 Martin Sebor <msebor@redhat.com>
16139 PR tree-optimization/83640
16140 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
16141 subtracting negative offset from size.
16142 (builtin_access::overlap): Adjust offset bounds of the access to fall
16143 within the size of the object if possible.
16145 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
16147 PR rtl-optimization/83699
16148 * expmed.c (extract_bit_field_1): Restrict the vector usage of
16149 extract_bit_field_as_subreg to cases in which the extracted
16150 value is also a vector.
16152 * lra-constraints.c (process_alt_operands): Test for the equivalence
16153 substitutions when detecting a possible reload cycle.
16155 2018-01-06 Jakub Jelinek <jakub@redhat.com>
16158 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
16159 by default if flag_selective_schedling{,2}. Formatting fixes.
16161 PR rtl-optimization/83682
16162 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
16163 if it has non-VECTOR_MODE element mode.
16164 (vec_duplicate_p): Likewise.
16166 PR middle-end/83694
16167 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
16168 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
16170 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16173 * config/i386/i386-builtin.def
16174 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
16175 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
16176 Require also OPTION_MASK_ISA_AVX512F in addition to
16177 OPTION_MASK_ISA_GFNI.
16178 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
16179 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
16180 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
16181 to OPTION_MASK_ISA_GFNI.
16182 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
16183 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
16184 OPTION_MASK_ISA_AVX512BW.
16185 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
16186 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
16187 addition to OPTION_MASK_ISA_GFNI.
16188 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
16189 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
16190 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
16191 to OPTION_MASK_ISA_GFNI.
16192 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
16193 a requirement for all ISAs rather than any of them with a few
16195 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
16197 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
16198 bitmasks to be enabled with 3 exceptions, instead of requiring any
16199 enabled ISA with lots of exceptions.
16200 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
16201 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
16202 Change avx512bw in isa attribute to avx512f.
16203 * config/i386/sgxintrin.h: Add license boilerplate.
16204 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
16205 to __AVX512F__ and __AVX512VL to __AVX512VL__.
16206 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
16207 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
16209 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
16210 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
16211 temporarily sse2 rather than sse if not enabled already.
16214 * config/i386/sse.md (VI248_VLBW): Rename to ...
16215 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
16216 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
16217 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
16218 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
16219 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
16220 mode iterator instead of VI248_VLBW.
16222 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
16224 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
16225 (record_modified): Skip clobbers; add debug output.
16226 (param_change_prob): Use sreal frequencies.
16228 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16230 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
16231 punt for user-aligned variables.
16233 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16235 * tree-chrec.c (chrec_contains_symbols): Return true for
16238 2018-01-05 Sudakshina Das <sudi.das@arm.com>
16241 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
16242 of (x|y) == x for BICS pattern.
16244 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16246 PR tree-optimization/83605
16247 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
16248 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
16251 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
16253 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
16254 * config/epiphany/rtems.h: New file.
16256 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16257 Uros Bizjak <ubizjak@gmail.com>
16260 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
16261 QIreg_operand instead of register_operand predicate.
16262 * config/i386/i386.c (ix86_rop_should_change_byte_p,
16263 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
16264 comments instead of -fmitigate[-_]rop.
16266 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16269 * cgraphunit.c (symbol_table::compile): Switch to text_section
16270 before calling assembly_start debug hook.
16271 * run-rtl-passes.c (run_rtl_passes): Likewise.
16274 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16276 * tree-vrp.c (extract_range_from_binary_expr_1): Check
16277 range_int_cst_p rather than !symbolic_range_p before calling
16278 extract_range_from_multiplicative_op_1.
16280 2018-01-04 Jeff Law <law@redhat.com>
16282 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
16283 redundant test in assertion.
16285 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16287 * doc/rtl.texi: Document machine_mode wrapper classes.
16289 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16291 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
16292 using tree_to_uhwi.
16294 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16296 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
16297 the VEC_PERM_EXPR fold to fail.
16299 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16302 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
16303 to switched_sections.
16305 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16308 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
16309 test for d.testing.
16311 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
16314 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
16315 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
16317 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16320 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
16321 is BLKmode and bitpos not zero or mode change is needed.
16323 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16326 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
16329 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
16332 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
16333 instead of MULT rtx. Update all corresponding splitters.
16334 (*saddl_se): Ditto.
16335 (*ssub<modesuffix>): Ditto.
16336 (*ssubl_se): Ditto.
16337 (*cmp_sadd_di): Update split patterns.
16338 (*cmp_sadd_si): Ditto.
16339 (*cmp_sadd_sidi): Ditto.
16340 (*cmp_ssub_di): Ditto.
16341 (*cmp_ssub_si): Ditto.
16342 (*cmp_ssub_sidi): Ditto.
16343 * config/alpha/predicates.md (const23_operand): New predicate.
16344 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
16345 Look for ASHIFT, not MULT inner operand.
16346 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
16348 2018-01-04 Martin Liska <mliska@suse.cz>
16350 PR gcov-profile/83669
16351 * gcov.c (output_intermediate_file): Add version to intermediate
16353 * doc/gcov.texi: Document new field 'version' in intermediate
16354 file format. Fix location of '-k' option of gcov command.
16356 2018-01-04 Martin Liska <mliska@suse.cz>
16359 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
16361 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16363 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
16365 2018-01-03 Martin Sebor <msebor@redhat.com>
16367 PR tree-optimization/83655
16368 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
16369 checking calls with invalid arguments.
16371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16373 * tree-vect-stmts.c (vect_get_store_rhs): New function.
16374 (vectorizable_mask_load_store): Delete.
16375 (vectorizable_call): Return false for masked loads and stores.
16376 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
16377 instead of gimple_assign_rhs1.
16378 (vectorizable_load): Handle IFN_MASK_LOAD.
16379 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
16381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16383 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
16385 (vectorizable_mask_load_store): ...here.
16386 (vectorizable_load): ...and here.
16388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16390 * tree-vect-stmts.c (vect_build_all_ones_mask)
16391 (vect_build_zero_merge_argument): New functions, split out from...
16392 (vectorizable_load): ...here.
16394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16396 * tree-vect-stmts.c (vect_check_store_rhs): New function,
16398 (vectorizable_mask_load_store): ...here.
16399 (vectorizable_store): ...and here.
16401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16403 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
16405 (vectorizable_mask_load_store): ...here.
16407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16409 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
16410 (vect_model_store_cost): Take a vec_load_store_type instead of a
16412 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
16413 (vect_model_store_cost): Take a vec_load_store_type instead of a
16415 (vectorizable_mask_load_store): Update accordingly.
16416 (vectorizable_store): Likewise.
16417 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
16419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16421 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
16422 IFN_MASK_LOAD calls here rather than...
16423 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
16425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16426 Alan Hayward <alan.hayward@arm.com>
16427 David Sherwood <david.sherwood@arm.com>
16429 * expmed.c (extract_bit_field_1): For vector extracts,
16430 fall back to extract_bit_field_as_subreg if vec_extract
16433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16434 Alan Hayward <alan.hayward@arm.com>
16435 David Sherwood <david.sherwood@arm.com>
16437 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
16438 they are variable or constant sized.
16439 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
16440 slots for constant-sized data.
16442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16443 Alan Hayward <alan.hayward@arm.com>
16444 David Sherwood <david.sherwood@arm.com>
16446 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
16447 handling COND_EXPRs with boolean comparisons, try to find a better
16448 basis for the mask type than the boolean itself.
16450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16452 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
16453 is calculated and how it can be overridden.
16454 * genmodes.c (max_bitsize_mode_any_mode): New variable.
16455 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
16457 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
16460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16461 Alan Hayward <alan.hayward@arm.com>
16462 David Sherwood <david.sherwood@arm.com>
16464 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
16465 Remove the mode argument.
16466 (aarch64_simd_valid_immediate): Remove the mode and inverse
16468 * config/aarch64/iterators.md (bitsize): New iterator.
16469 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
16470 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
16471 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
16472 aarch64_simd_valid_immediate.
16473 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
16474 (aarch64_reg_or_bic_imm): Likewise.
16475 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
16476 with an insn_type enum and msl with a modifier_type enum.
16477 Replace element_width with a scalar_mode. Change the shift
16478 to unsigned int. Add constructors for scalar_float_mode and
16479 scalar_int_mode elements.
16480 (aarch64_vect_float_const_representable_p): Delete.
16481 (aarch64_can_const_movi_rtx_p)
16482 (aarch64_simd_scalar_immediate_valid_for_move)
16483 (aarch64_simd_make_constant): Update call to
16484 aarch64_simd_valid_immediate.
16485 (aarch64_advsimd_valid_immediate_hs): New function.
16486 (aarch64_advsimd_valid_immediate): Likewise.
16487 (aarch64_simd_valid_immediate): Remove mode and inverse
16488 arguments. Rewrite to use the above. Use const_vec_duplicate_p
16489 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
16490 and aarch64_float_const_representable_p on the result.
16491 (aarch64_output_simd_mov_immediate): Remove mode argument.
16492 Update call to aarch64_simd_valid_immediate and use of
16493 simd_immediate_info.
16494 (aarch64_output_scalar_simd_mov_immediate): Update call
16497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16498 Alan Hayward <alan.hayward@arm.com>
16499 David Sherwood <david.sherwood@arm.com>
16501 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16502 (mode_nunits): Likewise CONST_MODE_NUNITS.
16503 * machmode.def (ADJUST_NUNITS): Document.
16504 * genmodes.c (mode_data::need_nunits_adj): New field.
16505 (blank_mode): Update accordingly.
16506 (adj_nunits): New variable.
16507 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16509 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16510 listed in adj_nunits.
16511 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16512 listed in adj_nunits. Don't emit case statements for such modes.
16513 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16514 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16515 nothing if adj_nunits is nonnull.
16516 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
16517 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
16518 (emit_mode_fbit): Update use of print_maybe_const_decl.
16519 (emit_move_size): Likewise. Treat the array as non-const
16521 (emit_mode_adjustments): Handle adj_nunits.
16523 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16525 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
16526 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
16527 (VECTOR_MODES): Use it.
16528 (make_vector_modes): Take the prefix as an argument.
16530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16531 Alan Hayward <alan.hayward@arm.com>
16532 David Sherwood <david.sherwood@arm.com>
16534 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16535 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16536 for MODE_VECTOR_BOOL.
16537 * machmode.def (VECTOR_BOOL_MODE): Document.
16538 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16539 (make_vector_bool_mode): New function.
16540 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16542 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16543 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16545 * stor-layout.c (int_mode_for_mode): Likewise.
16546 * tree.c (build_vector_type_for_mode): Likewise.
16547 * varasm.c (output_constant_pool_2): Likewise.
16548 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16549 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16550 for MODE_VECTOR_BOOL.
16551 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16552 of mode class checks.
16553 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16554 instead of a list of mode class checks.
16555 (expand_vector_scalar_condition): Likewise.
16556 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16558 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16559 Alan Hayward <alan.hayward@arm.com>
16560 David Sherwood <david.sherwood@arm.com>
16562 * machmode.h (mode_size): Change from unsigned short to
16564 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16565 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16566 or if measurement_type is not polynomial.
16567 (fixed_size_mode::includes_p): Check for constant-sized modes.
16568 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16569 return a poly_uint16 rather than an unsigned short.
16570 (emit_mode_size): Change the type of mode_size from unsigned short
16571 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16572 (emit_mode_adjustments): Cope with polynomial vector sizes.
16573 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16575 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16577 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16578 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16579 * caller-save.c (setup_save_areas): Likewise.
16580 (replace_reg_with_saved_mem): Likewise.
16581 * calls.c (emit_library_call_value_1): Likewise.
16582 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16583 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16584 (gen_lowpart_for_combine): Likewise.
16585 * convert.c (convert_to_integer_1): Likewise.
16586 * cse.c (equiv_constant, cse_insn): Likewise.
16587 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16588 (cselib_subst_to_values): Likewise.
16589 * dce.c (word_dce_process_block): Likewise.
16590 * df-problems.c (df_word_lr_mark_ref): Likewise.
16591 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16592 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16593 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16594 (rtl_for_decl_location): Likewise.
16595 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16596 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16597 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16598 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16599 (expand_expr_real_1): Likewise.
16600 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16601 (pad_below): Likewise.
16602 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16603 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16604 * ira.c (get_subreg_tracking_sizes): Likewise.
16605 * ira-build.c (ira_create_allocno_objects): Likewise.
16606 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16607 (ira_sort_regnos_for_alter_reg): Likewise.
16608 * ira-costs.c (record_operand_costs): Likewise.
16609 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16610 (resolve_simple_move): Likewise.
16611 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16612 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16613 (lra_constraints): Likewise.
16614 (CONST_POOL_OK_P): Reject variable-sized modes.
16615 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16616 (add_pseudo_to_slot, lra_spill): Likewise.
16617 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16618 * optabs-query.c (get_best_extraction_insn): Likewise.
16619 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16620 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16621 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16622 * recog.c (offsettable_address_addr_space_p): Likewise.
16623 * regcprop.c (maybe_mode_change): Likewise.
16624 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16625 * regrename.c (build_def_use): Likewise.
16626 * regstat.c (dump_reg_info): Likewise.
16627 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16628 (find_reloads, find_reloads_subreg_address): Likewise.
16629 * reload1.c (eliminate_regs_1): Likewise.
16630 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16631 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16632 (simplify_binary_operation_1, simplify_subreg): Likewise.
16633 * targhooks.c (default_function_arg_padding): Likewise.
16634 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16635 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16636 (verify_gimple_assign_ternary): Likewise.
16637 * tree-inline.c (estimate_move_cost): Likewise.
16638 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16639 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16640 (get_address_cost_ainc): Likewise.
16641 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16642 (vect_supportable_dr_alignment): Likewise.
16643 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16644 (vectorizable_reduction): Likewise.
16645 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16646 (vectorizable_operation, vectorizable_load): Likewise.
16647 * tree.c (build_same_sized_truth_vector_type): Likewise.
16648 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16649 * var-tracking.c (emit_note_insn_var_location): Likewise.
16650 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16651 (ADDR_VEC_ALIGN): Likewise.
16653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16654 Alan Hayward <alan.hayward@arm.com>
16655 David Sherwood <david.sherwood@arm.com>
16657 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16659 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16660 or if measurement_type is polynomial.
16661 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16662 * combine.c (make_extraction): Likewise.
16663 * dse.c (find_shift_sequence): Likewise.
16664 * dwarf2out.c (mem_loc_descriptor): Likewise.
16665 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16666 (extract_bit_field, extract_low_bits): Likewise.
16667 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16668 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16669 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16670 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16671 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16672 * reload.c (find_reloads): Likewise.
16673 * reload1.c (alter_reg): Likewise.
16674 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16675 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16676 * tree-if-conv.c (predicate_mem_writes): Likewise.
16677 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16678 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16679 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16680 * valtrack.c (dead_debug_insert_temp): Likewise.
16681 * varasm.c (mergeable_constant_section): Likewise.
16682 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16685 Alan Hayward <alan.hayward@arm.com>
16686 David Sherwood <david.sherwood@arm.com>
16688 * expr.c (expand_assignment): Cope with polynomial mode sizes
16689 when assigning to a CONCAT.
16691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16692 Alan Hayward <alan.hayward@arm.com>
16693 David Sherwood <david.sherwood@arm.com>
16695 * machmode.h (mode_precision): Change from unsigned short to
16697 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16699 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16700 or if measurement_type is not polynomial.
16701 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16702 in which the mode is already known to be a scalar_int_mode.
16703 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16704 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16706 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16707 for GET_MODE_PRECISION.
16708 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16709 for GET_MODE_PRECISION.
16710 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16712 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16713 (expand_field_assignment, make_extraction): Likewise.
16714 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16715 (get_last_value): Likewise.
16716 * convert.c (convert_to_integer_1): Likewise.
16717 * cse.c (cse_insn): Likewise.
16718 * expr.c (expand_expr_real_1): Likewise.
16719 * lra-constraints.c (simplify_operand_subreg): Likewise.
16720 * optabs-query.c (can_atomic_load_p): Likewise.
16721 * optabs.c (expand_atomic_load): Likewise.
16722 (expand_atomic_store): Likewise.
16723 * ree.c (combine_reaching_defs): Likewise.
16724 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16725 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16726 * tree.h (type_has_mode_precision_p): Likewise.
16727 * ubsan.c (instrument_si_overflow): Likewise.
16729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16730 Alan Hayward <alan.hayward@arm.com>
16731 David Sherwood <david.sherwood@arm.com>
16733 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16734 polynomial numbers of units.
16735 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16736 (valid_vector_subparts_p): New function.
16737 (build_vector_type): Remove temporary shim and take the number
16738 of units as a poly_uint64 rather than an int.
16739 (build_opaque_vector_type): Take the number of units as a
16740 poly_uint64 rather than an int.
16741 * tree.c (build_vector_from_ctor): Handle polynomial
16742 TYPE_VECTOR_SUBPARTS.
16743 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16744 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16745 (build_vector_from_val): If the number of units is variable,
16746 use build_vec_duplicate_cst for constant operands and
16747 VEC_DUPLICATE_EXPR otherwise.
16748 (make_vector_type): Remove temporary is_constant ().
16749 (build_vector_type, build_opaque_vector_type): Take the number of
16750 units as a poly_uint64 rather than an int.
16751 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16753 * cfgexpand.c (expand_debug_expr): Likewise.
16754 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16755 (store_constructor, expand_expr_real_1): Likewise.
16756 (const_scalar_mask_from_tree): Likewise.
16757 * fold-const-call.c (fold_const_reduction): Likewise.
16758 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16759 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16760 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16761 (fold_relational_const): Likewise.
16762 (native_interpret_vector): Likewise. Change the size from an
16763 int to an unsigned int.
16764 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16765 TYPE_VECTOR_SUBPARTS.
16766 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16767 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16768 duplicating a non-constant operand into a variable-length vector.
16769 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16770 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16771 * ipa-icf.c (sem_variable::equals): Likewise.
16772 * match.pd: Likewise.
16773 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16774 * print-tree.c (print_node): Likewise.
16775 * stor-layout.c (layout_type): Likewise.
16776 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16777 * tree-cfg.c (verify_gimple_comparison): Likewise.
16778 (verify_gimple_assign_binary): Likewise.
16779 (verify_gimple_assign_ternary): Likewise.
16780 (verify_gimple_assign_single): Likewise.
16781 * tree-pretty-print.c (dump_generic_node): Likewise.
16782 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16783 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16784 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16785 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16786 (vect_shift_permute_load_chain): Likewise.
16787 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16788 (expand_vector_condition, optimize_vector_constructor): Likewise.
16789 (lower_vec_perm, get_compute_type): Likewise.
16790 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16791 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16792 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16793 (vect_recog_mask_conversion_pattern): Likewise.
16794 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16795 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16796 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16797 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16798 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16799 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16800 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16801 (supportable_widening_operation): Likewise.
16802 (supportable_narrowing_operation): Likewise.
16803 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16805 * varasm.c (output_constant): Likewise.
16807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16808 Alan Hayward <alan.hayward@arm.com>
16809 David Sherwood <david.sherwood@arm.com>
16811 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16812 so that both the length == 3 and length != 3 cases set up their
16813 own permute vectors. Add comments explaining why we know the
16814 number of elements is constant.
16815 (vect_permute_load_chain): Likewise.
16817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16818 Alan Hayward <alan.hayward@arm.com>
16819 David Sherwood <david.sherwood@arm.com>
16821 * machmode.h (mode_nunits): Change from unsigned char to
16823 (ONLY_FIXED_SIZE_MODES): New macro.
16824 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16825 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16826 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16828 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16829 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16830 or if measurement_type is not polynomial.
16831 * genmodes.c (ZERO_COEFFS): New macro.
16832 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16834 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16835 Use ZERO_COEFFS when emitting initializers.
16836 * data-streamer.h (bp_pack_poly_value): New function.
16837 (bp_unpack_poly_value): Likewise.
16838 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16839 for GET_MODE_NUNITS.
16840 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16841 for GET_MODE_NUNITS.
16842 * tree.c (make_vector_type): Remove temporary shim and make
16843 the real function take the number of units as a poly_uint64
16844 rather than an int.
16845 (build_vector_type_for_mode): Handle polynomial nunits.
16846 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16847 * emit-rtl.c (const_vec_series_p_1): Likewise.
16848 (gen_rtx_CONST_VECTOR): Likewise.
16849 * fold-const.c (test_vec_duplicate_folding): Likewise.
16850 * genrecog.c (validate_pattern): Likewise.
16851 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16852 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16853 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16854 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16855 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16856 * rtlanal.c (subreg_get_info): Likewise.
16857 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16858 (vect_grouped_load_supported): Likewise.
16859 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16860 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16861 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16862 (simplify_const_unary_operation, simplify_binary_operation_1)
16863 (simplify_const_binary_operation, simplify_ternary_operation)
16864 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16865 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16866 instead of CONST_VECTOR_NUNITS.
16867 * varasm.c (output_constant_pool_2): Likewise.
16868 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16869 explicit-encoded elements in the XVEC for variable-length vectors.
16871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16873 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16876 Alan Hayward <alan.hayward@arm.com>
16877 David Sherwood <david.sherwood@arm.com>
16879 * coretypes.h (fixed_size_mode): Declare.
16880 (fixed_size_mode_pod): New typedef.
16881 * builtins.h (target_builtins::x_apply_args_mode)
16882 (target_builtins::x_apply_result_mode): Change type to
16883 fixed_size_mode_pod.
16884 * builtins.c (apply_args_size, apply_result_size, result_vector)
16885 (expand_builtin_apply_args_1, expand_builtin_apply)
16886 (expand_builtin_return): Update accordingly.
16888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16890 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16891 * cselib.c (cselib_hash_rtx): Likewise.
16892 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16893 CONST_VECTOR encoding.
16895 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16896 Jeff Law <law@redhat.com>
16899 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16900 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16901 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16902 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16905 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16906 explicitly probe *sp in a noreturn function if there were any callee
16907 register saves or frame pointer is needed.
16909 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16912 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16913 BLKmode for ternary, binary or unary expressions.
16916 * var-tracking.c (delete_vta_debug_insn): New inline function.
16917 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16918 insns from get_insns () to NULL instead of each bb separately.
16919 Use delete_vta_debug_insn. No longer static.
16920 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16921 delete_vta_debug_insns callers.
16922 * rtl.h (delete_vta_debug_insns): Declare.
16923 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16924 instead of variable_tracking_main.
16926 2018-01-03 Martin Sebor <msebor@redhat.com>
16928 PR tree-optimization/83603
16929 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16930 arguments past the endof the argument list in functions declared
16931 without a prototype.
16932 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16933 Avoid checking when arguments are null.
16935 2018-01-03 Martin Sebor <msebor@redhat.com>
16938 * doc/extend.texi (attribute const): Fix a typo.
16939 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16940 issuing -Wsuggest-attribute for void functions.
16942 2018-01-03 Martin Sebor <msebor@redhat.com>
16944 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16945 offset_int::from instead of wide_int::to_shwi.
16946 (maybe_diag_overlap): Remove assertion.
16947 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16948 * gimple-ssa-sprintf.c (format_directive): Same.
16949 (parse_directive): Same.
16950 (sprintf_dom_walker::compute_format_length): Same.
16951 (try_substitute_return_value): Same.
16953 2018-01-03 Jeff Law <law@redhat.com>
16955 PR middle-end/83654
16956 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16957 non-constant residual for zero at runtime and avoid probing in
16958 that case. Reorganize code for trailing problem to mirror handling
16961 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16963 PR tree-optimization/83501
16964 * tree-ssa-strlen.c (get_string_cst): New.
16965 (handle_char_store): Call get_string_cst.
16967 2018-01-03 Martin Liska <mliska@suse.cz>
16969 PR tree-optimization/83593
16970 * tree-ssa-strlen.c: Include tree-cfg.h.
16971 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16972 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16973 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16975 (strlen_dom_walker::before_dom_children): Call
16976 gimple_purge_dead_eh_edges. Dump tranformation with details
16978 (strlen_dom_walker::before_dom_children): Update call by adding
16979 new argument cleanup_eh.
16980 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16982 2018-01-03 Martin Liska <mliska@suse.cz>
16985 * cif-code.def (VARIADIC_THUNK): New enum value.
16986 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16989 2018-01-03 Jan Beulich <jbeulich@suse.com>
16991 * sse.md (mov<mode>_internal): Tighten condition for when to use
16992 vmovdqu<ssescalarsize> for TI and OI modes.
16994 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16996 Update copyright years.
16998 2018-01-03 Martin Liska <mliska@suse.cz>
17001 * ipa-visibility.c (function_and_variable_visibility): Skip
17002 functions with noipa attribure.
17004 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17006 * gcc.c (process_command): Update copyright notice dates.
17007 * gcov-dump.c (print_version): Ditto.
17008 * gcov.c (print_version): Ditto.
17009 * gcov-tool.c (print_version): Ditto.
17010 * gengtype.c (create_file): Ditto.
17011 * doc/cpp.texi: Bump @copying's copyright year.
17012 * doc/cppinternals.texi: Ditto.
17013 * doc/gcc.texi: Ditto.
17014 * doc/gccint.texi: Ditto.
17015 * doc/gcov.texi: Ditto.
17016 * doc/install.texi: Ditto.
17017 * doc/invoke.texi: Ditto.
17019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17021 * vector-builder.h (vector_builder::m_full_nelts): Change from
17022 unsigned int to poly_uint64.
17023 (vector_builder::full_nelts): Update prototype accordingly.
17024 (vector_builder::new_vector): Likewise.
17025 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
17026 (vector_builder::operator ==): Likewise.
17027 (vector_builder::finalize): Likewise.
17028 * int-vector-builder.h (int_vector_builder::int_vector_builder):
17029 Take the number of elements as a poly_uint64 rather than an
17031 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
17032 from unsigned int to poly_uint64.
17033 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
17034 (vec_perm_indices::new_vector): Likewise.
17035 (vec_perm_indices::length): Likewise.
17036 (vec_perm_indices::nelts_per_input): Likewise.
17037 (vec_perm_indices::input_nelts): Likewise.
17038 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
17039 number of elements per input as a poly_uint64 rather than an
17040 unsigned int. Use the original encoding for variable-length
17041 vectors, rather than clamping each individual element.
17042 For the second and subsequent elements in each pattern,
17043 clamp the step and base before clamping their sum.
17044 (vec_perm_indices::series_p): Handle polynomial element counts.
17045 (vec_perm_indices::all_in_range_p): Likewise.
17046 (vec_perm_indices_to_tree): Likewise.
17047 (vec_perm_indices_to_rtx): Likewise.
17048 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
17049 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
17050 (tree_vector_builder::new_binary_operation): Handle polynomial
17051 element counts. Return false if we need to know the number
17052 of elements at compile time.
17053 * fold-const.c (fold_vec_perm): Punt if the number of elements
17054 isn't known at compile time.
17056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17058 * vec-perm-indices.h (vec_perm_builder): Change element type
17059 from HOST_WIDE_INT to poly_int64.
17060 (vec_perm_indices::element_type): Update accordingly.
17061 (vec_perm_indices::clamp): Handle polynomial element_types.
17062 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17063 (vec_perm_indices::all_in_range_p): Likewise.
17064 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
17066 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
17067 polynomial vec_perm_indices element types.
17068 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
17069 * fold-const.c (fold_vec_perm): Likewise.
17070 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
17071 * tree-vect-generic.c (lower_vec_perm): Likewise.
17072 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17073 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
17074 element type to HOST_WIDE_INT.
17076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17077 Alan Hayward <alan.hayward@arm.com>
17078 David Sherwood <david.sherwood@arm.com>
17080 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
17081 rather than an int. Use plus_constant.
17082 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
17083 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
17085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17086 Alan Hayward <alan.hayward@arm.com>
17087 David Sherwood <david.sherwood@arm.com>
17089 * calls.c (emit_call_1, expand_call): Change struct_value_size from
17090 a HOST_WIDE_INT to a poly_int64.
17092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17093 Alan Hayward <alan.hayward@arm.com>
17094 David Sherwood <david.sherwood@arm.com>
17096 * calls.c (load_register_parameters): Cope with polynomial
17097 mode sizes. Require a constant size for BLKmode parameters
17098 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
17099 forces a parameter to be padded at the lsb end in order to
17100 fill a complete number of words, require the parameter size
17101 to be ordered wrt UNITS_PER_WORD.
17103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17104 Alan Hayward <alan.hayward@arm.com>
17105 David Sherwood <david.sherwood@arm.com>
17107 * reload1.c (spill_stack_slot_width): Change element type
17108 from unsigned int to poly_uint64_pod.
17109 (alter_reg): Treat mode sizes as polynomial.
17111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17112 Alan Hayward <alan.hayward@arm.com>
17113 David Sherwood <david.sherwood@arm.com>
17115 * reload.c (complex_word_subreg_p): New function.
17116 (reload_inner_reg_of_subreg, push_reload): Use it.
17118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17119 Alan Hayward <alan.hayward@arm.com>
17120 David Sherwood <david.sherwood@arm.com>
17122 * lra-constraints.c (process_alt_operands): Reject matched
17123 operands whose sizes aren't ordered.
17124 (match_reload): Refer to this check here.
17126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17127 Alan Hayward <alan.hayward@arm.com>
17128 David Sherwood <david.sherwood@arm.com>
17130 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
17131 that the mode size is in the set {1, 2, 4, 8, 16}.
17133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17134 Alan Hayward <alan.hayward@arm.com>
17135 David Sherwood <david.sherwood@arm.com>
17137 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
17138 Use plus_constant instead of gen_rtx_PLUS.
17140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17141 Alan Hayward <alan.hayward@arm.com>
17142 David Sherwood <david.sherwood@arm.com>
17144 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
17145 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
17146 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
17147 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
17148 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
17149 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
17150 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
17151 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
17152 * config/i386/i386.c (ix86_push_rounding): ...this new function.
17153 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
17155 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
17156 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
17157 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
17158 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
17159 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
17160 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
17161 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
17162 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
17163 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
17164 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
17166 * expr.c (emit_move_resolve_push): Treat the input and result
17167 of PUSH_ROUNDING as a poly_int64.
17168 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
17169 (emit_push_insn): Likewise.
17170 * lra-eliminations.c (mark_not_eliminable): Likewise.
17171 * recog.c (push_operand): Likewise.
17172 * reload1.c (elimination_effects): Likewise.
17173 * rtlanal.c (nonzero_bits1): Likewise.
17174 * calls.c (store_one_arg): Likewise. Require the padding to be
17175 known at compile time.
17177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17178 Alan Hayward <alan.hayward@arm.com>
17179 David Sherwood <david.sherwood@arm.com>
17181 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
17182 Use plus_constant instead of gen_rtx_PLUS.
17184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17185 Alan Hayward <alan.hayward@arm.com>
17186 David Sherwood <david.sherwood@arm.com>
17188 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
17189 rather than an int.
17191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17192 Alan Hayward <alan.hayward@arm.com>
17193 David Sherwood <david.sherwood@arm.com>
17195 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
17196 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
17197 via stack temporaries. Treat the mode size as polynomial too.
17199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17200 Alan Hayward <alan.hayward@arm.com>
17201 David Sherwood <david.sherwood@arm.com>
17203 * expr.c (expand_expr_real_2): When handling conversions involving
17204 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
17205 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
17206 as a poly_uint64 too.
17208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17209 Alan Hayward <alan.hayward@arm.com>
17210 David Sherwood <david.sherwood@arm.com>
17212 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
17214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17215 Alan Hayward <alan.hayward@arm.com>
17216 David Sherwood <david.sherwood@arm.com>
17218 * combine.c (can_change_dest_mode): Handle polynomial
17219 REGMODE_NATURAL_SIZE.
17220 * expmed.c (store_bit_field_1): Likewise.
17221 * expr.c (store_constructor): Likewise.
17222 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
17223 and polynomial REGMODE_NATURAL_SIZE.
17224 (gen_lowpart_common): Likewise.
17225 * reginfo.c (record_subregs_of_mode): Likewise.
17226 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
17233 numbers of elements.
17235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17236 Alan Hayward <alan.hayward@arm.com>
17237 David Sherwood <david.sherwood@arm.com>
17239 * match.pd: Cope with polynomial numbers of vector elements.
17241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17242 Alan Hayward <alan.hayward@arm.com>
17243 David Sherwood <david.sherwood@arm.com>
17245 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
17246 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
17253 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
17254 (ipa_simd_modify_function_body): Likewise.
17256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17257 Alan Hayward <alan.hayward@arm.com>
17258 David Sherwood <david.sherwood@arm.com>
17260 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
17261 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
17262 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
17263 (expand_vector_condition, vector_element): Likewise.
17264 (subparts_gt): New function.
17265 (get_compute_type): Use subparts_gt.
17266 (count_type_subparts): Delete.
17267 (expand_vector_operations_1): Use subparts_gt instead of
17268 count_type_subparts.
17270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17271 Alan Hayward <alan.hayward@arm.com>
17272 David Sherwood <david.sherwood@arm.com>
17274 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
17275 (vect_compile_time_alias): ...this new function. Do the calculation
17276 on poly_ints rather than trees.
17277 (vect_prune_runtime_alias_test_list): Update call accordingly.
17279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17280 Alan Hayward <alan.hayward@arm.com>
17281 David Sherwood <david.sherwood@arm.com>
17283 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
17285 (vect_schedule_slp_instance): Likewise.
17287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17288 Alan Hayward <alan.hayward@arm.com>
17289 David Sherwood <david.sherwood@arm.com>
17291 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
17292 constant and extern definitions for variable-length vectors.
17293 (vect_get_constant_vectors): Note that the number of units
17294 is known to be constant.
17296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17297 Alan Hayward <alan.hayward@arm.com>
17298 David Sherwood <david.sherwood@arm.com>
17300 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
17301 of units as polynomial. Choose between WIDE and NARROW based
17304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17305 Alan Hayward <alan.hayward@arm.com>
17306 David Sherwood <david.sherwood@arm.com>
17308 * tree-vect-stmts.c (simd_clone_subparts): New function.
17309 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
17311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17312 Alan Hayward <alan.hayward@arm.com>
17313 David Sherwood <david.sherwood@arm.com>
17315 * tree-vect-stmts.c (vectorizable_call): Treat the number of
17316 vectors as polynomial. Use build_index_vector for
17317 IFN_GOMP_SIMD_LANE.
17319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17320 Alan Hayward <alan.hayward@arm.com>
17321 David Sherwood <david.sherwood@arm.com>
17323 * tree-vect-stmts.c (get_load_store_type): Treat the number of
17324 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
17325 for variable-length vectors.
17326 (vectorizable_mask_load_store): Treat the number of units as
17327 polynomial, asserting that it is constant if the condition has
17328 already been enforced.
17329 (vectorizable_store, vectorizable_load): Likewise.
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-loop.c (vectorizable_live_operation): Treat the number
17336 of units as polynomial. Punt if we can't tell at compile time
17337 which vector contains the final result.
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-loop.c (vectorizable_induction): Treat the number
17344 of units as polynomial. Punt on SLP inductions. Use an integer
17345 VEC_SERIES_EXPR for variable-length integer reductions. Use a
17346 cast of such a series for variable-length floating-point
17349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17350 Alan Hayward <alan.hayward@arm.com>
17351 David Sherwood <david.sherwood@arm.com>
17353 * tree.h (build_index_vector): Declare.
17354 * tree.c (build_index_vector): New function.
17355 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
17356 of units as polynomial, forcibly converting it to a constant if
17357 vectorizable_reduction has already enforced the condition.
17358 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
17359 to create a {1,2,3,...} vector.
17360 (vectorizable_reduction): Treat the number of units as polynomial.
17361 Choose vectype_in based on the largest scalar element size rather
17362 than the smallest number of units. Enforce the restrictions
17365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17366 Alan Hayward <alan.hayward@arm.com>
17367 David Sherwood <david.sherwood@arm.com>
17369 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
17370 number of units as polynomial.
17372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17373 Alan Hayward <alan.hayward@arm.com>
17374 David Sherwood <david.sherwood@arm.com>
17376 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
17377 * target.def (autovectorize_vector_sizes): Return the vector sizes
17378 by pointer, using vector_sizes rather than a bitmask.
17379 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
17380 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
17381 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
17383 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
17384 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
17385 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
17386 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
17387 * omp-general.c (omp_max_vf): Likewise.
17388 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17389 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
17390 * tree-vect-loop.c (vect_analyze_loop): Likewise.
17391 * tree-vect-slp.c (vect_slp_bb): Likewise.
17392 * doc/tm.texi: Regenerate.
17393 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
17395 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
17396 the vector size as a poly_uint64 rather than an unsigned int.
17397 (current_vector_size): Change from an unsigned int to a poly_uint64.
17398 (get_vectype_for_scalar_type): Update accordingly.
17399 * tree.h (build_truth_vector_type): Take the size and number of
17400 units as a poly_uint64 rather than an unsigned int.
17401 (build_vector_type): Add a temporary overload that takes
17402 the number of units as a poly_uint64 rather than an unsigned int.
17403 * tree.c (make_vector_type): Likewise.
17404 (build_truth_vector_type): Take the number of units as a poly_uint64
17405 rather than an unsigned int.
17407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17408 Alan Hayward <alan.hayward@arm.com>
17409 David Sherwood <david.sherwood@arm.com>
17411 * target.def (get_mask_mode): Take the number of units and length
17412 as poly_uint64s rather than unsigned ints.
17413 * targhooks.h (default_get_mask_mode): Update accordingly.
17414 * targhooks.c (default_get_mask_mode): Likewise.
17415 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17416 * doc/tm.texi: Regenerate.
17418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17419 Alan Hayward <alan.hayward@arm.com>
17420 David Sherwood <david.sherwood@arm.com>
17422 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
17423 * omp-general.c (omp_max_vf): Likewise.
17424 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
17425 (expand_omp_simd): Handle polynomial safelen.
17426 * omp-low.c (omplow_simd_context): Add a default constructor.
17427 (omplow_simd_context::max_vf): Change from int to poly_uint64.
17428 (lower_rec_simd_input_clauses): Update accordingly.
17429 (lower_rec_input_clauses): Likewise.
17431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17432 Alan Hayward <alan.hayward@arm.com>
17433 David Sherwood <david.sherwood@arm.com>
17435 * tree-vectorizer.h (vect_nunits_for_cost): New function.
17436 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
17437 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
17438 (vect_analyze_slp_cost): Likewise.
17439 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
17440 (vect_model_load_cost): Likewise.
17442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17443 Alan Hayward <alan.hayward@arm.com>
17444 David Sherwood <david.sherwood@arm.com>
17446 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
17447 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
17448 from an unsigned int * to a poly_uint64_pod *.
17449 (calculate_unrolling_factor): New function.
17450 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
17452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17453 Alan Hayward <alan.hayward@arm.com>
17454 David Sherwood <david.sherwood@arm.com>
17456 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
17457 from an unsigned int to a poly_uint64.
17458 (_loop_vec_info::slp_unrolling_factor): Likewise.
17459 (_loop_vec_info::vectorization_factor): Change from an int
17461 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
17462 (vect_get_num_vectors): New function.
17463 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
17464 (vect_get_num_copies): Use vect_get_num_vectors.
17465 (vect_analyze_data_ref_dependences): Change max_vf from an int *
17466 to an unsigned int *.
17467 (vect_analyze_data_refs): Change min_vf from an int * to a
17469 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17470 than an unsigned HOST_WIDE_INT.
17471 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
17472 (vect_analyze_data_ref_dependence): Change max_vf from an int *
17473 to an unsigned int *.
17474 (vect_analyze_data_ref_dependences): Likewise.
17475 (vect_compute_data_ref_alignment): Handle polynomial vf.
17476 (vect_enhance_data_refs_alignment): Likewise.
17477 (vect_prune_runtime_alias_test_list): Likewise.
17478 (vect_shift_permute_load_chain): Likewise.
17479 (vect_supportable_dr_alignment): Likewise.
17480 (dependence_distance_ge_vf): Take the vectorization factor as a
17481 poly_uint64 rather than an unsigned HOST_WIDE_INT.
17482 (vect_analyze_data_refs): Change min_vf from an int * to a
17484 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
17485 vfm1 as a poly_uint64 rather than an int. Make the same change
17486 for the returned bound_scalar.
17487 (vect_gen_vector_loop_niters): Handle polynomial vf.
17488 (vect_do_peeling): Likewise. Update call to
17489 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
17490 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
17492 * tree-vect-loop.c (vect_determine_vectorization_factor)
17493 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
17494 (vect_get_known_peeling_cost): Likewise.
17495 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
17496 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
17497 (vect_transform_loop): Likewise. Use the lowest possible VF when
17498 updating the upper bounds of the loop.
17499 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17500 rather than an int.
17501 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17502 polynomial unroll factors.
17503 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17504 (vect_make_slp_decision): Likewise.
17505 (vect_supported_load_permutation_p): Likewise, and polynomial
17507 (vect_analyze_slp_cost): Handle polynomial vf.
17508 (vect_slp_analyze_node_operations): Likewise.
17509 (vect_slp_analyze_bb_1): Likewise.
17510 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17511 than an unsigned HOST_WIDE_INT.
17512 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17513 (vectorizable_load): Handle polynomial vf.
17514 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
17516 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
17518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17519 Alan Hayward <alan.hayward@arm.com>
17520 David Sherwood <david.sherwood@arm.com>
17522 * match.pd: Handle bit operations involving three constants
17523 and try to fold one pair.
17525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17527 * tree-vect-loop-manip.c: Include gimple-fold.h.
17528 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17529 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17530 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17531 Add a path that uses a step of VF instead of 1, but disable it
17533 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17534 and niters_no_overflow parameters. Update calls to
17535 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17536 Create a new SSA name if the latter choses to use a ste other
17537 than zero, and return it via niters_vector_mult_vf_var.
17538 * tree-vect-loop.c (vect_transform_loop): Update calls to
17539 vect_do_peeling, vect_gen_vector_loop_niters and
17540 slpeel_make_loop_iterate_ntimes.
17541 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17542 (vect_gen_vector_loop_niters): Update declarations after above changes.
17544 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17546 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17547 128-bit round to integer instructions.
17548 (ceil<mode>2): Likewise.
17549 (btrunc<mode>2): Likewise.
17550 (round<mode>2): Likewise.
17552 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17554 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17555 unaligned VSX load/store on P8/P9.
17556 (expand_block_clear): Allow the use of unaligned VSX
17557 load/store on P8/P9.
17559 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17561 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17563 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17564 swap associated with both a load and a store.
17566 2018-01-02 Andrew Waterman <andrew@sifive.com>
17568 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17569 * config/riscv/riscv.md (clear_cache): Use it.
17571 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17573 * web.c: Remove out-of-date comment.
17575 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17577 * expr.c (fixup_args_size_notes): Check that any existing
17578 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17579 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17580 (emit_single_push_insn): ...here.
17582 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17584 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17585 (const_vector_encoded_nelts): New function.
17586 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17587 (const_vector_int_elt, const_vector_elt): Declare.
17588 * emit-rtl.c (const_vector_int_elt_1): New function.
17589 (const_vector_elt): Likewise.
17590 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17591 of CONST_VECTOR_ELT.
17593 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17595 * expr.c: Include rtx-vector-builder.h.
17596 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17597 directly on the tree encoding.
17598 (const_vector_from_tree): Likewise.
17599 * optabs.c: Include rtx-vector-builder.h.
17600 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17601 sequence of "u" values.
17602 * vec-perm-indices.c: Include rtx-vector-builder.h.
17603 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17604 directly on the vec_perm_indices encoding.
17606 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17608 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17609 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17610 * rtx-vector-builder.h: New file.
17611 * rtx-vector-builder.c: Likewise.
17612 * rtl.h (rtx_def::u2): Add a const_vector field.
17613 (CONST_VECTOR_NPATTERNS): New macro.
17614 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17615 (CONST_VECTOR_DUPLICATE_P): Likewise.
17616 (CONST_VECTOR_STEPPED_P): Likewise.
17617 (CONST_VECTOR_ENCODED_ELT): Likewise.
17618 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17619 (unwrap_const_vec_duplicate): Likewise.
17620 (const_vec_series_p): Check for a non-duplicated vector encoding.
17621 Say that the function only returns true for integer vectors.
17622 * emit-rtl.c: Include rtx-vector-builder.h.
17623 (gen_const_vec_duplicate_1): Delete.
17624 (gen_const_vector): Call gen_const_vec_duplicate instead of
17625 gen_const_vec_duplicate_1.
17626 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17627 (gen_const_vec_duplicate): Use rtx_vector_builder.
17628 (gen_const_vec_series): Likewise.
17629 (gen_rtx_CONST_VECTOR): Likewise.
17630 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17631 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17632 Build a new vector rather than modifying a CONST_VECTOR in-place.
17633 (handle_special_swappables): Update call accordingly.
17634 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17635 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17636 Build a new vector rather than modifying a CONST_VECTOR in-place.
17637 (handle_special_swappables): Update call accordingly.
17639 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17641 * simplify-rtx.c (simplify_const_binary_operation): Use
17642 CONST_VECTOR_ELT instead of XVECEXP.
17644 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17646 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17647 the selector elements to be different from the data elements
17648 if the selector is a VECTOR_CST.
17649 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17650 ssizetype for the selector.
17652 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17654 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17655 before testing each element individually.
17656 * tree-vect-generic.c (lower_vec_perm): Likewise.
17658 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17660 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17661 * selftest-run-tests.c (selftest::run_tests): Call it.
17662 * vector-builder.h (vector_builder::operator ==): New function.
17663 (vector_builder::operator !=): Likewise.
17664 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17665 (vec_perm_indices::all_from_input_p): New function.
17666 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17667 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17668 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17669 instead of reading the VECTOR_CST directly. Detect whether both
17670 vector inputs are the same before constructing the vec_perm_indices,
17671 and update the number of inputs argument accordingly. Use the
17672 utility functions added above. Only construct sel2 if we need to.
17674 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17676 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17677 the broadcast of the low byte.
17678 (expand_mult_highpart): Use an explicit encoding for the permutes.
17679 * optabs-query.c (can_mult_highpart_p): Likewise.
17680 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17681 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17682 (vectorizable_bswap): Likewise.
17683 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17684 explicit encoding for the power-of-2 permutes.
17685 (vect_permute_store_chain): Likewise.
17686 (vect_grouped_load_supported): Likewise.
17687 (vect_permute_load_chain): Likewise.
17689 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17691 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17692 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17693 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17694 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17695 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17696 (vect_gen_perm_mask_any): Likewise.
17698 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17700 * int-vector-builder.h: New file.
17701 * vec-perm-indices.h: Include int-vector-builder.h.
17702 (vec_perm_indices): Redefine as an int_vector_builder.
17703 (auto_vec_perm_indices): Delete.
17704 (vec_perm_builder): Redefine as a stand-alone class.
17705 (vec_perm_indices::vec_perm_indices): New function.
17706 (vec_perm_indices::clamp): Likewise.
17707 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17708 (vec_perm_indices::new_vector): New function.
17709 (vec_perm_indices::new_expanded_vector): Update for new
17710 vec_perm_indices class.
17711 (vec_perm_indices::rotate_inputs): New function.
17712 (vec_perm_indices::all_in_range_p): Operate directly on the
17713 encoded form, without computing elided elements.
17714 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17715 encoding. Update for new vec_perm_indices class.
17716 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17717 the given vec_perm_builder.
17718 (expand_vec_perm_var): Update vec_perm_builder constructor.
17719 (expand_mult_highpart): Use vec_perm_builder instead of
17720 auto_vec_perm_indices.
17721 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17722 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17723 or double series encoding as appropriate.
17724 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17725 vec_perm_indices instead of auto_vec_perm_indices.
17726 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17727 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17728 (vect_permute_store_chain): Likewise.
17729 (vect_grouped_load_supported): Likewise.
17730 (vect_permute_load_chain): Likewise.
17731 (vect_shift_permute_load_chain): Likewise.
17732 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17733 (vect_transform_slp_perm_load): Likewise.
17734 (vect_schedule_slp_instance): Likewise.
17735 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17736 (vectorizable_mask_load_store): Likewise.
17737 (vectorizable_bswap): Likewise.
17738 (vectorizable_store): Likewise.
17739 (vectorizable_load): Likewise.
17740 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17741 vec_perm_indices instead of auto_vec_perm_indices. Use
17742 tree_to_vec_perm_builder to read the vector from a tree.
17743 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17744 vec_perm_builder instead of a vec_perm_indices.
17745 (have_whole_vector_shift): Use vec_perm_builder and
17746 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17747 truncation to calc_vec_perm_mask_for_shift.
17748 (vect_create_epilog_for_reduction): Likewise.
17749 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17750 from auto_vec_perm_indices to vec_perm_indices.
17751 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17752 instead of changing individual elements.
17753 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17754 the vector in d.perm.
17755 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17756 from auto_vec_perm_indices to vec_perm_indices.
17757 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17758 instead of changing individual elements.
17759 (arm_vectorize_vec_perm_const): Use new_vector to install
17760 the vector in d.perm.
17761 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17762 Update vec_perm_builder constructor.
17763 (rs6000_expand_interleave): Likewise.
17764 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17765 (rs6000_expand_interleave): Likewise.
17767 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17769 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17770 to qimode could truncate the indices.
17771 * optabs.c (expand_vec_perm_var): Likewise.
17773 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17775 * Makefile.in (OBJS): Add vec-perm-indices.o.
17776 * vec-perm-indices.h: New file.
17777 * vec-perm-indices.c: Likewise.
17778 * target.h (vec_perm_indices): Replace with a forward class
17780 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17781 * optabs.h: Include vec-perm-indices.h.
17782 (expand_vec_perm): Delete.
17783 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17784 (expand_vec_perm_const): Declare.
17785 * target.def (vec_perm_const_ok): Replace with...
17786 (vec_perm_const): ...this new hook.
17787 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17788 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17789 * doc/tm.texi: Regenerate.
17790 * optabs.def (vec_perm_const): Delete.
17791 * doc/md.texi (vec_perm_const): Likewise.
17792 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17793 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17794 expand_vec_perm for constant permutation vectors. Assert that
17795 the mode of variable permutation vectors is the integer equivalent
17796 of the mode that is being permuted.
17797 * optabs-query.h (selector_fits_mode_p): Declare.
17798 * optabs-query.c: Include vec-perm-indices.h.
17799 (selector_fits_mode_p): New function.
17800 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17801 is defined, instead of checking whether the vec_perm_const_optab
17802 exists. Use targetm.vectorize.vec_perm_const instead of
17803 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17804 fit in the vector mode before using a variable permute.
17805 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17806 vec_perm_indices instead of an rtx.
17807 (expand_vec_perm): Replace with...
17808 (expand_vec_perm_const): ...this new function. Take the selector
17809 as a vec_perm_indices rather than an rtx. Also take the mode of
17810 the selector. Update call to shift_amt_for_vec_perm_mask.
17811 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17812 Use vec_perm_indices::new_expanded_vector to expand the original
17813 selector into bytes. Check whether the indices fit in the vector
17814 mode before using a variable permute.
17815 (expand_vec_perm_var): Make global.
17816 (expand_mult_highpart): Use expand_vec_perm_const.
17817 * fold-const.c: Includes vec-perm-indices.h.
17818 * tree-ssa-forwprop.c: Likewise.
17819 * tree-vect-data-refs.c: Likewise.
17820 * tree-vect-generic.c: Likewise.
17821 * tree-vect-loop.c: Likewise.
17822 * tree-vect-slp.c: Likewise.
17823 * tree-vect-stmts.c: Likewise.
17824 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17826 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17827 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17828 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17829 (aarch64_vectorize_vec_perm_const): ...this new function.
17830 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17831 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17832 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17833 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17834 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17835 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17836 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17838 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17839 check for NEON modes.
17840 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17841 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17842 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17843 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17845 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17846 the old VEC_PERM_CONST conditions.
17847 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17848 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17849 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17850 (ia64_vectorize_vec_perm_const_ok): Merge into...
17851 (ia64_vectorize_vec_perm_const): ...this new function.
17852 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17853 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17854 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17855 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17856 * config/mips/mips.c (mips_expand_vec_perm_const)
17857 (mips_vectorize_vec_perm_const_ok): Merge into...
17858 (mips_vectorize_vec_perm_const): ...this new function.
17859 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17860 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17861 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17862 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17863 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17864 (rs6000_expand_vec_perm_const): Delete.
17865 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17867 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17868 (altivec_expand_vec_perm_const_le): Take each operand individually.
17869 Operate on constant selectors rather than rtxes.
17870 (altivec_expand_vec_perm_const): Likewise. Update call to
17871 altivec_expand_vec_perm_const_le.
17872 (rs6000_expand_vec_perm_const): Delete.
17873 (rs6000_vectorize_vec_perm_const_ok): Delete.
17874 (rs6000_vectorize_vec_perm_const): New function.
17875 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17876 an element count and rtx array.
17877 (rs6000_expand_extract_even): Update call accordingly.
17878 (rs6000_expand_interleave): Likewise.
17879 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17880 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17881 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17882 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17883 (rs6000_expand_vec_perm_const): Delete.
17884 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17885 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17886 (altivec_expand_vec_perm_const_le): Take each operand individually.
17887 Operate on constant selectors rather than rtxes.
17888 (altivec_expand_vec_perm_const): Likewise. Update call to
17889 altivec_expand_vec_perm_const_le.
17890 (rs6000_expand_vec_perm_const): Delete.
17891 (rs6000_vectorize_vec_perm_const_ok): Delete.
17892 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17893 reference to the SPE evmerge intructions.
17894 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17895 an element count and rtx array.
17896 (rs6000_expand_extract_even): Update call accordingly.
17897 (rs6000_expand_interleave): Likewise.
17898 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17899 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17901 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17903 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17905 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17906 vector mode and that that mode matches the mode of the data
17908 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17909 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17910 directly using expand_vec_perm_1 when forcing selectors into
17912 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17914 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17916 * optabs-query.h (can_vec_perm_p): Delete.
17917 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17918 * optabs-query.c (can_vec_perm_p): Split into...
17919 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17920 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17921 particular selector is valid.
17922 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17923 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17924 (vect_grouped_load_supported): Likewise.
17925 (vect_shift_permute_load_chain): Likewise.
17926 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17927 (vect_transform_slp_perm_load): Likewise.
17928 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17929 (vectorizable_bswap): Likewise.
17930 (vect_gen_perm_mask_checked): Likewise.
17931 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17932 implementations of variable permutation vectors into account
17933 when deciding which selector to use.
17934 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17935 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17936 with a false third argument.
17937 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17938 to test whether the constant selector is valid and can_vec_perm_var_p
17939 to test whether a variable selector is valid.
17941 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17943 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17944 * optabs-query.c (can_vec_perm_p): Likewise.
17945 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17946 instead of vec_perm_indices.
17947 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17948 (vect_gen_perm_mask_checked): Likewise,
17949 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17950 (vect_gen_perm_mask_checked): Likewise,
17952 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17954 * optabs-query.h (qimode_for_vec_perm): Declare.
17955 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17956 (qimode_for_vec_perm): ...this new function.
17957 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17959 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17961 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17962 does not have a conditional at the top.
17964 2018-01-02 Richard Biener <rguenther@suse.de>
17966 * ipa-inline.c (big_speedup_p): Fix expression.
17968 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17971 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17974 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17978 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17979 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17980 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17981 cond_taken_branch_cost 3->4.
17983 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17985 PR tree-optimization/83581
17986 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17987 TODO_cleanup_cfg if any changes have been made.
17989 PR middle-end/83608
17990 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17991 convert_modes if target mode has the right side, but different mode
17994 PR middle-end/83609
17995 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17996 last argument when extracting from CONCAT. If either from_real or
17997 from_imag is NULL, use expansion through memory. If result is not
17998 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17999 the parts directly to inner mode, if even that fails, use expansion
18002 PR middle-end/83623
18003 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
18004 check for bswap in mode rather than HImode and use that in expand_unop
18007 Copyright (C) 2018 Free Software Foundation, Inc.
18009 Copying and distribution of this file, with or without modification,
18010 are permitted in any medium without royalty provided the copyright
18011 notice and this notice are preserved.