1 2018-05-04 Tom de Vries <tom@codesourcery.com>
4 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7 2018-05-04 Richard Biener <rguenther@suse.de>
10 * tree-complex.c (update_complex_assignment): We are always in SSA form.
11 (expand_complex_div_wide): Likewise.
12 (expand_complex_operations_1): Likewise.
13 (expand_complex_libcall): Preserve EH info of the original stmt.
14 (tree_lower_complex): Handle removed blocks.
15 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
16 on complex multiplication and division libcall builtins.
18 2018-05-04 Richard Biener <rguenther@suse.de>
21 * fold-const.c (negate_expr_p): Restrict negation of operand
22 zero of a division to when we know that can happen without
24 (fold_negate_expr_1): Likewise.
26 2018-05-04 Jakub Jelinek <jakub@redhat.com>
29 * real.h (real_nextafter): Declare.
30 * real.c (real_nextafter): New function.
31 * fold-const-call.c (fold_const_nextafter): New function.
32 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
34 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
35 even when arg1_mode is different from arg0_mode.
37 2018-05-03 Nathan Sidwell <nathan@acm.org>
39 * doc/extend.texi (Deprecated Features): Remove
41 (Backwards Compatibility): Likewise.
42 * doc/invoke.texi (C++ Language Options): Likewise.
43 (C++ Dialect Options): Likewise.
45 2018-05-03 Jakub Jelinek <jakub@redhat.com>
48 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
49 _mm512_mask_mullox_epi64): New intrinsics.
51 2018-05-03 Tom de Vries <tom@codesourcery.com>
54 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
55 dump files): Add offload-tree.
57 2018-05-03 Richard Biener <rguenther@suse.de>
59 PR tree-optimization/85615
60 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
61 to loops not nested in BBs loop father to avoid creating multi-entry
64 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
66 PR tree-optimization/70291
67 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
68 arguments. Change return type to tree. Emit libcall as a new
69 statement rather than replacing existing one when inplace_p is true.
70 (expand_complex_multiplication_components): New function.
71 (expand_complex_multiplication): Expand floating-point complex
72 multiplication using the above.
73 (expand_complex_division): Rename inner_type parameter to type.
74 Update expand_complex_libcall call-site.
75 (expand_complex_operations_1): Update expand_complex_multiplication
76 and expand_complex_division call-sites.
78 2018-05-02 Jakub Jelinek <jakub@redhat.com>
81 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
82 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
83 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
84 the highest significant bit of the shift count mask is clear. In
85 check whether and[sq]i3 is needed verify that all significant bits
86 of the shift count other than the highest are set.
88 2018-05-02 Tom de Vries <tom@codesourcery.com>
91 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
92 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
93 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
94 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
95 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
96 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
97 __builtin_goacc_parlevel_size.
99 2018-05-02 Richard Biener <rguenther@suse.de>
101 PR tree-optimization/85597
102 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
103 do not use split vect_get_vec_defs call but call vect_get_slp_defs
106 2018-05-02 Tom de Vries <tom@codesourcery.com>
109 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
110 dump files): Add ltrans-tree.
112 2018-05-02 Tom de Vries <tom@codesourcery.com>
115 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
116 dump files): Add wpa-ipa.
118 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
120 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
121 powerpc*-*-linux*paired* target.
122 * config/rs6000/750cl.h: Delete.
123 * config/rs6000/paired.h: Delete.
124 * config/rs6000/paired.md: Delete.
125 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
127 * config/rs6000/rs6000-builtin.def: Remove paired float support.
128 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
129 comment. Remove paired float support.
130 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
131 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
133 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
134 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
136 * config/rs6000/rs6000.c: Remove paired float support.
137 (paired_expand_vector_init, paired_expand_vector_move,
138 paired_emit_vector_compare, paired_emit_vector_cond_expr,
139 (paired_expand_lv_builtin, paired_expand_stv_builtin,
140 paired_expand_builtin, paired_expand_predicate_builtin,
141 paired_init_builtins): Delete.
142 * config/rs6000/rs6000.h: Remove paired float support.
143 * config/rs6000/rs6000.md: Remove paired float support.
144 (move_from_CR_ov_bit): Delete.
145 * config/rs6000/rs6000.opt (mpaired): Delete.
146 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
147 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
149 2018-05-02 Richard Biener <rguenther@suse.de>
152 * gimplify.c (gimplify_save_expr): When in SSA form allow
153 SAVE_EXPRs to compute to SSA vars.
155 2018-05-02 Jakub Jelinek <jakub@redhat.com>
158 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
159 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
160 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
161 clobber operands[2], instead use a new pseudo. Formatting fixes.
163 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
165 PR tree-optimization/85586
166 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
167 exit early for statements in the same group if the accesses are
170 2018-05-02 Tom de Vries <tom@codesourcery.com>
173 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
176 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
178 PR tree-optimization/85143
179 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
181 2018-05-01 Tom de Vries <tom@codesourcery.com>
184 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
185 not found" error message.
187 2018-05-01 Tom de Vries <tom@codesourcery.com>
190 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
191 * vec.c (test_ordered_remove_if): New function.
192 (vec_c_tests): Call test_ordered_remove_if.
193 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
194 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
195 * tree-vect-patterns.c (vect_pattern_recog_1): Use
196 VEC_ORDERED_REMOVE_IF.
198 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
200 PR tree-optimization/82665
201 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
202 pointer subtraction where arguments come from a memchr call.
204 2018-05-01 Jakub Jelinek <jakub@redhat.com>
206 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
207 --push-state --as-needed and --pop-state instead of --as-needed and
208 --no-as-needed if ld supports it.
209 * configure: Regenerated.
212 * doc/install.texi2html: Replace _002d with - and _002a with * in
213 generated html files using sed.
215 2018-04-30 David Malcolm <dmalcolm@redhat.com>
218 * gcc-rich-location.c (blank_line_before_p): New function.
219 (use_new_line): New function.
220 (gcc_rich_location::add_fixit_insert_formatted): New function.
221 * gcc-rich-location.h
222 (gcc_rich_location::add_fixit_insert_formatted): New function.
224 2018-04-30 David Malcolm <dmalcolm@redhat.com>
226 * selftest.c (assert_streq): Rename "expected" and "actual" to
227 "val1" and "val2". Extend NULL-handling to cover both inputs
228 symmetrically, while still requiring both to be non-NULL for a pass.
229 * selftest.h (assert_streq): Rename "expected" and "actual" to
231 (ASSERT_EQ): Likewise.
232 (ASSERT_EQ_AT): Likewise.
233 (ASSERT_KNOWN_EQ): Likewise.
234 (ASSERT_KNOWN_EQ_AT): Likewise.
235 (ASSERT_NE): Likewise.
236 (ASSERT_MAYBE_NE): Likewise.
237 (ASSERT_MAYBE_NE_AT): Likewise.
238 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
239 the assertion to pass.
240 (ASSERT_STREQ_AT): Likewise.
242 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
244 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
245 interaction with -pie.
247 2018-04-30 David Malcolm <dmalcolm@redhat.com>
249 * selftest.h: Fix alphabetization of per-source-file selftest
252 2018-04-30 Jason Merrill <jason@redhat.com>
254 PR c++/61982 - dead stores to destroyed objects.
255 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
258 2018-04-30 Jason Merrill <jason@redhat.com>
260 * tree.c (build_clobber): New.
261 * tree.h: Declare it.
262 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
264 2018-04-30 David Malcolm <dmalcolm@redhat.com>
266 * diagnostic-show-locus.c (layout::layout): Update for
267 location_get_source_line returning a char_span.
268 (struct char_span): Move to input.h.
269 (struct correction): Update for fields in char_span becoming
271 (struct source_line): Update for location_get_source_line
272 returning a char_span.
273 (layout::print_line): Likewise.
274 * edit-context.c (edited_file::print_content): Likewise.
275 (edited_file::print_diff_hunk): Likewise.
276 (edited_file::print_run_of_changed_lines): Likewise.
277 (edited_file::get_num_lines): Likewise.
278 (edited_line::edited_line): Likewise.
279 * final.c (asm_show_source): Likewise.
280 * input.c (location_get_source_line): Convert return type
281 from const char * to char_span, losing the final "line_len"
283 (dump_location_info): Update for the above.
284 (get_substring_ranges_for_loc): Likewise. Use a char_span
285 when handling the literal within the line.
286 (test_reading_source_line): Update for location_get_source_line
287 returning a char_span.
288 * input.h (class char_span): Move here from
289 diagnostic-show-locus.c, converting from a struct to a class.
290 Make data members private.
291 (char_span::operator bool): New.
292 (char_span::length): New.
293 (char_span::get_buffer): New.
294 (char_span::operator[]): New.
295 (char_span::subspan): Make const.
296 (char_span::xstrdup): New.
297 (location_get_source_line): Convert return type from const char *
298 to char_span, losing the final "line_size" param.
300 2018-04-30 Jan Hubicka <jh@suse.cz>
302 * lto-wrapper.c (ltrans_priorities): New static var.
304 (run_gcc): Read priorities and if doing parallel build order
305 the Makefile by them.
307 2018-04-30 David Malcolm <dmalcolm@redhat.com>
309 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
311 2018-04-30 Richard Biener <rguenther@suse.de>
313 * tree-cfg.c (verify_address): Remove base argument, add
314 flag whether to check TREE_ADDRESSABLE and do that.
315 (verify_expr): Remove.
316 (verify_types_in_gimple_reference): Add pieces from verify_expr.
317 (verify_gimple_assign_single): Likewise.
318 (verify_gimple_switch): Likewise.
319 (verify_expr_location_1): Dereference tp once. Add (disabled)
320 piece from verify_expr.
321 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
323 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
325 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
327 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
329 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
330 (small_data_pattern): Likewise.
331 (arc_rewrite_small_data): Likewise.
332 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
333 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
334 (get_symbol_alignment): New function.
335 (legitimate_small_data_address_p): Likewise.
336 (legitimate_scaled_address): Update, call
337 legitimate_small_data_address_p.
338 (output_sdata): New static variable.
339 (arc_print_operand): Update how we handle small data operands.
340 (arc_print_operand_address): Likewise.
341 (arc_legitimate_address_p): Update, use
342 legitimate_small_data_address_p.
343 (arc_rewrite_small_data_p): Remove.
344 (arc_rewrite_small_data_1): Likewise.
345 (arc_rewrite_small_data): Likewise.
346 (small_data_pattern): Likewise.
347 (compact_sda_memory_operand): Update to use
348 legitimate_small_data_address_p and get_symbol_alignment.
349 (prepare_move_operands): Don't rewite sdata pattern.
350 (prepare_extend_operands): Remove.
351 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
353 (zero_extendqisi2): Likewise.
354 (zero_extendhisi2): Likewise.
355 (extendqihi2): Likewise.
356 (extendqisi2): Likewise.
357 (extendhisi2): Likewise.
361 * config/arc/constraints.md (Usd): Change it to memory constraint.
363 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
365 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
366 as source of std instructions.
367 * config/arc/arc.md (movsi_insn): Update pattern predicate to
368 allow 6-bit constants as source for store instructions.
369 (movdi_insn): Update instruction pattern to allow 6-bit constants
370 as source for store instructions.
372 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
374 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
376 2018-04-30 Nathan Sidwell <nathan@acm.org>
377 Sandra Loosemore <sandra@codesourcery.com>
379 * dumpfile.c (dump_open): Allow '-' for stdout.
380 * doc/invoke.texi (Developer Options): Document dump filename
381 determination early. Document stdin/stdout selection.
383 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
385 Microblaze Target: PIC data text relative
387 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
388 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
390 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
391 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
392 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
393 New addressing mode for data-text relative position indepenedent code.
394 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
395 'ADDRESS_SYMBOLIC_TXT_REL'.
396 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
397 (microblaze_legitimate_pic_operand): Exclude function calls from
398 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
399 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
401 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
402 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
403 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
404 for 'address + offset'.
405 (microblaze_expand_prologue): Add new function prologue call for
407 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
408 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
409 table in case of TARGET_PIC_DATA_TEXT_REL.
410 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
411 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
412 Add new macros 'UNSPEC_TEXT',
413 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
414 + exclude function calls from 'UNSPEC_PLT' in case of data text
416 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
417 new target hook for generating address diff vector tables in case of
419 * doc/tm.texi : Regenerate.
420 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
421 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
422 of addr diff vector generation.
423 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
424 target hook definition.
425 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
426 Add default function for generate_pic_addr_diff_vec -> flag_pic.
427 * doc/invoke.texi (Add new pic option): Add new microblaze pic
428 option for data text relative.
430 2018-04-30 Richard Biener <rguenther@suse.de>
432 * tree-chrec.h (evolution_function_is_constant_p): Remove
434 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
436 2018-04-30 Richard Biener <rguenther@suse.de>
439 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
441 2018-04-30 Richard Biener <rguenther@suse.de>
443 PR tree-optimization/28364
444 PR tree-optimization/85275
445 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
446 copying first exit test.
448 2018-04-28 Mark Wielaard <mark@klomp.org>
450 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
452 (dwarf_AT): Handle DW_AT_addr_base.
453 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
455 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
458 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
459 (*ashl<dwi>3_doubleword_mask_1): Ditto.
460 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
461 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
463 2018-04-28 Richard Biener <rguenther@suse.de>
465 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
466 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
467 to reflect use. Only add interesting stmts.
469 2018-04-27 Martin Jambor <mjambor@suse.cz>
472 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
473 the jump function allows for passing through aggregate values.
475 2018-04-27 David Malcolm <dmalcolm@redhat.com>
477 * input.h (in_system_header_at): Convert from macro to inline
479 (from_macro_expansion_at): Likewise.
480 (from_macro_definition_at): Likewise.
482 2018-04-27 Jeff Law <law@redhat.com>
484 * config.gcc: Mark tile* targets as deprecated/obsolete.
486 2018-04-27 Richard Biener <rguenther@suse.de>
488 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
491 2018-04-27 Richard Biener <rguenther@suse.de>
493 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
495 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
497 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
498 with Yd constraint. Set "preferred_for_speed" attribute from
499 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
501 (*movdi_internal): Ditto.
502 (movti_interunit splitters): Remove
503 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
504 (movdi_interunit splitters): Ditto.
505 * config/i386/constraints.md (Ye): Remove.
506 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
508 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
511 * config/aarch64/constraints.md (Usg): Limit to 31.
514 2018-04-27 Jakub Jelinek <jakub@redhat.com>
516 PR tree-optimization/85529
517 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
518 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
519 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
520 zero extension or masking of the MSB bit.
521 (optimize_range_tests): Add FIRST_BB argument, pass it through
522 to optimize_range_tests_var_bound.
523 (maybe_optimize_range_tests, reassociate_bb): Adjust
524 optimize_range_tests callers.
526 2018-04-26 Richard Biener <rguenther@suse.de>
527 Jakub Jelinek <jakub@redhat.com>
529 * cgraph.h (symbol_table): Just declare debug method here.
530 * symtab.c (symbol_table::debug): Define.
532 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
534 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
536 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
538 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
539 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
540 (*movdi_internal): Substitute Yi and Yj constraint with x
541 and Ym and Yn constraint with y constraint. Update "isa"
542 attribute and set "preferred_for_speed" attribute from
543 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
544 (*movsi_internal): Ditto.
545 (*movdf_internal): Ditto.
546 (*movsf_internal): Ditto.
547 (*zero_extendsidi2): Ditto.
548 * config/i386/sse.md (vec_set<mode>_0): Ditto.
549 (sse2_loadld): Ditto.
550 (*vec_extract<ssevecmodelower>_0): Ditto.
551 (*vec_extractv4si_0_zext_sse4): Ditto.
552 (vec_concatv2di): Ditto.
553 (*vec_dup<mode>): Ditto.
554 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
555 * config/i386/constraints.md (Yi): Remove.
560 2018-04-26 Nathan Sidwell <nathan@acm.org>
562 * dumpfile.c (dump_open): New.
563 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
564 (dump_finish): Detect stdio/stderr by value not name.
566 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
568 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
570 2018-04-26 Tom de Vries <tom@codesourcery.com>
573 * config/nvptx/nvptx.c (verify_neutering_jumps)
574 (verify_neutering_labels): New function
575 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
577 2018-04-26 Tom de Vries <tom@codesourcery.com>
580 * config/nvptx/nvptx.c (needs_neutering_p): New function.
581 (nvptx_single): Use needs_neutering_p to skip over insns that do not
584 2018-04-26 Richard Biener <rguenther@suse.de>
585 Tom de Vries <tom@codesourcery.com>
588 * lto-streamer-out.c (output_function): Fixup loops if required to match
589 discovery done in the reader.
591 2018-04-26 Richard Biener <rguenther@suse.de>
593 PR tree-optimization/85116
594 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
595 have a loop exit from the single latch predecessor. Remove
596 case of header with just condition.
597 (ch_base::copy_headers): Exclude infinite loops from any
599 (pass_ch::execute): Record exits.
601 2018-04-26 Richard Biener <rguenther@suse.de>
603 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
604 prologue cost vector and pass it to vect_get_load_cost.
605 (vect_get_peeling_costs_all_drs): Likewise.
606 (vect_peeling_hash_get_lowest_cost): Likewise.
607 (vect_enhance_data_refs_alignment): Likewise.
609 2018-04-26 Richard Biener <rguenther@suse.de>
612 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
613 checking of integer<->pointer conversions.
614 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
615 sign-/zero-extending pointer types.
616 (expand_omp_for_static_chunk): Likewise.
618 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
619 Jean Lee <xiaoyur347@gmail.com>
621 * config/mips/mips.c (mips_asan_shadow_offset): New function.
622 (TARGET_ASAN_SHADOW_OFFSET): Define.
623 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
624 true for -fsanitize=address.
626 2018-04-25 Mark Wielaard <mark@klomp.org>
628 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
631 2018-04-25 Jakub Jelinek <jakub@redhat.com>
633 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
634 than "alu", remove explicit "memory" and "imm_disp" attributes.
635 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
638 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
639 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
642 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
645 * config/i386/i386.c (ix86_expand_builtin): Change memory
646 operand to XI, extend p0 to Pmode.
647 * config/i386/i386.md: Change unspec volatile and operand
648 1 mode to XI, change operand 0 mode to P.
650 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
652 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
653 GET_MODE_MASK before any checking.
654 (nds32_can_use_bset_p): Likewise.
655 (nds32_can_use_btgl_p): Likewise.
657 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
659 * config/nds32/nds32-doubleword.md: New define_split pattern for
660 illegal register number.
662 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
664 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
666 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
668 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
670 2018-04-25 Richard Biener <rguenther@suse.de>
672 * lto-streamer.h (LTO_major_version): Bump to 8.
674 2018-04-25 Jakub Jelinek <jakub@redhat.com>
676 * BASE-VER: Set to 9.0.0.
678 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
680 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
681 in __abskf2 and __powikf2.
683 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
686 * config/aarch64/constraints.md (Usg, Usj): New constraints.
687 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
688 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
689 Use the above on operand 2. Reindent.
690 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
692 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
695 * common/config/i386/i386-common.c (ix86_handle_option): Don't
697 * config/i386/i386.opt (mcet): Removed.
698 * doc/install.texi: Remove -mcet documentation.
699 * doc/invoke.texi: Likewise.
701 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
704 * doc/install.texi: Remove -mcet from bootstrap-cet.
706 2018-04-24 Jakub Jelinek <jakub@redhat.com>
709 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
710 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
714 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
715 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
716 containing a CONST_VECTOR.
718 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
720 * doc/install.texi: Update newlib dependency for nvptx.
722 2018-04-24 Jakub Jelinek <jakub@redhat.com>
725 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
726 instead of INTVAL when shifting x left.
728 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
730 PR tree-optimization/85478
731 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
732 vect_grouped_store_supported for single element vectors.
734 2018-04-24 Richard Biener <rguenther@suse.de>
737 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
738 load cost increase to the case of non-constant step.
740 2018-04-24 Jakub Jelinek <jakub@redhat.com>
743 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
744 destination if any_malformed_asm.
746 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
749 * expr.c (store_field): In the bitfield case, if the value comes from
750 a function call and is returned in registers by means of a PARALLEL,
751 do not change the mode of the temporary unless BLKmode and VOIDmode.
753 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
755 PR rtl-optimization/85423
756 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
757 dependencies to debug insns when the previous insn is non-debug.
759 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
761 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
762 enums into a single definition.
763 (fls): Fix predicates and printing.
766 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
768 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
769 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
770 and short u6 immediate.
771 (check_if_valid_sleep_operand): Remove.
772 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
774 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
776 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
777 flag_always_save_lp condition.
778 * config/nds32/nds32.opt (malways-save-lp): New option.
780 2018-04-22 Shiva Chen <shiva0217@gmail.com>
782 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
783 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
784 * config/nds32/nds32.h
785 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
786 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
788 2018-04-22 Shiva Chen <shiva0217@gmail.com>
790 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
791 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
793 2018-04-22 Shiva Chen <shiva0217@gmail.com>
794 Chung-Ju Wu <jasonwucj@gmail.com>
796 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
798 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
799 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
801 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
803 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
805 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
807 * config/nds32/nds32-protos.h (nds32_data_alignment,
808 nds32_local_alignment): Declare.
809 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
810 nds32_local_alignment): New functions.
811 (TARGET_CONSTANT_ALIGNMENT): Define.
812 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
814 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
816 * config/nds32/nds32.c
817 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
818 (TARGET_MODES_TIEABLE_P): Likewise.
820 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
822 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
825 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
826 Chung-Ju Wu <jasonwucj@gmail.com>
828 * config/nds32/constants.md (unspec_volatile_element): Add enum values
829 for unaligned access.
830 * config/nds32/nds32-intrinsic.c: Implementation of expanding
832 * config/nds32/nds32-intrinsic.md: Likewise.
833 * config/nds32/nds32_intrinsic.h: Likewise.
834 * config/nds32/nds32.h (nds32_builtins): Likewise.
835 * config/nds32/nds32.opt (munaligned-access): New option.
836 * config/nds32/nds32.c (nds32_asm_file_start): Display
837 flag_unaligned_access status.
839 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
841 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
842 -mno-relax is present.
843 * config/riscv/linux.h (LINK_SPEC): Ditto.
845 2018-04-20 Martin Sebor <msebor@redhat.com>
848 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
850 (gimple_fold_builtin_stxcpy_chk): Same.
851 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
853 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
856 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
857 __powikf2 when long double is IEEE 128-bit.
859 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
861 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
862 step to make sure stack always aligned.
864 2018-04-20 Carl Love <cel@us.ibm.com>
867 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
870 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
871 Tom de Vries <tom@codesourcery.com>
874 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
875 Emit insns for calls too.
876 (nvptx_find_par): Always look for worker-level predecessor insn.
877 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
879 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
880 (nvptx_process_pars): Propagate frames for calls.
882 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
885 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
887 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
888 (ix86_handle_option): Don't handle OPT_mibt.
889 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
891 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
893 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
894 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
895 (ix86_target_macros): Define __CET__ with flag_cf_protection
897 * config/i386/i386.c (isa2_opts): Remove -mibt.
898 * config/i386/i386.h (TARGET_IBT): Removed.
899 (TARGET_IBT_P): Likewise.
900 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
901 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
902 * config/i386/i386.opt (mcet): Update help message.
905 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
906 -mcet as an alias for -mshstk.
908 2018-04-20 Richard Biener <rguenther@suse.de>
911 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
912 complexity by forcing a single use of the multiply operand.
914 2018-04-20 Martin Jambor <mjambor@suse.cz>
917 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
918 recursion dependency to only apply to non-clones.
920 2018-04-20 Martin Jambor <mjambor@suse.cz>
923 * ipa-cp.c (create_specialized_node): Check that clones of
924 self-recursive edges exist during IPA-CP.
926 2018-04-19 Toon Moene <toon@moene.org>
928 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
931 2018-04-19 Jakub Jelinek <jakub@redhat.com>
933 PR tree-optimization/85467
934 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
935 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
936 VECTOR_CST element to type.
938 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
941 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
942 * config/i386/i386.md (builtin_setjmp_setup): Removed.
943 (builtin_longjmp): Likewise.
944 (save_stack_nonlocal): New pattern.
945 (restore_stack_nonlocal): Likewise.
947 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
950 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
951 Replace ASM_OUTPUT_LABEL with fprintf.
953 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
956 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
957 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
958 * config/i386/i386-c.c (ix86_target_macros_internal): Also
959 define __IBT__ and __SHSTK__ for -fcf-protection.
960 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
962 (ix86_trampoline_init): Likewise.
963 (x86_output_mi_thunk): Likewise.
964 (ix86_notrack_prefixed_insn_p): Likewise.
965 (ix86_option_override_internal): Don't disallow -fcf-protection.
966 * config/i386/i386.md (rdssp<mode>): Also enable for
968 (incssp<mode>): Likewise.
969 (nop_endbr): Likewise.
970 * config/i386/i386.opt (mcet): Change help message to built-in
974 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
975 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
976 enable CET built-in functions.
978 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
980 * common/config/i386/i386-common.c
981 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
982 OPTION_MASK_ISA_MOVDIRI_UNSET,
983 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
984 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
985 * config.gcc (movdirintrin.h): New header.
986 * config/i386/cpuid.h (bit_MOVDIRI,
987 bit_MOVDIR64B): New bits.
988 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
990 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
991 (VOID, PVOID, PCVOID)): New function types.
992 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
993 __builtin_ia32_directstoreu_u64,
994 __builtin_ia32_movdir64b): New builtins.
995 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
996 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
998 (ix86_valid_target_attribute_inner_p): Ditto.
999 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1000 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1001 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1002 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1003 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1004 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1005 (movdiri<mode>, movdir64b_<mode>): New.
1006 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1007 * config/i386/immintrin.h: Include movdirintrin.h.
1008 * config/i386/movdirintrin.h: New file.
1009 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1011 2018-04-19 Richard Biener <rguenther@suse.de>
1014 * cfg.c (clear_bb_flags): When loop state says we have
1015 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1017 2018-04-19 Richard Biener <rguenther@suse.de>
1019 PR tree-optimization/84737
1020 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1021 copying restrict info.
1022 (vect_setup_realignment): Use it.
1023 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1024 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1025 the first DR to all generated stores.
1026 (vectorizable_load): Likewise for loads.
1028 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1030 PR tree-optimization/85446
1031 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1032 the integral and pointer types to have the same precision.
1034 * doc/install.texi: Document --disable-cet being the default and
1037 2018-04-18 Martin Liska <mliska@suse.cz>
1039 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1042 2018-04-18 Martin Liska <mliska@suse.cz>
1045 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1048 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1049 arguments if they are comparable.
1051 2018-04-18 Martin Liska <mliska@suse.cz>
1054 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1057 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1060 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1063 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1064 ENDBR after calling __morestack.
1066 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1069 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1070 by using gcc_base_ver to generate a gcc_driver_version, and use
1071 it when generating GCC_DRIVER_NAME.
1072 * configure: Regenerate.
1074 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1077 * config.gcc: Obsolete powerpc*-*-*spe*.
1079 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1082 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1083 (stabstr_D): Change type of unum from unsigned int to
1084 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1087 2018-04-17 Jim Wilson <jimw@sifive.com>
1090 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1091 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1092 Set arg_pointer_offset after using pretend_args_size.
1094 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1096 PR rtl-optimization/85431
1097 * dse.c (record_store): Ignore zero width stores.
1100 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1101 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1102 __builtin_stack_restore rather than after it.
1103 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1104 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1105 argument instead of virtual_dynamic_stack_rtx.
1107 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1109 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1111 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1112 Add note to error message to explain internal mapping of overloaded
1113 built-in function name to non-overloaded built-in function name.
1114 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1117 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1120 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1121 where the inputs overlap with the output.
1123 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1125 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1126 (=v, v) alternative and explicit "memory" attribute.
1127 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1128 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1130 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1131 "sselog1" type instead of "sselog".
1132 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1133 "sselog". Remove explicit "memory" attribute.
1134 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1135 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1137 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1138 "sselog1" type instead of "sselog". Remove explicit "memory"
1140 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1141 use "sselog1" type instead of "sselog". Remove explicit "memory"
1143 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1144 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1146 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1147 "sselog1" type instead of "sselog". Remove explicit "memory"
1149 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1150 use "sselog1" type instead of "sselog". Remove explicit "memory"
1154 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1157 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1160 2018-04-17 Martin Jambor <mjambor@suse.cz>
1163 * ipa-cp.c (create_specialized_node): Call
1164 expand_all_artificial_thunks if necessary.
1166 2018-04-17 Martin Liska <mliska@suse.cz>
1169 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1170 in message, remote space in between '_G' and '('.
1172 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1175 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1176 avx512f_vmcmp<mode>3<round_saeonly_name>,
1177 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1178 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1179 avx512f_rndscale<mode><round_saeonly_name>,
1180 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1181 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1182 Use %<iptr>2 instead of %2 for -masm=intel.
1183 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1184 avx512f_vcvttss2usi<round_saeonly_name>,
1185 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1187 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1188 avx512f_vcvttsd2usi<round_saeonly_name>,
1189 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1190 Use %q1 instead of %1 for -masm=intel.
1191 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1192 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1193 of %3 for -masm=intel.
1194 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1196 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1198 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1200 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1201 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1203 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1205 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1206 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1207 %g1 and one with %0 and %1.
1208 (avx512er_vmrcp28<mode><round_saeonly_name>,
1209 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1211 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1212 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1213 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1214 of %0 and %{%4%} for -masm=intel.
1215 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1216 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1217 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1218 order of %0 and %{%5%}%{z%} for -masm=intel.
1220 2018-04-17 Jan Hubicka <jh@suse.cz>
1223 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1225 2018-04-17 Martin Liska <mliska@suse.cz>
1228 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1229 for target_clone error message. Make default implementation
1230 clone to be a local declaration.
1231 (separate_attrs): Add new argument and check for an empty
1233 (expand_target_clones): Handle it.
1234 (ipa_target_clone): Make redirection just for target_clones
1237 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1238 Tom de Vries <tom@codesourcery.com>
1241 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1242 tiled basic blocks without omp continue statements.
1244 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1247 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1248 vec_extract expression as having side effects to make sure it gets
1251 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1254 * config/i386/i386.c (get_builtin_code_for_version): Check
1257 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1260 * gcc/config.gcc: Support "skylake".
1261 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1263 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1264 (processor_target_table): Add "skylake".
1265 (ix86_option_override_internal): Add "skylake".
1266 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1267 PROCESSOR_CANNONLAKE.
1268 (get_builtin_code_for_version): Fix priority for
1269 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1270 PROCESSOR_SKYLAKE-AVX512.
1271 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1272 (processor_type): Add PROCESSOR_SKYLAKE.
1274 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1275 Jason Merrill <jason@redhat.com>
1278 * convert.c (convert_to_integer_1): Use direct recursion for
1279 enumeral types and types with a precision less than the number
1280 of bits in their mode.
1282 2018-04-16 Julia Koval <julia.koval@intel.com>
1285 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1286 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1288 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1291 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1292 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1293 and -mno-direct-move.
1295 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1298 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1299 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1300 Ensure negative shifts result in {0}.
1302 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1304 PR rtl-optimization/79916
1305 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1306 regs (if any) to define how to gnerate SD moves when LRA is in
1309 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1311 PR rtl-optimization/85393
1312 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1313 * except.c (expand_dw2_landing_pad_for_region): Make static.
1314 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1315 a label and unconditional jump to old_bb, rather than
1316 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1319 PR rtl-optimization/85376
1320 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1321 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1322 instead of a specific value.
1324 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1325 Bin Cheng <bin.cheng@arm.com>
1327 PR tree-optimization/82965
1328 PR tree-optimization/83991
1329 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1330 by_profile_only parameter.
1331 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1332 information if the loop was predicted to iterate too many times.
1333 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1335 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1338 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1341 2018-04-13 Martin Liska <mliska@suse.cz>
1342 Jakub Jelinek <jakub@redhat.com>
1345 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1346 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1347 * builtins.c (expand_builtin_memory_copy_args): Use
1348 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1349 handle dest_addr == pc_rtx.
1351 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1354 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1355 asked to not generate direct moves.
1356 (fix_trunc<mode>si2_stfiwx): Similar.
1357 (fix_trunc<mode>si2_internal): Similar.
1359 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1362 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1363 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1364 lookup if dest in some wider mode is known to be const0_rtx and
1365 if so, record permanent equivalence for it to be ZERO_EXTEND of
1366 the narrower mode destination.
1368 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1370 * lto-streamer-out.c (output_function): Revert 259346.
1371 * omp-expand.c (expand_oacc_for): Likewise.
1373 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1375 PR rtl-optimization/85354
1376 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1377 * sel-sched.c (sel_global_init): ... here.
1379 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1382 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1383 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1384 mode for PE-COFF targets.
1385 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1386 (i386_pe_asm_lto_end): Likewise.
1387 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1388 (TARGET_ASM_LTO_END): Likewise.
1389 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1390 (i386_pe_asm_lto_start): New function.
1391 (i386_pe_asm_lto_end): Likewise.
1393 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1394 Richard Biener <rguenther@suse.de>
1397 * lto-streamer-out.c (output_function): Fix CFG loop state before
1399 * omp-expand.c (expand_oacc_for): Handle calls to internal
1400 functions like regular functions.
1402 2018-04-12 Richard Biener <rguenther@suse.de>
1405 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1406 for the early LTO debug to properly generate references to it
1407 during DIE emission. Do not re-use that for the skeleton for
1409 (dwarf2out_early_finish): Likewise.
1411 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1414 * config/i386/sse.md
1415 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1416 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1417 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1418 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1419 and output is a reg, avoid creating invalid lowpart subreg, but
1420 instead split into a 512-bit move. Don't split if not AVX512VL,
1421 input is xmm16+ reg and output is a mem.
1422 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1423 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1424 xmm16+ reg and output is a mem.
1426 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1428 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1429 also for flag_dwarf2_cfi_asm.
1431 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1433 PR rtl-optimization/85342
1434 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1435 a bool scalar var inside of the loop instead. Don't try to update
1436 recog_data.operand after failed apply_change_group.
1438 2018-04-12 Tom de Vries <tom@codesourcery.com>
1441 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1442 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1443 array with flexible array member as array without given dimension.
1444 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1445 argument for undefined param to true.
1447 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1450 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1451 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1452 from PowerPC section.
1453 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1454 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1455 help text that is too long.
1456 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1457 help text that is too long.
1458 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1459 help text that is too long.
1461 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1463 * config/alpha/alpha.md (stack_probe_internal): Rename
1464 from "probe_stack". Update all callers.
1466 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1468 PR rtl-optimization/84566
1469 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1470 sched_macro_fuse_insns.
1472 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1475 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1476 (compute_block_dependences): ... from here.
1478 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1480 PR tree-optimization/85331
1481 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1482 from int to HOST_WIDE_INT.
1484 2018-04-11 Martin Jambor <mjambor@suse.cz>
1487 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1488 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1489 not the same as the source val.
1490 (cgraph_edge_brings_value_p): New parameter.
1491 (gather_edges_for_value): Pass destination value to
1492 cgraph_edge_brings_value_p.
1493 (perhaps_add_new_callers): Likewise.
1494 (get_info_about_necessary_edges): Likewise and exclude values brought
1495 only by self-recursive edges.
1496 (create_specialized_node): Redirect only clones of self-calling edges.
1497 (+self_recursive_pass_through_p): New function.
1498 (find_more_scalar_values_for_callers_subset): Use it.
1499 (find_aggregate_values_for_callers_subset): Likewise.
1500 (known_aggs_to_agg_replacement_list): Removed.
1501 (decide_whether_version_node): Re-calculate known constants for all
1502 remaining context clones.
1504 2018-04-11 Richard Biener <rguenther@suse.de>
1507 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1508 from early DWARF output.
1509 (dwarf2out_early_finish): Output line info unconditionally into
1510 early DWARF and add reference to it.
1512 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1515 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1516 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1517 other than V2DFmode using iptr mode attribute.
1518 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1520 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1522 PR rtl-optimization/84659
1523 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1525 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1528 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1530 (output_loc_list): Pass address of a dummy size variable even in the
1531 locview handling loop.
1532 (index_location_lists): Add comment on why skip_loc_list_entry can't
1535 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1538 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1541 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1544 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1545 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1546 and -mstring-compare-inline-limit.
1548 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1551 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1552 for stack clash protection in a register whenever we need it to be in
1555 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1557 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1558 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1560 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1563 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1565 (mlong-double-): Ditto.
1566 * config/rs6000/sysv4.opt (msdata=): Ditto.
1567 (mtls-size=): Ditto.
1569 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1571 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1572 erroneous entries for
1573 "vector int vec_ldl (int, long int *)", and
1574 "vector unsigned int vec_ldl (int, unsigned long int *)".
1575 Add comments and entries for
1576 "vector bool char vec_ldl (int, bool char *)",
1577 "vector bool short vec_ldl (int, bool short *)",
1578 "vector bool int vec_ldl (int, bool int *)",
1579 "vector bool long long vec_ldl (int, bool long long *)",
1580 "vector pixel vec_ldl (int, pixel *)",
1581 "vector long long vec_ldl (int, long long *)",
1582 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1583 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1584 type tree bool_long_long_type_node and correct definition of
1585 bool_V2DI_type_node to make reference to this new type tree.
1586 (rs6000_mangle_type): Replace erroneous reference to
1587 bool_long_type_node with bool_long_long_type_node.
1588 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1589 comments to emphasize sign distinctions for char and int types and
1590 replace RS6000_BTI_bool_long constant with
1591 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1592 use of RS6000_BTI_pixel.
1593 (bool_long_type_node): Remove this macro definition.
1594 (bool_long_long_type_node): New macro definition
1596 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1598 PR rtl-optimization/85300
1599 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1600 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1601 simplify_unary_operation fails.
1603 2018-04-10 Martin Liska <mliska@suse.cz>
1605 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1606 cgraph_edge and ipa_ref.
1608 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1612 * config/i386/sse.md
1613 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1614 computation of the VEC_MERGE selector from mask.
1615 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1616 Fix decoding of the VEC_MERGE selector into mask.
1618 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1620 PR tree-optimization/85286
1621 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1623 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1625 * final.c (final_1): Set insn_last_address as well as
1626 insn_current_address.
1628 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1631 * explow.c (emit_stack_probe): Call validize_mem on memory location
1632 before passing it to gen_probe_stack. Create address operand and
1633 legitimize it for the probe_stack_address case.
1635 2018-04-09 Jan Hubicka <jh@suse.cz>
1638 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1639 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1640 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1642 (free_lang_data): Rebuild type inheritance graph.
1644 2018-04-09 Martin Sebor <msebor@redhat.com>
1646 * invoke.texi (-finline-small-functions): Mention other optimization
1648 (-findirect-inlining, -fpartial-inlining): Same.
1649 (-finline-functions-called-once): Same.
1650 (-freorder-blocks-and-partition): Same.
1652 2018-04-09 Jan Hubicka <jh@suse.cz>
1655 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1656 jumps; choose last target that matches the criteria (i.e.
1657 no partition changes for non-crossing jumps).
1658 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1659 support for redirecting crossing jumps to non-crossing.
1661 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1663 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1664 also for naked functions.
1666 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1668 * config/arc/arc.md (add_shift): New pattern.
1669 (add_shift2): Likewise.
1670 (sub_shift): Likewise.
1671 (sub_shift_cmp0_noout): Likewise.
1672 (compare_si_ashiftsi): Likewise.
1673 (xbfu_cmp0_noout): New combine pattern.
1674 (xbfu_cmp0"): Likewise.
1675 (movsi_set_cc_insn): Place the predicable variant first.
1676 (commutative_binary_cmp0_noout): Remove clobber.
1677 (commutative_binary_cmp0): New pattern.
1678 (noncommutative_binary_cmp0): Likewise.
1679 (noncommutative_binary_cmp0_noout): Likewise.
1680 (noncommutative_binary_comparison_result_used): Removed.
1681 (rsub_cmp0): New pattern.
1682 (rsub_cmp0_noout): Likewise.
1683 (extzvsi): Changed, keep only meaningful variants.
1684 (SQH, SEZ): New iterators.
1685 (SQH_postfix): New mode attribute.
1686 (SEZ_prefix): New code attribute.
1687 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1688 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1689 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1691 (noncommutative_operator): Check the availability of barrel
1694 2018-04-09 Richard Biener <rguenther@suse.de>
1696 PR tree-optimization/85284
1697 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1698 Only use the niter constraining form of simple_iv when the exit
1701 2018-04-09 Tom de Vries <tom@codesourcery.com>
1704 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1705 (define_expand "*memory_barrier"): New define_expand.
1706 (define_insn "memory_barrier"): New insn.
1708 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1710 PR rtl-optimization/80463
1711 PR rtl-optimization/83972
1712 PR rtl-optimization/83480
1714 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1715 correct producer for the insn.
1716 (tidy_control_flow): Fixup seqnos in case of debug insns.
1718 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1720 PR rtl-optimization/83913
1722 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1723 different sched-times when merging exprs.
1725 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1727 PR rtl-optimization/83962
1729 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1730 tidy_fallthru_edge and tidy_control_flow.
1732 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1734 PR rtl-optimization/83530
1736 * sel-sched.c (force_next_insn): New global variable.
1737 (remove_insn_for_debug): When force_next_insn is true, also leave only
1738 next insn in the ready list.
1739 (sel_sched_region): When the region wasn't scheduled, make another pass
1740 over it with force_next_insn set to 1.
1742 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1744 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1746 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1747 for interrupt control.
1748 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1749 functions for interrupt control.
1750 * config/nds32/nds32-intrinsic.md: Likewise.
1751 * config/nds32/nds32_intrinsic.h: Likewise.
1752 * config/nds32/nds32.h (nds32_builtins): Likewise.
1754 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1756 * config/nds32/nds32.c (nds32_init_machine_status,
1757 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1758 strict_aligned_p field.
1759 (nds32_expand_to_rtl_hook): New function.
1760 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1761 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1763 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1764 Chung-Ju Wu <jasonwucj@gmail.com>
1766 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1767 * config/nds32/nds32-n7.md: New file.
1768 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1769 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1771 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1772 * config/nds32/nds32.md (pipeline_model): Add n7.
1773 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1774 * config/nds32/pipelines.md: Include n7 settings.
1776 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1777 Chung-Ju Wu <jasonwucj@gmail.com>
1779 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1780 * config/nds32/nds32-e8.md: New file.
1781 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1782 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1784 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1785 * config/nds32/nds32.md (pipeline_model): Add e8.
1786 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1787 * config/nds32/pipelines.md: Include e8 settings.
1789 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1790 Chung-Ju Wu <jasonwucj@gmail.com>
1792 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1793 * config/nds32/nds32-n8.md: New file.
1794 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1795 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1797 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1798 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1799 * config/nds32/nds32.md (pipeline_model): Add n8.
1800 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1801 * config/nds32/pipelines.md: Include n8 settings.
1803 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1804 Chung-Ju Wu <jasonwucj@gmail.com>
1806 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1807 * config/nds32/nds32-n9-2r1w.md: New file.
1808 * config/nds32/nds32-n9-3r2w.md: New file.
1809 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1810 nds32_register_ports): New or modify for cpu n9.
1811 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1813 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1814 * config/nds32/nds32-utils.c: New file.
1815 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1816 TARGET_MUL_SLOW): Define.
1817 * config/nds32/nds32.md (pipeline_model): New attribute.
1818 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1819 New options that support cpu n9.
1820 * config/nds32/pipelines.md: Include n9 settings.
1821 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1823 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1825 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1826 information if necessary.
1827 (output_cond_branch_compare_zero): Likewise.
1828 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1829 (nds32_target_alignment): Refine for alignment.
1830 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1831 (FUNCTION_BOUNDARY): Modify.
1832 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1834 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1836 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1838 * config/nds32/constants.md (unspec_volatile_element): Add values for
1839 TLB operation and data prefetch.
1840 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1841 functions for TLB operation and data prefetch.
1842 * config/nds32/nds32-intrinsic.md: Likewise.
1843 * config/nds32/nds32_intrinsic.h: Likewise.
1844 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1845 (nds32_print_operand): Likewise.
1846 * config/nds32/nds32.h (nds32_builtins): Likewise.
1848 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1849 Andrew Pinski <pinsika@gcc.gnu.org>
1852 * match.pd: Use constant_boolean_node of correct type instead of
1853 boolean_true_node or boolean_false_node for simplifying
1854 pointer comparisons to zero.
1856 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1858 PR tree-optimization/80021
1859 * tree.c (verify_type_variant): Make error call in verify_variant_match
1860 translatable and remove final full stop.
1862 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1864 * config/nds32/constants.md (unspec_volatile_element): Add
1865 UNSPEC_VOLATILE_EH_RETURN.
1866 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1867 nds32_output_stack_pop): Support dwarf exception handling process.
1868 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1869 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1870 exception handling process.
1871 (nds32_compute_stack_frame): Likewise.
1872 (nds32_return_addr_rtx): Likewise.
1873 (nds32_initial_elimination_offset): Likewise.
1874 (nds32_expand_prologue): Likewise.
1875 (nds32_expand_epilogue): Likewise.
1876 (nds32_dynamic_chain_address): New function.
1877 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1879 (DYNAMIC_CHAIN_ADDRESS): Define.
1880 (EH_RETURN_DATA_REGNO): Define.
1881 (EH_RETURN_STACKADJ_RTX): Define.
1882 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1883 patterns for dwarf exception handling.
1885 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1887 * config/nds32/nds32.h: Clean up obsolete macros.
1889 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1891 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1892 Add enum values for particular instructions.
1893 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1894 particular intrinsic functions.
1895 * config/nds32/nds32-intrinsic.md: Likewise.
1896 * config/nds32/nds32_intrinsic.h: Likewise.
1897 * config/nds32/nds32.h (nds32_builtins): Likewise.
1898 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1899 (btst, ave): New patterns for particular instructions.
1901 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1903 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1904 Add enum values for atomic load/store and memory sync.
1905 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1907 * config/nds32/nds32-intrinsic.md: Likewise.
1908 * config/nds32/nds32_intrinsic.h: Likewise.
1909 * config/nds32/nds32.h (nds32_builtins): Likewise.
1911 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1913 PR tree-optimization/85257
1914 * fold-const.c (native_encode_vector): If not all elts could fit
1915 and off is -1, return 0 rather than offset.
1916 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1917 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1918 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1919 adjust buffer in native_interpret_expr call.
1921 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1923 * config/nds32/constants.md (unspec_volatile_element): Add cache
1924 control enum values.
1925 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1926 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1927 * config/nds32/nds32.c (nds32_cctl_names): New.
1928 (nds32_print_operand): Handle cache control register names.
1929 * config/nds32/nds32.h (nds32_builtins): New enum values.
1930 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1932 * config/nds32/nds32.md (type): Add mmu.
1933 * config/nds32/pipelines.md (simple_insn): Add mmu.
1935 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1937 * config/nds32/nds32.md (type): Remove call.
1938 * config/nds32/pipelines.md (simple_insn): Likewise.
1940 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1942 * config/nds32/constants.md (unspec_volatile_element): Add
1943 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1944 UNSPEC_VOLATILE_FMFCFG.
1945 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1946 description for fmfcfg and fmfcsr.
1947 (bdesc_1arg): Add fmtcsr.
1948 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1949 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1950 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1951 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1952 unspec_fmfcfg): New patterns.
1953 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1954 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1955 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1956 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1957 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1958 __nds32__fmfcfg): Define.
1960 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1962 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1963 intrinsic register names.
1964 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1965 intrinsic register enum values and macros.
1967 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1969 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1970 for load/store addressing form.
1971 (nds32_print_operand_address): Likewise.
1973 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1976 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1977 based on LABEL_REF. Remove useless assertion.
1978 (pic_address_needs_scratch): Fix formatting.
1979 (sparc_legitimize_pic_address): Minor tweaks.
1980 (sparc_delegitimize_address): Adjust assertion accordingly.
1981 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1982 into symbolic_operand.
1983 (movsi_high_pic_label_ref): Likewise.
1984 (movsi_lo_sum_pic_label_ref): Likewise.
1985 (movdi_pic_label_ref): Likewise.
1986 (movdi_high_pic_label_ref): Likewise.
1987 (movdi_lo_sum_pic_label_ref): Likewise.
1989 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1991 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1992 custom LIB_SPEC setup.
1994 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1995 Kito Cheng <kito.cheng@gmail.com>
1997 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1998 * config/riscv/freebsd.h: New.
2000 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2002 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2003 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2006 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2007 Kito Cheng <kito.cheng@gmail.com>
2009 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2010 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2011 * config/nds32/nds32-protos.h (nds32_output_call,
2012 nds32_output_return): Declare.
2013 * config/nds32/nds32.md: Refine all the call and return patterns.
2015 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2018 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2019 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2021 PR rtl-optimization/84872
2022 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2023 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2026 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2028 * expr.c (copy_blkmode_to_reg): Revert 254862.
2029 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2031 2018-04-06 Richard Biener <rguenther@suse.de>
2034 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2035 after seeing a component reference with an adjacent field. Treat
2036 refs to arrays at struct end of external decls similar to
2037 refs to unconstrained commons.
2039 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2042 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2043 look through SAVE_EXPRs with non-side-effects argument. Adjust
2045 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2048 2018-04-06 Richard Biener <rguenther@suse.de>
2051 * alias.c (find_base_term): New wrapper around find_base_term
2052 unwinding CSELIB_VAL_PTR changes.
2053 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2056 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2058 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2060 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2061 constant definitions.
2062 ("nop"): lr 0,0 -> nopr r0
2063 ("nop_lr0", "nop_lr1"): New insn definitions.
2065 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2067 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2068 NDS32_V3PUSH_AVAILABLE_P macro.
2070 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2071 Chung-Ju Wu <jasonwucj@gmail.com>
2073 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2074 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2075 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2076 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2077 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2078 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2079 * config/nds32/constraints.md: New constraints and checking for hard
2080 float configuration.
2081 * config/nds32/iterators.md: New mode iterator and attribute for hard
2082 float configuration.
2083 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2085 * config/nds32/nds32-fpu.md: New file.
2086 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2087 deal with hard float code generation.
2088 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2090 (abi_type, float_reg_number): New enum type.
2091 * config/nds32/nds32-predicates.c: New predicates for hard float.
2092 * config/nds32/nds32-protos.h: Declare functions for hard float.
2093 * config/nds32/nds32.c: Implementation for hard float configuration.
2094 * config/nds32/nds32.h: Definitions for hard float configuration.
2095 * config/nds32/nds32.md: Include hard float machine description and
2096 modify patterns for hard float configuration.
2097 * config/nds32/nds32.opt: New options for hard float configuration.
2098 * config/nds32/predicates.md: New predicates for hard float
2101 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2103 * common/config/nds32/nds32-common.c
2104 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2106 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2109 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2110 CONSTRUCTOR_ELT (ctor, ...)->value.
2112 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2115 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2117 2018-04-05 Tom de Vries <tom@codesourcery.com>
2120 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2123 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2124 Kito Cheng <kito.cheng@gmail.com>
2126 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2127 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2128 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2129 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2131 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2132 Kito Cheng <kito.cheng@gmail.com>
2134 * config/nds32/constraints.md (Ufe): New memory constraint.
2135 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2136 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2137 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2139 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2140 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2142 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2144 * config/nds32/nds32.md: Use optimize_size in the condition for
2145 alu-shift instructions.
2147 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2149 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2151 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2153 * config/nds32/nds32.md (negsi2): Refine pattern.
2155 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2156 Chung-Ju Wu <jasonwucj@gmail.com>
2158 * config/nds32/iterators.md (shift_rotate): New code iterator.
2159 (shift): New code attribute.
2160 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2161 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2162 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2163 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2164 bit-wise operations.
2165 (andsi3, *andsi3): Ditto.
2166 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2167 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2168 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2169 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2170 nds32_ior_operand, nds32_xor_operand): New predicates.
2172 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2174 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2175 (addsi3, subsi3): ... this.
2177 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2179 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2181 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2183 * config/nds32/nds32.md: Adjust indention.
2185 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2187 * config/nds32/nds32.md (feature): New attribute.
2189 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2191 * config/nds32/nds32.md (subtype): New attribute.
2193 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2196 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2197 expansion to perform a bitwise AND of the argument followed by a
2198 boolean negation of the result.
2200 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2202 PR rtl-optimization/84878
2203 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2204 the basic block. Assert the use reference is not artificial and that
2205 it has an associated insn.
2207 2018-04-04 Michael Matz <matz@suse.de>
2209 * builtins.c (compute_objsize): Pass correct operand
2210 to array_at_struct_end_p.
2212 2018-04-04 Richard Biener <rguenther@suse.de>
2215 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2216 from contexts for DINFO_LEVEL_TERSE and below.
2218 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2220 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2221 resiter_operand condition.
2222 * config/nds32/nds32.md (*move<mode>): Ditto.
2224 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2225 Monk Chiang <sh.chiang04@gmail.com>
2227 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2229 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2231 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2233 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2234 Kito Cheng <kito.cheng@gmail.com>
2236 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2237 nds32_cond_code_str, output_cond_branch,
2238 output_cond_branch_compare_zero, nds32_expand_cbranch,
2239 nds32_expand_cstore, nds32_expand_movcc,
2240 nds32_output_cbranchsi4_equality_zero,
2241 nds32_output_cbranchsi4_equality_reg,
2242 nds32_output_cbranchsi4_equality_reg_or_const_int,
2243 nds32_output_cbranchsi4_greater_less_zero: New functions.
2244 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2245 nds32_expand_cstore, nds32_expand_movcc,
2246 nds32_output_cbranchsi4_equality_zero,
2247 nds32_output_cbranchsi4_equality_reg,
2248 nds32_output_cbranchsi4_equality_reg_or_const_int,
2249 nds32_output_cbranchsi4_greater_less_zero): Declare.
2250 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2251 nds32_rimm11s_operand): New predicates.
2252 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2253 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2256 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2258 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2259 * config/nds32/nds32.md: Ditto.
2260 * config/nds32/pipelines.md: Ditto.
2262 2018-04-04 Richard Biener <rguenther@suse.de>
2264 PR tree-optimization/85168
2265 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2266 propagating abnormals.
2268 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2270 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2272 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2273 Kito Cheng <kito.cheng@gmail.com>
2275 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2276 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2277 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2278 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2279 * config/nds32/nds32.md (sibcall_internal): New.
2280 (sibcall_register): Remove.
2281 (sibcall_immediate): Remove.
2282 (sibcall_value_internal): New.
2283 (sibcall_value_register): Remove.
2284 (sibcall_value_immediate): Remove.
2285 * config/nds32/predicates.md (nds32_general_register_operand): New.
2286 (nds32_call_address_operand): New.
2288 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2290 PR rtl-optimization/85167
2291 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2292 bb_defs if *split_p, instead preinitialize it to NULL.
2294 PR tree-optimization/85156
2295 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2296 evaluating the argument multiple times.
2298 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2300 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2302 (_mm_cvtpd_ps): Likewise.
2303 (_mm_cvttpd_epi32): Likewise.
2304 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2305 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2306 vector, pixel, and bool following altivec.h include.
2308 2018-04-03 Martin Sebor <msebor@redhat.com>
2310 * doc/extend.texi (Common Function Attributes): Clarify.
2311 (const attribute): Likewise.
2312 (pure attribute): Likewise.
2314 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2317 * config/i386/i386.c (ix86_expand_vector_set): Use
2318 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2320 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2322 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2323 instructions when changing rounding bits to preserve precision bits
2324 in the x87 control word.
2326 2018-04-03 Martin Liska <mliska@suse.cz>
2328 PR tree-optimization/82491
2329 * rtl.h (strip_offset_and_add): Replace += suboffset with
2330 poly_uint64 () + suboffset.
2332 2018-03-29 Martin Liska <mliska@suse.cz>
2333 Martin Jambor <mjambor@suse.cz>
2336 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2337 param_type is not an integral or pointer type.
2339 2018-04-03 Richard Biener <rguenther@suse.de>
2341 * sese.h (recompute_all_dominators): Remove.
2343 2018-04-02 Martin Sebor <msebor@redhat.com>
2345 * doc/invoke.texi (-Wrestrict): Fix typos.
2347 2018-04-02 Jim Wilson <jimw@sifive.com>
2349 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2350 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2351 (<optab>di3, <optab>si3_extend): Likewise.
2352 (<optab>si3_mask, <optab>si3_mask_1): New.
2353 (<optab>di3_mask, <optab>di3_mask_1): New.
2354 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2355 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2356 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2358 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2360 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2363 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2365 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2366 (nds32_canonicalize_comparison): New function.
2368 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2369 Kito Cheng <kito.cheng@gmail.com>
2370 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2372 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2373 * config/nds32/constants.md (unspec_volatile_element): Add
2374 UNSPEC_VOLATILE_RELAX_GROUP.
2375 * config/nds32/nds32-relax-opt.c: New file.
2376 * config/nds32/nds32-predicates.c
2377 (nds32_symbol_load_store_p): New function.
2378 * config/nds32/nds32-protos.h
2379 (nds32_symbol_load_store_p): Declare function.
2380 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2381 * config/nds32/nds32.c
2382 (nds32_register_pass): New function to register pass.
2383 (nds32_register_passes): New function to register passes.
2384 * config/nds32/nds32.md (relax_group): New pattern.
2385 * config/nds32/nds32.opt (mrelax-hint): New option.
2386 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2388 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2390 * config/nds32/t-nds32: Modify files dependency.
2392 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2394 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2395 (PROFILE_HOOK): Define its implementation.
2397 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2399 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2400 type and 32-bit size.
2402 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2405 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2406 (V_128_256): New mode iterator.
2407 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2408 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2409 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2411 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2412 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2414 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2417 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2418 NaN inputs correctly.
2420 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2423 * config/rs6000/vsx.md (??r): New mode attribute.
2424 (*vsx_mov<mode>_64bit): Use it.
2425 (*vsx_mov<mode>_32bit): Likewise.
2427 2018-03-30 Martin Sebor <msebor@redhat.com>
2429 PR tree-optimization/84818
2430 * builtins.c (check_access): Use warning_n.
2432 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2435 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2437 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2440 2018-03-30 Julia Koval <julia.koval@intel.com>
2443 * x86-tune.def (movx, partial_reg_dependency): Enable for
2446 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2449 * lra-constraints.c (process_alt_operands): Move setting
2450 this_alternative_matches below.
2452 2018-03-29 Martin Liska <mliska@suse.cz>
2455 * doc/invoke.texi: Document how LTO works with debug info.
2456 Describe auto-load support of binutils. Mention 'x86-64'
2457 as valid option value of -march option.
2459 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2461 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2464 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2465 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2466 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2469 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2472 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2473 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2474 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2475 for RS6000_BTM_POWERPC64.
2476 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2477 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2478 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2483 2018-03-28 Carl Love <cel@us.ibm.com>
2486 2017-09-27 Carl Love <cel@us.ibm.com>
2488 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2489 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2490 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2493 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2495 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2496 instead of __vector bool.
2497 (_mm_max_pu8): Likewise.
2498 (_mm_min_pi16): Likewise.
2500 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2503 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2504 (DIVWEUO): Likewise.
2506 (DIVDEUO): Likewise.
2507 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2508 DIVWEUO and DIVDEUO.
2509 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2510 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2511 (div_extend): Likewise.
2512 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2514 (__builtin_divweuo): Likewise.
2515 (__builtin_divdeo): Likewise.
2516 (__builtin_divdeuo): Likewise.
2518 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2521 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2522 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2524 PR tree-optimization/82004
2525 * gimple-match-head.c (optimize_pow_to_exp): New function.
2526 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2527 Don't fold to exp if optimize_pow_to_exp is false.
2529 2018-03-28 Martin Liska <mliska@suse.cz>
2532 * calls.c (initialize_argument_information): Fix trailing space.
2533 * common.opt: Fix typo and provide better explanation for
2534 -fsanitize-coverage option.
2535 * config/i386/i386.opt: Fix typo.
2537 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2538 Martin Liska <mliska@suse.cz>
2541 * gimplify.c (asan_poison_variable): Don't do the check for
2542 gimplify_omp_ctxp here.
2543 (gimplify_decl_expr): Do it here.
2544 (gimplify_target_expr): Likewise.
2546 2018-03-28 Martin Liska <mliska@suse.cz>
2549 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2550 chkp_type_bounds_count if MPX is not enabled.
2552 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2554 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2556 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2559 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2560 function to create the function decl for complex long double
2561 multiply and divide for -mabi=ieeelongdouble.
2562 (init_float128_ieee): Call it.
2564 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2567 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2568 -fcf-protection=branch -mibt.
2569 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2571 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2574 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2576 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2579 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2580 extern array declarations.
2582 2018-03-27 Richard Biener <rguenther@suse.de>
2585 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2586 explicit single_use checks.
2588 2018-03-27 Richard Biener <rguenther@suse.de>
2590 PR tree-optimization/85082
2591 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2594 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2596 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2597 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2598 Turn on fasynchronous-unwind-tables and funwind-tables.
2600 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2603 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2604 (*bmi_blsr_<mode>_ccz): Ditto.
2606 2018-03-26 Tom de Vries <tom@codesourcery.com>
2608 PR tree-optimization/85063
2609 * omp-general.c (offloading_function_p): New function. Factor out
2611 * omp-offload.c (pass_omp_target_link::gate): ... here.
2612 * omp-general.h (offloading_function_p): Declare.
2613 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2614 with attribute omp declare target for offloading functions.
2616 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2618 PR tree-optimization/84005
2619 * tree-data-ref.h (get_base_for_alignment): Declare.
2620 * tree-data-ref.c (get_base_for_alignment_1): New function.
2621 (get_base_for_alignment): Likewise.
2622 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2623 get_base_for_alignment to find a suitable base object, instead
2624 of always using drb->base_address.
2626 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2629 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2630 known size by default.
2632 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2635 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2636 and non BLKmode operands.
2638 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2641 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2642 Clean up attributes.
2644 2018-03-23 Richard Biener <rguenther@suse.de>
2647 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2648 we are going to emit early debug for LTO.
2650 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2653 * function.c (match_asm_constraints_1): Don't optimize if input
2654 doesn't satisfy general_operand predicate for output's mode.
2657 * alias.c (write_dependence_p): Don't require for x_canonicalized
2658 non-VOIDmode if x has VOIDmode.
2661 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2662 just don't try to optimize it rather than assert it never happens.
2664 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2666 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2667 macro expansions for definition of ST_INTERNAL_<mode> and
2668 LD_INTERNAL_<mode> builtins.
2669 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2671 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2673 (altivec_expand_st_builtin): Likewise.
2674 (altivec_expand_builtin): Remove calls to deleted functions.
2675 (rs6000_address_for_altivec): Delete this function.
2676 * config/rs6000/vector.md: Remove expands for
2677 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2679 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2682 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2683 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2684 re-computing once computed.
2685 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2686 (arm_init_machine_status): Initialize
2687 machine->static_chain_stack_bytes.
2689 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2692 * doc/extend.texi: Add four new prototypes for vec_ld.
2693 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2694 definitions for more logical presentation.
2695 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2696 entries for V1TI variants of __builtin_altivec_ld builtin.
2697 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2698 handling of V1TI variant of LVX icode pattern.
2699 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2700 (rs6000_gimple_fold_builtin): Likewise.
2701 (altivec_init_builtins): Add code to define
2702 __builtin_altivec_lvx_v1ti function.
2704 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2707 * function.c (match_asm_constraints_1): Don't do the optimization
2708 if input isn't a REG, SUBREG, MEM or constant.
2710 2018-03-22 Tom de Vries <tom@codesourcery.com>
2712 PR tree-optimization/84956
2713 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2714 bb_has_abnormal_pred.
2716 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2719 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2720 DECL_INITIAL (decl) to decl at the end.
2721 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2724 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2726 * doc/extend.texi (__builtin_tgmath): Document when complex
2727 integer types are treated as _Complex _Float64.
2729 2018-03-21 Tom de Vries <tom@codesourcery.com>
2731 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2733 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2735 PR tree-optimization/84960
2736 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2737 if it is ENTRY block, move them into single succ of ENTRY in that case.
2739 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2741 PR tree-optimization/84811
2742 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2743 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2744 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2745 (known_subrange_p): Update accordingly. Cast each value involved
2746 in the size comparison, rather than casting the result of the
2749 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2751 PR tree-optimization/84982
2752 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2753 by flipping the least significant bit rather than all bits from
2754 bitpos to bitpos + bitsize - 1.
2756 2018-03-21 Nathan Sidwell <nathan@acm.org>
2758 * doc/extend.texi (Deprecated Features): Remove mention of
2759 long-deleted deprecations.
2761 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2764 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2765 * configure: Regenerate.
2767 2018-03-21 Tom de Vries <tom@codesourcery.com>
2769 PR tree-optimization/83126
2770 * tree-parloops.c (num_phis): New function.
2771 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2773 2018-03-21 Nathan Sidwell <nathan@acm.org>
2775 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2776 mention anon-struct/union members and trailing attributes.
2778 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2780 PR tree-optimization/84969
2781 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2782 builtin memset partitions if they set different rhs values.
2784 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2786 PR rtl-optimization/84989
2787 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2788 VEC_DUPLICATE with scalar result mode.
2790 2018-03-21 Martin Liska <mliska@suse.cz>
2793 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2794 not intended return statement.
2796 2018-03-21 Martin Liska <mliska@suse.cz>
2799 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2800 (chkp_find_bound_slots_1): Limit number of iterations.
2802 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2805 * Minor grammar fixes for x86 options.
2807 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2810 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2811 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2813 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2816 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2817 (altivec_lvx_<mode>_1op): Likewise.
2818 (altivec_stvx_<mode>_2op): Likewise.
2819 (altivec_stvx_<mode>_1op): Likewise.
2820 (altivec_lvx_<VM2:mode>): New define_expand.
2821 (altivec_stvx_<VM2:mode>): Likewise.
2822 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2823 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2824 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2825 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2826 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2827 (rs6000_gen_lvx): Likewise.
2828 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2829 (altivec_expand_stv_builtin): Likewise.
2830 (altivec_expand_builtin): Likewise.
2831 * config/rs6000/vector.md: Likewise.
2833 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2836 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2839 2018-03-20 Richard Biener <rguenther@suse.de>
2842 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2843 sign-conversions as zero, fall back to standard scalar_stmt
2846 2018-03-20 Martin Liska <mliska@suse.cz>
2849 * predict.c (rebuild_frequencies): Handle case when we have
2850 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2852 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2855 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2856 flag_section_anchors.
2857 * varasm.c (use_blocks_for_decl_p): Remove hack for
2858 dw2_force_const_mem.
2861 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2863 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2864 be created, use lowpart_subreg of operands[0] rather than operands[0]
2866 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2867 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2868 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2869 and n constraint instead of aarch64_shift_imm_di and Usd.
2870 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2871 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2873 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2876 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2877 to favor GPR over NEON registers.
2878 (<shift>di3_neon): Likewise.
2880 2018-03-20 Tom de Vries <tom@codesourcery.com>
2883 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2884 (nvptx_process_pars): Emit bar.sync asap and alap.
2886 2018-03-20 Tom de Vries <tom@codesourcery.com>
2889 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2890 seen_label if seen_label is already set.
2892 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2895 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2896 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2897 Use 1U instead of 1. Formatting fixes.
2900 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2901 instead of TREE_TYPE (s1) for the return value.
2903 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2905 PR tree-optimization/84946
2906 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2907 bitsize + bitsize in poly_uint64 rather than poly_int64.
2910 * dwarf2asm.c: Include fold-const.c.
2911 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2912 of decl rather than decl itself.
2914 PR rtl-optimization/84643
2915 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2917 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2920 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2921 calling assemble_variable.
2923 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2926 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2927 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2929 2018-03-19 Jim Wilson <jimw@sifive.com>
2932 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2933 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2934 (riscv_first_stack_step): Likewise.
2935 (riscv_option_override): Use STACK_BOUNDARY instead of
2937 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2939 (BIGGEST_ALIGNMENT): Set to 128.
2940 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2941 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2944 2018-03-19 Richard Biener <rguenther@suse.de>
2946 PR tree-optimization/84933
2947 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2948 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2950 2018-03-19 Richard Biener <rguenther@suse.de>
2952 PR tree-optimization/84859
2953 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2954 (cond_if_else_store_replacement): Perform sinking operation on
2955 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2956 Generalize what a BB with a single eligible store is.
2958 2018-03-19 Richard Biener <rguenther@suse.de>
2960 PR tree-optimization/84929
2961 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2962 chrec_is_positive against non-chrec arg.
2964 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2967 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2969 2018-03-18 Martin Liska <mliska@suse.cz>
2971 PR rtl-optimization/84635
2972 * regrename.c (build_def_use): Use matches_mode only when
2975 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2977 PR tree-optimization/84913
2978 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2979 vectorize chains of COND_EXPRs.
2981 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2983 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2985 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2987 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2989 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2991 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2993 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2994 Kito Cheng <kito.cheng@gmail.com>
2996 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2997 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2998 (nds32_adjust_reg_alloc_order): New function.
2999 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3001 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3003 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3004 nds32_print_operand, nds32_print_operand_address): Use
3005 HOST_WIDE_INT_PRINT_DEC instead.
3007 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3009 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3011 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3014 * config/i386/i386.c (initial_ix86_tune_features,
3015 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3017 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3018 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3019 rather than 1u << ix86_tune. Formatting fix.
3020 (ix86_option_override_internal): Change ix86_arch_mask from
3021 unsigned int to unsigned HOST_WIDE_INT, initialize to
3022 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3023 (ix86_function_specific_restore): Likewise.
3025 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3028 * postreload.c (reload_combine_recognize_pattern): Perform
3029 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3030 truncate_int_for_mode the result for the destination's mode.
3033 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3034 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3036 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3039 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3042 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3043 regno_allocno_class_array and sorted_pseudos.
3044 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3045 insns where regno is used.
3047 2018-03-16 Martin Liska <mliska@suse.cz>
3050 * multiple_target.c (create_dispatcher_calls): Redirect
3051 reference in the symbol table.
3053 2018-03-16 Martin Liska <mliska@suse.cz>
3056 * multiple_target.c (create_dispatcher_calls): Redirect also
3059 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3063 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3064 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3065 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3067 2018-03-16 Julia Koval <julia.koval@intel.com>
3069 * doc/invoke.texi (Skylake Server): Add CLWB.
3070 Cannonlake): Remove CLWB.
3072 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3074 PR tree-optimization/84841
3075 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3077 (FLOAT_ONE_CONST_TYPE): Define.
3078 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3079 (sort_by_operand_rank): Put entries with higher constant_type last
3080 rather than first to match comments.
3082 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3084 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3087 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3090 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3091 check and use address of target always.
3093 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3096 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3097 (indirect_thunk_bnd_needed): Likewise.
3098 (indirect_thunks_used): Likewise.
3099 (indirect_thunks_bnd_used): Likewise.
3100 (indirect_return_needed): New.
3101 (indirect_return_bnd_needed): Likewise.
3102 (output_indirect_thunk_function): Add a bool argument for
3104 (output_indirect_thunk_function): Don't generate alias for
3105 function return thunk.
3106 (ix86_code_end): Call output_indirect_thunk_function to generate
3107 function return thunks.
3108 (ix86_output_function_return): Set indirect_return_bnd_needed
3109 and indirect_return_needed instead of indirect_thunk_bnd_needed
3110 and indirect_thunk_needed.
3112 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3114 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3115 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3116 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3118 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3119 Paul Hua <paul.hua.gm@gmail.com>
3122 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3124 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3126 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3127 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3130 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3133 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3134 instead of GET_MODE_SIZE when comparing Units.
3136 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3139 * varasm.c (hash_section): Return an unchangeble hash value
3140 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3141 Return !aarch64_can_use_per_function_literal_pools_p ().
3143 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3146 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3147 rather than address of cmode as last argument to prepare_cmp_insn.
3149 2018-03-15 Julia Koval <julia.koval@intel.com>
3151 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3152 F_AVX512VNNI, F_AVX512BITALG): New.
3154 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3157 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3158 insn for floating-point loads and stores.
3160 2018-03-14 Carl Love <cel@us.ibm.com>
3162 * config/rs6000/rs6000-c.c: Add macro definitions for
3163 ALTIVEC_BUILTIN_VEC_PERMXOR.
3164 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3165 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3166 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3167 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3169 * config/doc/extend.texi: Add prototypes for vec_permxor.
3171 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3174 * diagnostic-show-locus.c (class layout_point): Convert m_line
3175 from int to linenum_type.
3176 (line_span::comparator): Use linenum "compare" function when
3177 comparing line numbers.
3178 (test_line_span): New function.
3179 (layout_range::contains_point): Convert param "row" from int to
3181 (layout_range::intersects_line_p): Likewise.
3182 (layout::will_show_line_p): Likewise.
3183 (layout::print_source_line): Likewise.
3184 (layout::should_print_annotation_line_p): Likewise.
3185 (layout::print_annotation_line): Likewise.
3186 (layout::print_leading_fixits): Likewise.
3187 (layout::annotation_line_showed_range_p): Likewise.
3188 (struct line_corrections): Likewise for field m_row.
3189 (line_corrections::line_corrections): Likewise for param "row".
3190 (layout::print_trailing_fixits): Likewise.
3191 (layout::get_state_at_point): Likewise.
3192 (layout::get_x_bound_for_row): Likewise.
3193 (layout::print_line): Likewise.
3194 (diagnostic_show_locus): Likewise for locals "last_line" and
3196 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3197 * input.c (selftest::test_linenum_comparisons): New function.
3198 (selftest::input_c_tests): Call it.
3199 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3200 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3201 * selftest.h (ASSERT_GT): New macro.
3202 (ASSERT_GT_AT): New macro.
3203 (ASSERT_LT): New macro.
3204 (ASSERT_LT_AT): New macro.
3206 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3208 PR rtl-optimization/84780
3209 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3211 2018-03-14 Martin Liska <mliska@suse.cz>
3213 * tree.c (record_node_allocation_statistics): Use
3214 get_stats_node_kind.
3215 (get_stats_node_kind): New function extracted from
3216 record_node_allocation_statistics.
3217 (free_node): Use get_stats_node_kind.
3219 2018-03-14 Richard Biener <rguenther@suse.de>
3221 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3222 that the value-set of ANTIC_IN doesn't grow.
3225 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3227 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3228 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3230 2018-03-14 Julia Koval <julia.koval@intel.com>
3232 * config.gcc (icelake-client, icelake-server): New.
3234 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3235 (initial_ix86_arch_features): Ditto.
3236 (PTA_SKYLAKE): Add SGX.
3237 (PTA_ICELAKE): Remove.
3238 (PTA_ICELAKE_CLIENT): New.
3239 (PTA_ICELAKE_SERVER): New.
3240 (ix86_option_override_internal): Split up icelake on icelake client and
3242 (get_builtin_code_for_version): Ditto.
3243 (fold_builtin_cpu): Ditto.
3244 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3245 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3246 * config/i386/i386.h (processor_type): Ditto.
3247 * doc/invoke.texi: Ditto.
3249 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3252 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3253 INTEGER_CST offset, add it together with bitpos / 8 and
3254 sign extend based on POINTER_SIZE.
3258 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3261 * config/i386/constraints.md (Yc): New register constraint.
3262 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3263 Use Yc constraint for alternative 2 of operand 0. Remove
3264 preferred_for_speed attribute.
3266 2018-03-14 Richard Biener <rguenther@suse.de>
3268 PR tree-optimization/84830
3269 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3270 with the old one to avoid oscillations.
3272 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3275 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3277 (assign_by_spills): Return a flag of reload assignment failure.
3278 Do not process the reload assignment failures. Do not spill other
3279 reload pseudos if they has the same reg class. Update n if
3281 (lra_assign): Add a return arg. Set up from the result of
3282 assign_by_spills call.
3283 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3284 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3285 usage_insns if it is not NULL.
3286 (spill_hard_reg_in_range): New function.
3287 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3288 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3289 function prototypes.
3290 (lra_assign): Change prototype.
3291 * lra.c (lra): Add code to deal with fails by splitting hard reg
3294 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3296 * config/riscv/riscv.opt (mrelax): New option.
3297 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3298 "norelax" when riscv_mrelax is disabled.
3299 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3301 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3304 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3305 reassociation for int modes.
3307 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3309 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3310 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3312 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3313 * config/aarch64/aarch64-sve.md
3314 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3315 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3316 (*extend<mode><Vwide>2): Rename to...
3317 (aarch64_sve_extend<mode><Vwide>2): ...this.
3318 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3319 renaming the old pattern to...
3320 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3322 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3323 define_expand, renaming the old pattern to...
3324 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3325 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3326 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3327 account when deciding which SVE instruction the optab should use.
3328 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3330 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3332 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3333 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3334 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3335 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3336 (tlsdesc_small_<mode>): Turn a define_expand and use
3337 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3338 (tlsdesc_small_advsimd_<mode>): ...this.
3339 (tlsdesc_small_sve_<mode>): New pattern.
3341 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3343 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3344 (UNSPEC_UMUL_HIGHPART): New constants.
3345 (MUL_HIGHPART): New int iteraor.
3346 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3347 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3349 (*<su>mul<mode>3_highpart): New define_insn.
3351 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3354 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3357 2018-03-13 Martin Liska <mliska@suse.cz>
3360 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3362 (sem_item_optimizer::~sem_item_optimizer): Release it.
3363 (sem_item_optimizer::merge_classes): Register variable aliases.
3364 (sem_item_optimizer::fixup_pt_set): New function.
3365 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3366 * ipa-icf.h: Declare new variables and functions.
3368 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3371 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3372 integer_pow2p@2 and test integer_pow2p in condition.
3373 (A < 0 ? C : 0): Similarly for @1.
3376 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3377 characters starting at p contain '\0' character, don't look beyond
3381 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3382 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3385 * reg-stack.c (change_stack): Change update_end var from int to
3386 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3387 also call set_block_for_insn on the newly added insns and rescan.
3390 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3391 on the last operand.
3394 * tree.c (stabilize_reference_1): Return save_expr (e) for
3395 STATEMENT_LIST even if it doesn't have side-effects.
3397 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3399 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3401 2018-03-12 Renlin Li <renlin.li@arm.com>
3403 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3404 aarch64_output_scalar_simd_mov_immediate.
3406 2018-03-12 Martin Sebor <msebor@redhat.com>
3408 PR tree-optimization/83456
3409 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3410 for perfectly overlapping calls to memcpy.
3411 (gimple_fold_builtin_memory_chk): Same.
3412 (gimple_fold_builtin_strcpy): Handle no-warning.
3413 (gimple_fold_builtin_stxcpy_chk): Same.
3414 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3416 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3418 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3419 parameter. Use it for SFmode.
3420 (rs6000_function_arg_advance_1): Adjust.
3421 (rs6000_function_arg): Adjust.
3422 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3424 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3426 PR rtl-optimization/84169
3427 PR rtl-optimization/84780
3428 * combine.c (can_combine_p): Check for a 2-insn combination whether
3429 the destination register is used between the two insns, too.
3431 2018-03-12 Richard Biener <rguenther@suse.de>
3433 PR tree-optimization/84803
3434 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3435 for refs DR analysis didn't process.
3437 2018-03-12 Richard Biener <rguenther@suse.de>
3439 PR tree-optimization/84777
3440 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3441 force-vectorize loops ignore whether we are optimizing for size.
3443 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3445 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3446 (TARGET_MD_ASM_ADJUST): Define.
3448 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3449 Kito Cheng <kito.cheng@gmail.com>
3450 Chung-Ju Wu <jasonwucj@gmail.com>
3452 * config/nds32/nds32.c (nds32_compute_stack_frame,
3453 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3454 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3455 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3456 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3457 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3458 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3459 * config/nds32/nds32.md (prologue, epilogue): Use macro
3460 NDS32_V3PUSH_AVAILABLE_P to do checking.
3462 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3465 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3466 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3467 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3468 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3469 addition of most attributes on !orig_type_die or the attribute not
3470 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3472 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3473 Chung-Ju Wu <jasonwucj@gmail.com>
3475 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3477 * config/nds32/nds32.opt (mvh): New option.
3479 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3480 Chung-Ju Wu <jasonwucj@gmail.com>
3482 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3484 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3485 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3488 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3489 Chung-Ju Wu <jasonwucj@gmail.com>
3491 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3493 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3494 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3496 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3497 Kito Cheng <kito.cheng@gmail.com>
3498 Chung-Ju Wu <jasonwucj@gmail.com>
3500 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3501 UNSPEC_FFMISM and UNSPEC_FLMISM.
3502 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3503 for ffb, ffmism and flmism.
3504 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3505 (unspec_ffmism): Ditto.
3506 (unspec_flmism): Ditto.
3507 (nds32_expand_builtin_impl): Check if string extension is available.
3508 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3509 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3511 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3514 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3517 * lra-assigns.c (assign_by_spills): Return a flag of reload
3518 assignment failure. Do not process the reload assignment
3519 failures. Do not spill other reload pseudos if they has the same
3521 (lra_assign): Add a return arg. Set up from the result of
3522 assign_by_spills call.
3523 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3524 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3525 usage_insns if it is not NULL.
3526 (spill_hard_reg_in_range): New function.
3527 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3528 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3529 function prototypes.
3530 (lra_assign): Change prototype.
3531 * lra.c (lra): Add code to deal with fails by splitting hard reg
3534 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3537 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3539 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3542 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3543 (dw_val_node): Add val_symbolic_view.
3544 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3545 (symview_upper_bound): New.
3546 (new_line_info_table): Initialize symviews_since_reset.
3547 (dwarf2out_source_line): Count symviews_since_reset and set
3548 symview_upper_bound.
3549 (dw_val_equal_p): Handle symview.
3550 (add_AT_symview): New.
3551 (print_dw_val): Handle symview.
3552 (attr_checksum, attr_checksum_ordered): Likewise.
3553 (same_dw_val_p, size_of_die): Likewise.
3554 (value_format, output_die): Likewise.
3555 (add_high_low_attributes): Use add_AT_symview for entry_view.
3556 (dwarf2out_finish): Reset symview_upper_bound, clear
3559 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3562 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3563 Add strict argument and use it.
3564 (rs6000_split_multireg_move): Update for new strict argument.
3565 (mem_operand_gpr): Disallow all non-offsettable addresses.
3566 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3568 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3571 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3572 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3573 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3576 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3577 decl, use remap_type if we want to use the type.
3579 2018-03-09 Martin Sebor <msebor@redhat.com>
3581 PR tree-optimization/84526
3582 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3584 (builtin_access::generic_overlap): Be prepared to handle non-array
3587 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3589 PR rtl-optimization/84682
3590 * lra-constraints.c (process_address_1): Check is_address flag
3591 for address constraints.
3592 (process_alt_operands): Likewise.
3593 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3594 preprocess_constraints.
3595 * recog.h (preprocess_constraints): Add oploc parameter.
3597 * recog.c (preprocess_constraints): Test address_operand for
3598 CT_ADDRESS constraints.
3600 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3603 * lra-assigns.c (assign_by_spills): Return a flag of reload
3604 assignment failure. Do not process the reload assignment
3605 failures. Do not spill other reload pseudos if they has the same
3607 (lra_assign): Add a return arg. Set up from the result of
3608 assign_by_spills call.
3609 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3610 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3611 usage_insns if it is not NULL.
3612 (spill_hard_reg_in_range): New function.
3613 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3614 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3615 function prototypes.
3616 (lra_assign): Change prototype.
3617 * lra.c (lra): Add code to deal with fails by splitting hard reg
3620 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3623 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3624 Accept complain bool parameter. Only emit errors if it is true.
3625 (arm_parse_cpu_option_name): Likewise.
3626 (arm_target_thumb_only): Adjust callers of the above.
3627 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3628 prototype to take a default true bool parameter.
3629 (arm_parse_arch_option_name): Likewise.
3631 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3632 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3636 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3637 * configure: Regenerate.
3638 * configure.ac ("linker --version-script option"): New.
3639 ("linker soname option"): New.
3641 2018-03-09 Richard Biener <rguenther@suse.de>
3643 PR tree-optimization/84775
3644 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3645 immediate uses of predicate stmts and mark them modified.
3648 PR tree-optimization/84178
3649 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3651 (version_loop_for_if_conversion): Delay update_ssa call.
3652 (tree_if_conversion): Delay update_ssa until after predicate
3655 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3658 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3659 when the function accesses prior frames.
3661 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3664 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3665 gen_llsym, otherwise call maybe_gen_llsym.
3668 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3669 has ',' character inside of it.
3671 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3674 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3675 as clobbering CC_REGNUM.
3677 2018-03-08 Richard Biener <rguenther@suse.de>
3680 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3681 (follow_copies_to_constant): Do not follow SSA names registered
3684 2018-03-08 Richard Biener <rguenther@suse.de>
3686 PR tree-optimization/84178
3687 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3689 (version_loop_for_if_conversion): Delay update_ssa call.
3690 (tree_if_conversion): Delay update_ssa until after predicate
3693 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3695 PR tree-optimization/84178
3696 * tree-if-conv.c (release_bb_predicate): Remove the
3697 the assertion that the stmts have NULL use_ops.
3698 Discard the statements, asserting that they haven't
3699 yet been added to a BB.
3701 2018-03-08 Richard Biener <rguenther@suse.de>
3703 PR tree-optimization/84746
3704 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3705 (phi_translate): Pass in destination ANTIC_OUT set.
3706 (phi_translate_1): Likewise. For a simplified result lookup
3707 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3708 (phi_translate_set): Adjust.
3709 (do_pre_regular_insertion): Likewise.
3710 (do_pre_partial_partial_insertion): Likewise.
3712 2018-03-08 Martin Liska <mliska@suse.cz>
3714 PR gcov-profile/84735
3715 * doc/gcov.texi: Document usage of profile files.
3716 * gcov-io.h: Document changes in the format.
3718 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3722 * dwarf2out.c (struct dw_line_info_table): Update comments for
3724 (FORCE_RESET_NEXT_VIEW): New.
3725 (FORCE_RESETTING_VIEW_P): New.
3726 (RESETTING_VIEW_P): Check for -1 too.
3727 (ZERO_VIEW_P): Likewise.
3728 (new_line_info_table): Force-reset next view.
3729 (dwarf2out_begin_function): Likewise.
3730 (dwarf2out_source_line): Simplify zero_view_p initialization.
3731 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3732 view directly. Omit view when omitting .loc at line 0.
3734 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3736 PR tree-optimization/84740
3737 * tree-switch-conversion.c (process_switch): Call build_constructors
3738 only if info.phi_count is non-zero.
3740 PR tree-optimization/84739
3741 * tree-tailcall.c (find_tail_calls): Check call arguments against
3742 DECL_ARGUMENTS (current_function_decl) rather than
3743 DECL_ARGUMENTS (func) when checking for tail recursion.
3745 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3747 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3748 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3749 Volker Reichelt's entry and add entries for people that perform
3750 GCC fuzzy testing and report numerous bugs.
3752 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3755 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3756 readonly data in sdata, if that is disabled.
3757 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3758 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3759 -mreadonly-in-sdata option.
3761 2018-03-07 Martin Sebor <msebor@redhat.com>
3763 PR tree-optimization/84468
3764 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3765 basic block when looking for nul assignment.
3767 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3770 * except.h (output_function_exception_table): Adjust prototype.
3771 * except.c (output_function_exception_table): Remove FNNAME parameter
3772 and add SECTION parameter. Ouput one part of the table at a time.
3773 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3774 the first part of the exception table and emit unwind directives.
3775 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3776 (i386_pe_seh_cold_init): Likewise.
3777 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3778 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3779 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3780 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3781 * config/i386/winnt.c: Include except.h.
3782 (struct seh_frame_state): Add reg_offset, after_prologue and
3783 in_cold_section fields.
3784 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3785 (i386_pe_seh_cold_init): New function.
3786 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3787 to seh->in_cold_section.
3788 (seh_emit_push): Record the offset of the push.
3789 (seh_emit_save): Record the offet of the save.
3790 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3791 Test seh->after_prologue to disregard the epilogue.
3792 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3793 (i386_pe_end_cold_function): New function.
3795 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3798 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3799 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3802 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3804 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3805 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3808 * multiple_target.c: Include tree-inline.h and intl.h.
3809 (expand_target_clones): Diagnose and fail if node->definition and
3810 !tree_versionable_function_p (node->decl).
3812 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3814 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3816 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3817 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3818 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3820 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3823 * combine.c (try_combine): Use reg_or_subregno instead of handling
3824 just paradoxical SUBREGs and REGs.
3826 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3828 * config/arc/arc.c (arc_finalize_pic): Remove function.
3829 (arc_must_save_register): We use single base PIC register, remove
3830 checks to save/restore the PIC register.
3831 (arc_expand_prologue): Likewise.
3832 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3834 (arc_verify_short): Likewise.
3835 (arc_attr_type): Likewise.
3836 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3837 (walk_stores): Likewise.
3838 (arc_address_cost): Make it static.
3839 (arc_verify_short): Likewise.
3840 (branch_dest): Likewise.
3841 (arc_attr_type): Likewise.
3842 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3843 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3844 (arc_final_prescan_insn): Remove inserting the nops due to
3845 hardware hazards. It is done in reorg step.
3846 (insn_length_variant_t): Remove.
3847 (insn_length_parameters_t): Likewise.
3848 (arc_insn_length_parameters): Likewise.
3849 (arc_get_insn_variants): Likewise.
3850 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3852 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3855 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3858 PR tree-optimization/84687
3859 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3861 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3863 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3865 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3866 Rename to ppc_speculation_barrier.
3867 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3868 __builtin_ppc_speculation_barrier.
3870 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3873 * combine.c (combine_simplify_rtx): Don't try to simplify if
3874 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3877 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3879 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3880 to 32 bytes when compiling for POWER9.
3882 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3885 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3886 regparm >= 3 with no arg reg available also for calls with
3887 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3890 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3892 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3894 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3897 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3899 2018-03-05 Richard Biener <rguenther@suse.de>
3901 PR tree-optimization/84486
3902 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3903 When inserting a __builtin_assume_aligned call set the LHS
3904 SSA name alignment info accordingly.
3906 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3908 PR tree-optimization/84114
3909 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3910 Avoid reassociation of FLOAT_MODE addition.
3912 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3914 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3915 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3916 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3917 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3918 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3919 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3920 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3922 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3923 __builtin_ia32_wbinvd): New builtins.
3924 (SPECIAL_ARGS2): New.
3925 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3926 (SPECIAL_ARGS2): New.
3927 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3928 (ix86_valid_target_attribute_inner_p): Ditto.
3929 (ix86_init_mmx_sse_builtins): Add special_args2.
3930 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3931 TARGET_WBNOINVD_P): New.
3932 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3933 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3934 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3935 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3936 * config/i386/pconfigintrin.h: New file.
3937 * config/i386/wbnoinvdintrin.h: Ditto.
3938 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3939 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3941 2018-03-05 Richard Biener <rguenther@suse.de>
3943 PR tree-optimization/84670
3944 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3946 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3947 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3948 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3949 doesn't grow if all successors (recursively) were visited at least
3952 2018-03-05 Richard Biener <rguenther@suse.de>
3954 PR tree-optimization/84650
3955 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3956 if executed in the loop pipeline.
3958 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3960 * doc/configfiles.texi (Configuration Files): Move info about
3961 conditionalizing $target-protos.h to...
3962 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3963 differs from $target-protos.h.
3965 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3966 Chung-Ju Wu <jasonwucj@gmail.com>
3968 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3969 * config/nds32/nds32-multiple.md (setmemsi): Define.
3970 * config/nds32/nds32-memory-manipulation.c
3971 (nds32_gen_dup_4_byte_to_word_value): New.
3972 (emit_setmem_word_loop): New.
3973 (emit_setmem_byte_loop): New.
3974 (nds32_expand_setmem_loop): New.
3975 (nds32_expand_setmem_loop_v3m): New.
3976 (nds32_expand_setmem_unroll): New.
3977 (nds32_expand_setmem): New.
3979 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3980 Chung-Ju Wu <jasonwucj@gmail.com>
3982 * config/nds32/nds32-memory-manipulation.c
3983 (nds32_emit_load_store): New.
3984 (nds32_emit_post_inc_load_store): New.
3985 (nds32_emit_mem_move): New.
3986 (nds32_emit_mem_move_block): New.
3987 (nds32_expand_movmemsi_loop_unknown_size): New.
3988 (nds32_expand_movmemsi_loop_known_size): New.
3989 (nds32_expand_movmemsi_loop): New.
3990 (nds32_expand_movmemsi_unroll): New.
3991 (nds32_expand_movmemqi): Rename ...
3992 (nds32_expand_movmemsi): ... to this.
3993 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3994 (movmemsi): ... to this.
3995 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3996 (nds32_expand_movmemsi): ... to this.
3998 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3999 Monk Chiang <sh.chiang04@gmail.com>
4000 Chung-Ju Wu <jasonwucj@gmail.com>
4002 * config/nds32/nds32-protos.h
4003 (nds32_expand_load_multiple): New arguments.
4004 (nds32_expand_store_multiple): Ditto.
4005 (nds32_valid_multiple_load_store): Rename ...
4006 (nds32_valid_multiple_load_store_p): ... to this.
4007 * config/nds32/nds32-memory-manipulation.c
4008 (nds32_expand_load_multiple): Refine implementation.
4009 (nds32_expand_store_multiple): Ditto.
4010 * config/nds32/nds32-multiple.md
4011 (load_multiple): Update nds32_expand_load_multiple interface.
4012 (store_multiple): Update nds32_expand_store_multiple interface.
4013 * config/nds32/nds32-predicates.c
4014 (nds32_valid_multiple_load_store): Rename ...
4015 (nds32_valid_multiple_load_store_p): ... to this and refine
4017 * config/nds32/predicates.md
4018 (nds32_load_multiple_and_update_address_operation): New predicate.
4019 (nds32_store_multiple_and_update_address_operation): New predicate.
4021 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4022 Chung-Ju Wu <jasonwucj@gmail.com>
4024 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4025 (combo): New attribute.
4026 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4028 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4030 * config/nds32/nds32.opt: Change -mcmodel= default value.
4032 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4033 Monk Chiang <sh.chiang04@gmail.com>
4034 Chung-Ju Wu <jasonwucj@gmail.com>
4036 * config/nds32/constants.md (unspec_element): New enum.
4037 * config/nds32/constraints.md (Umw): New constraint.
4038 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4039 * config/nds32/nds32-intrinsic.md: Likewise.
4040 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4041 (nds32_valid_smw_lwm_base_p): New.
4042 (nds32_output_smw_single_word): New.
4043 (nds32_output_lmw_single_word): New.
4044 (nds32_expand_unaligned_load): New.
4045 (nds32_expand_unaligned_store): New.
4046 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4047 (nds32_output_smw_single_word): Declare.
4048 (nds32_output_lmw_single_word): Declare.
4049 (nds32_expand_unaligned_load): Declare.
4050 (nds32_expand_unaligned_store): Declare.
4051 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4052 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4053 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4054 NDS32_BUILTIN_UASTORE_DW.
4055 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4058 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4059 Kito Cheng <kito.cheng@gmail.com>
4060 Chung-Ju Wu <jasonwucj@gmail.com>
4062 * config/nds32/nds32-intrinsic.c
4063 (nds32_expand_builtin_null_ftype_reg): Delete.
4064 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4065 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4066 (nds32_read_argument): New.
4067 (nds32_legitimize_target): Ditto.
4068 (nds32_legitimize_argument): Ditto.
4069 (nds32_check_constant_argument): Ditto.
4070 (nds32_expand_unop_builtin): Ditto.
4071 (nds32_expand_unopimm_builtin): Ditto.
4072 (nds32_expand_binop_builtin): Ditto.
4073 (nds32_builtin_decl_impl): Ditto.
4074 (builtin_description): Ditto.
4075 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4076 (nds32_init_builtins_impl): Ditto.
4077 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4078 (nds32_builtin_decl): New.
4079 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4080 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4082 2018-03-02 Jeff Law <law@redhat.com>
4084 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4085 (redundant_insn, fill_simple_delay_slots): Likewise.
4086 (fill_slots_from_thread): Likewise.
4087 * resource.c (mark_referenced_resources): Likewise.
4088 (mark_set_resources, find_dead_or_set_registers): Likewise.
4090 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4092 * substring-locations.h (format_warning_va): Formatting fix for
4094 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4096 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4097 * substring-locations.c: Include intl.h.
4098 (format_warning_va): Turned into small wrapper around
4099 format_warning_n_va, renamed to ...
4100 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4101 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4103 (format_warning_at_substring_n): New function.
4104 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4105 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4106 format_warning_at_substring with just a shorter name instead of
4107 const function pointer.
4108 (fmtwarn_n): New function.
4109 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4110 appropriate, get rid of all the fmtstr temporaries, move conditionals
4111 with G_() wrapped string literals directly into fmtwarn arguments,
4112 cast dir.len to (int), formatting fixes.
4114 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4116 * doc/invoke.texi: Remove "Cilk Plus" references.
4118 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4119 Richard Biener <rguenther@suse.de>
4122 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4123 for error or warning attributes if CALL_FROM_THUNK_P is set.
4126 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4129 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4130 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4133 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4134 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4136 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4137 instead of -1U in last predictors element's probability member.
4139 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4142 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4143 arguments if they are comparable.
4145 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4147 PR tree-optimization/84634
4148 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4149 masks and masked_loop_p with a single loop_masks, making sure it's
4150 null for bb vectorization.
4152 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4154 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4155 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4156 loop->force_vectorize to check whether there is no alias.
4158 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4161 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4163 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4165 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4166 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4167 instead of a loop around prev_real_insn.
4168 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4172 * config/i386/i386.c (ix86_print_operand): Use conditional
4173 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4176 2018-03-02 Richard Biener <rguenther@suse.de>
4178 PR tree-optimization/84427
4179 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4180 (bitmap_set_subtract_values): Rewrite to handle multiple
4183 (prune_clobbered_mems): Likewise.
4184 (phi_translate): Take edge instead of pred/phiblock.
4185 (phi_translate_1): Likewise.
4186 (phi_translate_set): Likewise. Insert all translated
4187 exprs for a value into the set, keeping possibly multiple
4188 expressions per value.
4189 (compute_antic_aux): Adjust for phi_translate changes.
4190 When intersecting union the expressions and prune those
4191 not in the final value set, keeping possibly multiple
4192 expressions per value. Do not use value-insertion
4193 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4194 all expressions. Add verification that the value-sets
4195 only shrink during iteration.
4196 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4197 (do_pre_regular_insertion): Likewise.
4198 (do_pre_partial_partial_insertion): Likewise.
4200 2018-03-02 Richard Biener <rguenther@suse.de>
4203 * config/darwin.c (saved_debug_info_level): New static global.
4204 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4205 (darwin_asm_lto_end): Restore debug info generation settings.
4207 2018-03-01 Martin Liska <mliska@suse.cz>
4210 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4213 2018-03-01 Richard Biener <rguenther@suse.de>
4216 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4217 type annotation with LTO when debug was disabled at compile-time.
4219 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4221 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4223 (mips_final_postscan_insn): Likewise.
4225 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4227 PR rtl-optimization/84528
4228 * alias.c (init_alias_target): Add commentary.
4229 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4230 a unique base value if the frame pointer is not eliminated
4231 to the stack pointer.
4233 2018-03-01 Tom de Vries <tom@codesourcery.com>
4235 PR rtl-optimization/83327
4236 * lra-int.h (hard_regs_spilled_into): Declare.
4237 * lra.c (hard_regs_spilled_into): Define.
4238 (init_reg_info): Init hard_regs_spilled_into.
4239 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4240 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4241 (process_bb_lives): Handle hard_regs_spilled_into.
4242 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4243 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4245 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4247 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4248 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4249 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4250 * config/rs6000/aix72.h: New file.
4252 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4254 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4255 instead of warning_at with conditional singular and plural messages
4259 * stor-layout.c (update_alignment_for_field): For
4260 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4261 && !DECL_PACKED (field), do the alignment update, just use
4262 only desired_align instead of MAX (type_align, desired_align)
4264 (place_field): Don't do known_align < desired_align handling
4265 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4266 is non-NULL, instead do it after rli->prev_field handling and
4267 only if not within a bitfield word. For DECL_PACKED (field)
4268 use type_align of BITS_PER_UNIT.
4270 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4272 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4273 superfluous parentheses and trailing spaces.
4275 2018-02-28 Richard Biener <rguenther@suse.de>
4277 PR tree-optimization/84584
4278 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4279 SCoPs with fake exit edge.
4281 2018-02-28 Martin Liska <mliska@suse.cz>
4284 * timevar.c (timer::print): Fix format to properly print 100%
4287 2018-02-28 Richard Biener <rguenther@suse.de>
4290 * genmatch.c (capture_info::walk_match): Do not mark
4291 captured expressions without operands as expr_p given
4292 they act more like predicates and should be subject to
4293 "lost tail" side-effect preserving.
4295 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4297 PR rtl-optimization/81611
4298 * auto-inc-dec.c (attempt_change): Move dead note from
4299 mem_insn if it's the next use of regno
4300 (find_address): Take address use of reg holding
4301 non-incremented value. Add parm to limit search to the named
4303 (merge_in_block): Attempt to use a mem insn that is the next
4304 use of the original regno.
4306 2018-02-27 Martin Sebor <msebor@redhat.com>
4309 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4310 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4312 2018-02-27 Martin Sebor <msebor@redhat.com>
4314 PR translation/84207
4315 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4316 n argument to unsigned HOST_WIDE_INT.
4317 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4318 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4319 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4320 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4322 2018-02-27 Richard Biener <rguenther@suse.de>
4324 PR tree-optimization/84512
4325 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4326 Do not use the estimate returned from record_stmt_cost for
4327 the scalar iteration cost but sum properly using add_stmt_cost.
4329 2018-02-27 Richard Biener <rguenther@suse.de>
4331 PR tree-optimization/84466
4332 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4333 Adjust last change to less strictly validate use operands.
4335 2018-02-27 Martin Liska <mliska@suse.cz>
4337 PR gcov-profile/84548
4338 * gcov.c (process_file): Allow partial overlap and consider it
4339 also as group functions.
4340 (output_lines): Properly calculate range of lines for a group.
4342 2018-02-27 Martin Liska <mliska@suse.cz>
4344 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4345 'ggc' suffixes. Change first column width.
4346 (timer::print): Fix formatting of the column.
4348 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4350 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4351 preserve inline entry blocks for the sake of debug inline
4352 entry point markers alone.
4353 (remove_unused_locals): Suggest in comments a better place to
4354 force the preservation of inline entry blocks that are
4355 otherwise unused, but do not preserve them.
4357 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4359 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4361 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4364 * config/i386/constraints.md (Bs): Replace
4365 ix86_indirect_branch_register with
4366 TARGET_INDIRECT_BRANCH_REGISTER.
4368 * config/i386/i386.md (indirect_jump): Likewise.
4369 (tablejump): Likewise.
4370 (*sibcall_memory): Likewise.
4371 (*sibcall_value_memory): Likewise.
4372 Peepholes of indirect call and jump via memory: Likewise.
4373 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4374 (*sibcall_value_GOT_32): Likewise.
4375 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4376 (GOT_memory_operand): Likewise.
4377 (call_insn_operand): Likewise.
4378 (sibcall_insn_operand): Likewise.
4379 (GOT32_symbol_operand): Likewise.
4380 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4382 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4384 PR rtl-optimization/83496
4385 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4386 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4387 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4388 redundant insn, if any.
4389 (relax_delay_slots): Likewise.
4390 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4392 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4394 PR tree-optimization/83965
4395 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4396 that grouped statements are part of a reduction chain. Return
4397 true if the statement is not marked as a reduction itself but
4399 (vect_recog_dot_prod_pattern): Don't check whether the statement
4400 is part of a group here.
4401 (vect_recog_sad_pattern): Likewise.
4402 (vect_recog_widen_sum_pattern): Likewise.
4404 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4407 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4409 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4412 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4414 (ix86_output_indirect_function_return): New prototype.
4415 (ix86_split_simple_return_pop_internal): Likewise.
4416 * config/i386/i386.c (indirect_return_via_cx): New.
4417 (indirect_return_via_cx_bnd): Likewise.
4418 (indirect_thunk_name): Handle return va CX_REG.
4419 (output_indirect_thunk_function): Create alias for
4420 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4421 (ix86_output_indirect_jmp): Remove the bool argument.
4422 (ix86_output_indirect_function_return): New function.
4423 (ix86_split_simple_return_pop_internal): Likewise.
4424 * config/i386/i386.md (*indirect_jump): Don't pass false
4425 to ix86_output_indirect_jmp.
4426 (*tablejump_1): Likewise.
4427 (simple_return_pop_internal): Change it to define_insn_and_split.
4428 Call ix86_split_simple_return_pop_internal to split it for
4430 (simple_return_indirect_internal): Call
4431 ix86_output_indirect_function_return instead of
4432 ix86_output_indirect_jmp.
4434 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4437 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4438 memset and value initialization afterwards.
4440 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4442 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4444 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4447 * common/config/aarch64/aarch64-common.c
4448 (aarch_option_optimization_table[]): Switch
4449 off fomit-frame-pointer
4451 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4452 Chung-Ju Wu <jasonwucj@gmail.com>
4454 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4456 (store_multiple): Ditto.
4458 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4460 * config.gcc: Add --with-cpu support for nds32 target.
4461 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4462 * config/nds32/nds32.opt: Add -mcpu= option.
4464 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4466 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4467 isel=yes): Warn for these deprecated options.
4469 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4471 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4472 ISA_2_5_MASKS_EMBEDDED.
4474 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4476 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4477 p->max as pointers rather than using iterative_hash_expr.
4479 2018-02-23 Carl Love <cel@us.ibm.com>
4481 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4482 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4484 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4485 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4486 P8V_BUILTIN_VEC_VUNSIGNED2.
4488 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4491 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4492 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4494 * lra-constraints.c (curr_insn_transform): Set up
4495 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4497 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4498 * lra-eliminations.c (spill_pseudos): Ditto.
4499 (process_insn_for_elimination): Ditto.
4500 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4501 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4502 LRA_NON_CLOBBERED_ALT.
4504 2018-02-22 Martin Sebor <msebor@redhat.com>
4506 PR tree-optimization/84480
4507 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4508 to maybe_diag_stxncpy_trunc. Call it.
4509 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4510 from gimple_fold_builtin_strcpy. Print inlining stack.
4511 (handle_builtin_stxncpy): Print inlining stack.
4512 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4514 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4517 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4518 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4519 and -fcheck-pointer-bounds are used together.
4520 (indirect_thunk_prefix): New enum.
4521 (indirect_thunk_need_prefix): New function.
4522 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4523 "_nt" instead of "_bnd" for NOTRACK prefix.
4524 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4525 (output_indirect_thunk_function): Likewise.
4527 (ix86_code_end): Update output_indirect_thunk_function calls.
4528 (ix86_output_indirect_branch_via_reg): Replace
4529 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4530 (ix86_output_indirect_branch_via_push): Likewise.
4531 (ix86_output_function_return): Likewise.
4532 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4533 incompatible with -fcf-protection=branch and
4534 -fcheck-pointer-bounds.
4536 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4539 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4540 Change gcc_assert call to output_operand_lossage.
4542 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4544 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4546 2018-02-22 DJ Delorie <dj@redhat.com>
4547 Sebastian Perta <sebastian.perta@renesas.com>
4548 Oleg Endo <olegendo@gcc.gnu.org>
4550 * config/rx/rx.c (rx_rtx_costs): New function.
4551 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4553 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4555 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4557 2018-02-22 Martin Liska <mliska@suse.cz>
4560 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4561 Add "native" as a possible value.
4563 2018-02-22 Martin Liska <mliska@suse.cz>
4566 * config/i386/i386.c (ix86_option_override_internal):
4567 Add "native" as a possible value for -march and -mtune.
4569 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4572 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4573 to all type variants.
4575 PR tree-optimization/84503
4576 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4577 width as info->bitpos + info->bitsize - start.
4578 (merged_store_group::merge_overlapping): Simplify width computation.
4579 (check_no_overlap): New function.
4580 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4581 start + width and last_order of the group, fail if check_no_overlap
4583 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4584 to group if check_no_overlap fails.
4586 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4588 * config/rs6000/altivec.md: Delete contraint arguments to
4589 define_expand, define_split, and define_peephole2, and in
4590 define_insn_and_split if always unused.
4591 * config/rs6000/darwin.md: Ditto.
4592 * config/rs6000/dfp.md: Ditto.
4593 * config/rs6000/rs6000.md: Ditto.
4594 * config/rs6000/sync.md: Ditto.
4595 * config/rs6000/vector.md: Ditto.
4596 * config/rs6000/vsx.md: Ditto.
4598 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4600 * config/rs6000/altivec.md: Write output control strings as braced
4601 blocks instead of double-quoted strings.
4602 * config/rs6000/darwin.md: Ditto.
4603 * config/rs6000/rs6000.md: Ditto.
4604 * config/rs6000/vector.md: Ditto.
4605 * config/rs6000/vsx.md: Ditto.
4607 2018-02-21 Jason Merrill <jason@redhat.com>
4609 PR c++/84314 - ICE with templates and fastcall attribute.
4610 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4612 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4614 * ipa-cp.c (determine_versionability): Fix comment typos.
4616 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4619 * ipa-cp.c (determine_versionability): Do not version functions caling
4622 2018-02-21 Martin Liska <mliska@suse.cz>
4625 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4626 Add "native" as a possible value.
4627 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4628 the macro when native cpu detection is available.
4630 2018-02-21 Martin Liska <mliska@suse.cz>
4633 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4634 Add "native" as a possible value.
4635 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4636 when native cpu detection is available.
4638 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4639 Martin Sebor <msebor@redhat.com>
4641 PR tree-optimization/84478
4642 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4644 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4645 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4646 support which is conservatively correct, for 2 only stay conservative
4647 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4648 argument to the 2 argument get_range_strlen, adjust 6 arg
4649 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4651 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4652 (gimple_fold_builtin_strlen): Pass true as last argument to
4655 2018-02-20 Martin Sebor <msebor@redhat.com>
4658 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4659 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4660 (builtin_memref::builtin_memref): Factor out parts into
4661 set_base_and_offset and call it.
4663 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4666 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4667 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4668 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4669 search at the associated MODE_INT.
4671 2018-02-20 Jeff Law <law@redhat.com>
4674 PR tree-optimization/81592
4676 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4677 for range data rather than using global data.
4678 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4679 range data rather than using global data.
4680 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4681 pass it to children as needed.
4682 (struct directive::fmtresult): Similarly.
4683 (struct directive::set_width): Similarly.
4684 (struct directive::set_precision): Similarly.
4685 (format_integer, format_directive, parse_directive): Similarly.
4686 (format_none): Accept unnamed vr_values parameter.
4687 (format_percent, format_floating, format_character): Similarly.
4688 (format_string, format_plain): Similarly.
4689 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4690 the EVRP range analyzer for range data rather than using global data.
4691 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4692 gimple-ssa-evrp-analyze.h
4693 (class sprintf_dom_walker): Add after_dom_children member function.
4694 Add evrp_range_analyzer member.
4695 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4696 range analyzer as needed.
4697 (sprintf_dom_walker::after_dom_children): New member function.
4698 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4700 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4701 (evrp_range_analyzer::pop_to_marker): Likewise.
4703 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4705 PR tree-optimization/84419
4706 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4707 with the required type if its current type is compatible but
4710 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4713 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4714 after vectorization.
4716 2018-02-20 Martin Liska <mliska@suse.cz>
4719 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4720 possible values if we don't have a hint.
4722 2018-02-20 Martin Liska <mliska@suse.cz>
4726 * final.c (shorten_branches): Build align_tab array with one
4728 * opts.c (finish_options): Add alignment option limit check.
4729 (MAX_CODE_ALIGN): Likewise.
4730 (MAX_CODE_ALIGN_VALUE): Likewise.
4731 * doc/invoke.texi: Document maximum allowed option value for
4732 all -falign-* options.
4734 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4737 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4738 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4739 * var-tracking.c (emit_note_insn_var_location): Remove all references
4740 to NOTE_INSN_CALL_ARG_LOCATION.
4741 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4742 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4743 Use copy_rtx_if_shared.
4744 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4745 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4746 (dwarf2out_var_location): Remove handling of
4747 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4749 * final.c (final_scan_insn): Remove all references to
4750 NOTE_INSN_CALL_ARG_LOCATION.
4751 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4752 before dumping final insns.
4753 * except.c (emit_note_eh_region_end): Remove all references to
4754 NOTE_INSN_CALL_ARG_LOCATION.
4755 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4756 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4757 * config/arc/arc.c (hwloop_optimize): Likewise.
4758 * config/arm/arm.c (create_fix_barrier): Likewise.
4759 * config/s390/s390.c (s390_chunkify_start): Likewise.
4760 * config/sh/sh.c (find_barrier): Likewise.
4761 * config/i386/i386.c (rest_of_insert_endbranch,
4762 ix86_seh_fixup_eh_fallthru): Likewise.
4763 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4764 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4765 * config/frv/frv.c (frv_function_prologue): Likewise.
4766 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4768 (note_outside_basic_block_p): Remove all references to
4769 NOTE_INSN_CALL_ARG_LOCATION.
4770 * gengtype.c (adjust_field_rtx_def): Likewise.
4771 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4773 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4774 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4777 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4780 PR tree-optimization/84452
4781 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4782 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4785 2018-02-19 Martin Liska <mliska@suse.cz>
4788 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4790 2018-02-19 Martin Liska <mliska@suse.cz>
4791 Richard Sandiford <richard.sandiford@linaro.org>
4793 PR tree-optimization/82491
4794 * gimple-fold.c (get_base_constructor): Make earlier bail out
4797 2018-02-19 Carl Love <cel@us.ibm.com>
4799 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4800 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4802 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4803 P8V_BUILTIN_VEC_NEG.
4805 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4807 * config/rl78/rl78.md (movdf): New define expand.
4809 2018-02-19 Martin Liska <mliska@suse.cz>
4812 * doc/invoke.texi: Fix typo.
4813 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4815 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4817 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4818 handle rs6000_single_float and rs6000_double_float specially for
4821 2018-02-16 Jeff Law <law@redhat.com>
4823 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4826 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4829 * ipa-inline.c (inline_small_functions): Fix a typo.
4831 2018-02-16 Nathan Sidwell <nathan@acm.org>
4833 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4835 2018-02-16 Carl Love <cel@us.ibm.com>
4837 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4838 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4839 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4840 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4841 expansion to P8V_BUILTIN_VEC_FLOAT2.
4843 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4845 PR rtl-optimization/70023
4846 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4847 src_regno into account.
4849 2018-02-16 Carl Love <cel@us.ibm.com>
4851 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4852 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4853 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4854 * config/rs6000/rs6000.c: Remove case statements for
4855 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4856 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4857 and P9V_BUILTIN_VEC_VINSERT4B.
4858 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4859 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4860 * config/rs6000/vsx.md:
4861 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4864 2018-02-16 Carl Love <cel@us.ibm.com>
4866 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4868 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4870 * config/rs6000/rs6000-c.c: Add the definitions for
4871 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4872 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4873 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4874 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4875 definition for insert4b and define insn *insert3b_internal.
4876 * doc/extend.texi: Add documentation for vec_extract4b.
4878 2018-02-16 Nathan Sidwell <nathan@acm.org>
4880 * doc/extend.texi (Backwards Compatibility): Mention friend
4881 injection. Note for-scope is deprecated.
4882 * doc/invoke.texi (-ffriend-injection): Deprecate.
4884 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4886 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4887 that moved to I2, also allow destinations that are a paradoxical
4888 subreg (instead of a normal reg).
4890 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4893 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4896 2018-02-16 Richard Biener <rguenther@suse.de>
4898 PR tree-optimization/84037
4899 PR tree-optimization/84016
4901 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4902 Adjust vec_construct for the fact we need additional higher latency
4903 128bit inserts for AVX256 and AVX512 vector builds.
4904 (ix86_add_stmt_cost): Scale vector construction cost for
4907 2018-02-16 Richard Biener <rguenther@suse.de>
4909 PR tree-optimization/84417
4910 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4911 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4912 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4914 2018-02-16 Martin Liska <mliska@suse.cz>
4917 * internal-fn.def (ASAN_CHECK): Set proper flags.
4918 (ASAN_MARK): Likewise.
4920 2018-02-16 Julia Koval <julia.koval@intel.com>
4922 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4923 from PTA_CANNONLAKE.
4925 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4928 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4929 Use ++iter rather than iter++ for std::list iterators.
4930 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4931 defer deleting them until all nodes in the forest are processed. Do
4932 free even leaf nodes. Change to_process into auto_vec.
4935 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4936 * vec.h (vec_default_construct): Use memset instead of placement new
4937 if BROKEN_VALUE_INITIALIZATION is defined.
4938 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4939 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4942 PR rtl-optimization/83723
4943 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4944 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4945 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4947 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4949 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4951 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4953 PR rtl-optimization/81443
4954 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4955 from inner REGs to paradoxical SUBREGs.
4957 2018-02-16 Richard Biener <rguenther@suse.de>
4959 PR tree-optimization/84399
4960 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4961 For operands we can analyze at their definition make sure we can
4962 analyze them at each use as well.
4964 2018-02-16 Richard Biener <rguenther@suse.de>
4966 PR tree-optimization/84190
4967 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4968 volatile accesses if the decl isn't volatile.
4970 2018-02-15 Jason Merrill <jason@redhat.com>
4972 PR c++/84314 - ICE with templates and fastcall attribute.
4973 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4974 TYPE_CANONICAL on an existing type.
4976 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4978 PR tree-optimization/84383
4979 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4980 dstoff nor call operand_equal_p if dstbase is NULL.
4982 PR tree-optimization/84334
4983 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4984 also a CONSTANT_CLASS_P, punt.
4986 2018-02-14 Jim Wilson <jimw@sifive.com>
4988 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4989 first SMALL_OPERAND check. New local min_second_step. Move assert
4990 to where locals are set. Add TARGET_RVC support.
4991 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4993 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4995 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4997 2018-02-14 Martin Sebor <msebor@redhat.com>
4999 PR tree-optimization/83698
5000 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5001 arrays constrain the offset range to their bounds.
5002 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5003 (builtin_access::overlap): Avoid setting the size of overlap if it's
5005 (maybe_diag_overlap): Also consider arrays when deciding what values
5006 of offsets to include in diagnostics.
5008 2018-02-14 Martin Sebor <msebor@redhat.com>
5011 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5012 that correspond to the kind of a declaration.
5014 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5017 * config/pa/pa.md: Load address of PIC label using the linkage table
5018 if the label is nonlocal.
5020 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5022 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5023 warning message if user requests -maltivec=be.
5024 * doc/invoke.texi: Document deprecation of -maltivec=be.
5026 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5029 * config/rs6000/rs6000-c.c: Update definitions for
5030 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5031 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5033 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5036 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5037 add _get_ssp intrinsics. Remove argument from
5038 __builtin_ia32_rdssp[d|q].
5039 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5040 * config/i386/i386-builtin.def: Remove argument from
5041 __builtin_ia32_rdssp[d|q].
5042 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5043 ix86_expand_special_args_builtin for _rdssp[d|q].
5044 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5045 Clear register before usage.
5046 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5047 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5049 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5051 PR tree-optimization/84357
5052 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5053 operand 1 of an ARRAY_REF too.
5055 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5058 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5059 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5061 (set_of_reg): New struct.
5062 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5063 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5064 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5066 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5067 Split into bitclr, bitset, bitinvert patterns if appropriate.
5068 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5069 use rx_fuse_in_memory_bitop.
5070 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5071 to named insn, correct maximum insn length.
5073 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5076 * machmode.def: Define a complex mode for PARTIAL_INT.
5077 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5079 * doc/rtl.texi: Document CSPImode.
5080 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5082 (msp430_hard_regno_nregs_with_padding): Likewise.
5084 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5087 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5089 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5091 PR rtl-optimization/84169
5092 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5093 we generated a parallel as new i3 and we split that to new i2 and i3
5094 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5095 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5096 those to i2, not i1. Partially rewrite this scan code.
5098 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5101 * stor-layout.c (place_field): For variable length fields, adjust
5102 offset_align afterwards not just based on the field's alignment,
5103 but also on the size.
5106 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5107 of exps and logs in the use_exp2 case.
5109 2018-02-13 Jeff Law <law@redhat.com>
5111 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5114 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5117 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5121 * common.opt (gas-loc-support, gas-locview-support): New.
5122 (ginline-points, ginternal-reset-location-views): New.
5123 * doc/invoke.texi: Document them. Use @itemx where intended.
5124 (gvariable-location-views): Adjust.
5125 * target.def (reset_location_view): New.
5126 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5127 (TARGET_RESET_LOCATION_VIEW): New.
5128 * doc/tm.texi: Rebuilt.
5129 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5130 (dwarf2out_default_as_locview_support): New.
5131 (output_asm_line_debug_info): Use option variables.
5132 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5133 (output_loc_list): Likewise.
5134 (add_high_low_attributes): Check option variables.
5135 Don't output entry view attribute in strict mode.
5136 (gen_inlined_subroutine_die): Check option variables.
5137 (dwarf2out_inline_entry): Likewise.
5138 (init_sections_and_labels): Likewise.
5139 (dwarf2out_early_finish): Likewise.
5140 (maybe_reset_location_view): New, from...
5141 (dwarf2out_var_location): ... here. Call it.
5142 * debug.h (dwarf2out_default_as_loc_support): Declare.
5143 (dwarf2out_default_as_locview_support): Declare.
5144 * hooks.c (hook_int_rtx_insn_0): New.
5145 * hooks.h (hook_int_rtx_insn_0): Declare.
5146 * toplev.c (process_options): Take -gas-loc-support and
5147 -gas-locview-support from dwarf2out. Enable
5148 -gvariable-location-views by default only with locview
5149 assembler support. Enable -ginternal-reset-location-views by
5150 default only if the target defines the corresponding hook.
5151 Enable -ginline-points by default if location views are
5152 enabled; force it disabled if statement frontiers are
5154 * tree-inline.c (expand_call_inline): Check option variables.
5155 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5157 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5159 PR tree-optimization/84321
5160 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5161 handling. Also check whether the anti-range contains any values
5162 that satisfy the mask; switch to a VR_RANGE if not.
5164 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5167 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5169 2018-02-13 Martin Jambor <mjambor@suse.cz>
5172 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5173 of call statements, also set location of a load to a temporary.
5175 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5177 * config/rl78/rl78.c (add_vector_labels): New function.
5178 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5179 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5180 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5181 which checks that no arguments are passed.
5182 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5183 * doc/extend.texi: Documentation for the new attribute.
5185 2018-02-13 Andreas Schwab <schwab@suse.de>
5187 * config/riscv/linux.h (CPP_SPEC): Define.
5189 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5192 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5193 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5194 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5195 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5196 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5197 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5198 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5199 temporarily for AES and PCLMUL builtins.
5201 PR tree-optimization/84339
5202 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5203 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5207 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5208 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5209 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5211 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5213 * omp-simd-clone.h: New file.
5214 * omp-simd-clone.c: Include omp-simd-clone.h.
5215 (expand_simd_clones): No longer static.
5216 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5217 cgraph.h and omp-simd-clone.h.
5218 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5219 (vect_recog_widen_shift_pattern): Formatting fix.
5220 (vect_pattern_recog_1): Don't check optab for calls.
5223 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5224 operands[2] into a REG before using gen_lowpart on it.
5226 2018-02-12 Jeff Law <law@redhat.com>
5229 * config/sh/sh.c (find_barrier): Consider a sibling call
5232 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5233 successfully back substituting a reg.
5235 2018-02-12 Richard Biener <rguenther@suse.de>
5237 PR tree-optimization/84037
5238 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5239 parameter, move visited init to caller.
5240 (vect_slp_analyze_operations): Separate cost from validity
5241 check, initialize visited once for all instances.
5242 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5244 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5246 (vect_model_promotion_demotion_cost): Likewise.
5247 (vectorizable_bswap): Guard cost modeling with !slp_node
5248 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5250 (vectorizable_call): Likewise.
5251 (vectorizable_conversion): Likewise.
5252 (vectorizable_assignment): Likewise.
5253 (vectorizable_shift): Likewise.
5254 (vectorizable_operation): Likewise.
5255 (vectorizable_store): Likewise.
5256 (vectorizable_load): Likewise.
5257 (vectorizable_condition): Likewise.
5258 (vectorizable_comparison): Likewise.
5260 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5263 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5264 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5265 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5267 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5270 * params.def (inline-min-speedup): Increase from 8 to 15.
5271 (max-inline-insns-auto): Decrease from 40 to 30.
5272 * ipa-split.c (consider_split): Add some buffer for function to
5273 be considered inlining candidate.
5274 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5277 2018-02-12 Richard Biener <rguenther@suse.de>
5279 PR tree-optimization/84037
5280 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5281 matched stmts if we cannot swap the non-matched ones.
5283 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5285 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5286 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5287 _mm_maskz_scalef_round_ss): New intrinsics.
5288 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5289 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5290 __builtin_ia32_scalefss_round): Remove.
5291 (__builtin_ia32_scalefsd_mask_round,
5292 __builtin_ia32_scalefss_mask_round): New intrinsics.
5293 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5294 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5295 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5296 "<round_constraint>")): Changed to ...
5297 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5298 "<round_scalar_constraint>")): ... this.
5299 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5300 %0, %1, %2<round_op3>}"): Changed to ...
5301 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5302 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5303 %2<round_scalar_mask_op3>}"): ... this.
5304 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5306 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5308 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5309 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5310 (_mm_maskz_sqrt_round_ss): New intrinsics.
5311 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5312 (__builtin_ia32_sqrtsd_mask_round)
5313 (__builtin_ia32_sqrtss_mask_round): New builtins.
5314 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5315 (__builtin_ia32_sqrtss_round): Remove.
5316 (__builtin_ia32_sqrtsd_mask_round)
5317 (__builtin_ia32_sqrtss_mask_round): New builtins.
5318 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5319 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5320 ((match_operand:VF_128 1 "vector_operand"
5321 "xBm,<round_constraint>")): Changed to ...
5322 ((match_operand:VF_128 1 "vector_operand"
5323 "xBm,<round_scalar_constraint>")): ... this.
5324 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5325 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5326 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5327 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5328 %<iptr>1<round_scalar_mask_op3>}): ... this.
5329 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5330 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5332 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5335 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5336 Cast vec_cmpeq result to correct type.
5337 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5338 Cast vec_cmpgt result to correct type.
5340 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5342 * final.c (final_scan_insn_1): Renamed from...
5343 (final_scan_insn): ... this. New wrapper, to recover
5344 seen from the outermost call in recursive ones.
5345 * config/sparc/sparc.c (output_return): Drop seen from call.
5346 (output_sibcall): Likewise.
5347 * config/visium/visium.c (output_branch): Likewise.
5349 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5351 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5354 2018-02-10 Alan Modra <amodra@gmail.com>
5357 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5358 Specify LR as an input.
5360 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5363 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5364 remove_member_access_dummy_vars): New functions.
5365 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5366 lower_omp_1, execute_lower_omp): Use them.
5368 PR rtl-optimization/84308
5369 * shrink-wrap.c (spread_components): Release todo vector.
5371 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5373 PR rtl-optimization/57193
5374 * ira-color.c (struct allocno_color_data): Add member
5375 conflict_allocno_hard_prefs.
5376 (update_conflict_allocno_hard_prefs): New.
5377 (bucket_allocno_compare_func): Add a preference based on
5378 conflict_allocno_hard_prefs.
5379 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5380 (color_allocnos): Remove a dead code. Initiate
5381 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5383 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5386 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5387 constraint from =wa to wa. Avoid a subreg on the output operand,
5388 instead use a pseudo and subreg it in a move.
5389 (p9_xxbrd_<mode>): Changed to ...
5390 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5391 (p9_xxbrd_v2df): New expander.
5392 (p9_xxbrw_<mode>): Changed to ...
5393 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5394 (p9_xxbrw_v4sf): New expander.
5396 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5398 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5400 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5403 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5404 multiply in 32-bit mode.
5405 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5406 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5409 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5411 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5412 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5413 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5414 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5416 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5419 * dwarf2out.c (is_trivial_indirect_ref): New function.
5420 (dwarf2out_late_global_decl): Do not generate a location
5421 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5422 and that are not defined in the current unit.
5424 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5426 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5427 instead of a libcall for UNORDERED.
5429 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5432 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5433 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5435 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5438 * config/s390/s390.c (s390_set_current_function): Invoke
5439 s390_indirect_branch_settings also if fndecl didn't change.
5441 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5443 * config/rs6000/rs6000.md (blockage): Set length to zero.
5445 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5447 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5449 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5452 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5453 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5457 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5458 PARALLEL incoming that failed vt_get_decl_and_offset check.
5461 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5462 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5464 (get_variable_section): For decls in named .bss* sections pass true as
5465 second argument to bss_initializer_p.
5467 2018-02-09 Marek Polacek <polacek@redhat.com>
5468 Jakub Jelinek <jakub@redhat.com>
5471 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5472 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5473 Sync some changes from cxx_fold_indirect_ref.
5475 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5477 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5479 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5480 (BLOCK_INLINE_ENTRY_LABEL): New.
5481 (dwarf2out_var_location): Disregard inline entry markers.
5482 (inline_entry_data): New struct.
5483 (inline_entry_data_hasher): New hashtable type.
5484 (inline_entry_data_hasher::hash): New.
5485 (inline_entry_data_hasher::equal): New.
5486 (inline_entry_data_table): New variable.
5487 (add_high_low_attributes): Add DW_AT_entry_pc and
5488 DW_AT_GNU_entry_view attributes if a pending entry is found
5489 in inline_entry_data_table. Add old entry_pc attribute only
5490 if debug nonbinding markers are disabled.
5491 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5492 markers are enabled.
5493 (block_within_block_p, dwarf2out_inline_entry): New.
5494 (dwarf2out_finish): Check that no entries remained in
5495 inline_entry_data_table.
5496 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5497 (final_scan_insn, notice_source_line): Likewise.
5498 (rest_of_clean_state): Skip inline entry markers.
5499 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5501 * gimple.c (gimple_build_debug_inline_entry): New.
5502 * gimple.h (enum gimple_debug_subcode): Add
5503 GIMPLE_DEBUG_INLINE_ENTRY.
5504 (gimple_build_debug_inline_entry): Declare.
5505 (gimple_debug_inline_entry_p): New.
5506 (gimple_debug_nonbind_marker_p): Adjust.
5507 * insn-notes.def (INLINE_ENTRY): New.
5508 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5509 inline entry marker notes.
5510 (print_insn): Likewise.
5511 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5512 (INSN_DEBUG_MARKER_KIND): Likewise.
5513 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5514 * tree-inline.c (expand_call_inline): Build and insert
5515 debug_inline_entry stmt.
5516 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5517 inline entry blocks early, if nonbind markers are enabled.
5518 (dump_scope_block): Dump fragment info.
5519 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5520 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5521 (gimple_build_debug_inline_entry): New.
5522 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5523 Enable/disable inline entry points too.
5524 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5525 (DEBUG_INSN): Describe inline entry markers.
5527 * common.opt (gvariable-location-views): New.
5528 (gvariable-location-views=incompat5): New.
5529 * config.in: Rebuilt.
5530 * configure: Rebuilt.
5531 * configure.ac: Test assembler for view support.
5532 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5533 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5534 * dwarf2out.c (var_loc_view): New typedef.
5535 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5536 (dwarf2out_locviews_in_attribute): New.
5537 (dwarf2out_locviews_in_loclist): New.
5538 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5539 (enum dw_line_info_opcode): Add LI_adv_address.
5540 (struct dw_line_info_table): Add view.
5541 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5542 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5543 (zero_view_p): New variable.
5544 (ZERO_VIEW_P): New macro.
5545 (output_asm_line_debug_info): New.
5546 (struct var_loc_node): Add view.
5547 (add_AT_view_list, AT_loc_list): New.
5548 (add_var_loc_to_decl): Add view param. Test it against last.
5549 (new_loc_list): Add view params. Record them.
5550 (AT_loc_list_ptr): Handle loc and view lists.
5551 (view_list_to_loc_list_val_node): New.
5552 (print_dw_val): Handle dw_val_class_view_list.
5553 (size_of_die): Likewise.
5554 (value_format): Likewise.
5555 (loc_list_has_views): New.
5556 (gen_llsym): Set vl_symbol too.
5557 (maybe_gen_llsym, skip_loc_list_entry): New.
5558 (dwarf2out_maybe_output_loclist_view_pair): New.
5559 (output_loc_list): Output view list or entries too.
5560 (output_view_list_offset): New.
5561 (output_die): Handle dw_val_class_view_list.
5562 (output_dwarf_version): New.
5563 (output_compilation_unit_header): Use it.
5564 (output_skeleton_debug_sections): Likewise.
5565 (output_rnglists, output_line_info): Likewise.
5566 (output_pubnames, output_aranges): Update version comments.
5567 (output_one_line_info_table): Output view numbers in asm comments.
5568 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5569 Call maybe_gen_llsym.
5570 (loc_list_from_tree_1): Adjust.
5571 (add_AT_location_description): Create view list attribute if
5572 needed, check it's absent otherwise.
5573 (convert_cfa_to_fb_loc_list): Adjust.
5574 (maybe_emit_file): Call output_asm_line_debug_info for test.
5575 (dwarf2out_var_location): Reset views as needed. Precompute
5576 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5577 attribute. Set view.
5578 (new_line_info_table): Reset next view.
5579 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5580 (dwarf2out_source_line): Likewise. Output view resets and labels to
5581 the assembler, or select appropriate line info opcodes.
5582 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5583 (optimize_string_length): Catch it. Adjust.
5584 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5585 dw_val_class_view_list, and remove it if no longer needed.
5586 (hash_loc_list): Hash view numbers.
5587 (loc_list_hasher::equal): Compare them.
5588 (optimize_location_lists): Check whether a view list symbol is
5589 needed, and whether the locview attribute is present, and
5590 whether they match. Remove the locview attribute if no longer
5592 (index_location_lists): Call skip_loc_list_entry for test.
5593 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5594 Use output_dwarf_version.
5595 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5596 (struct dw_val_node): Add val_view_list.
5597 * final.c (SEEN_NEXT_VIEW): New.
5598 (set_next_view_needed): New.
5599 (clear_next_view_needed): New.
5600 (maybe_output_next_view): New.
5601 (final_start_function): Rename to...
5602 (final_start_function_1): ... this. Take pointer to FIRST,
5603 add SEEN parameter. Emit param bindings in the initial view.
5604 (final_start_function): Reintroduce SEEN-less interface.
5605 (final): Rename to...
5606 (final_1): ... this. Take SEEN parameter. Output final pending
5607 next view at the end.
5608 (final): Reintroduce seen-less interface.
5609 (final_scan_insn): Output pending next view before switching
5610 sections or ending a block. Mark the next view as needed when
5611 outputting variable locations. Notify debug backend of section
5612 changes, and of location view changes.
5613 (rest_of_handle_final): Adjust.
5614 * toplev.c (process_options): Autodetect value for debug variable
5615 location views option. Warn on incompat5 without -gdwarf-5.
5616 * doc/invoke.texi (gvariable-location-views): New.
5617 (gvariable-location-views=incompat5): New.
5618 (gno-variable-location-views): New.
5620 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5622 PR tree-optimization/84136
5623 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5624 that the result of find_edge is non-NULL.
5626 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5629 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5630 storing integer register in SImode. Fix cost of 256 and 512
5631 byte aligned SSE register store.
5633 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5635 * config/i386/i386.c (ix86_multiplication_cost): Fix
5636 multiplication cost for TARGET_AVX512DQ.
5638 2018-02-08 Marek Polacek <polacek@redhat.com>
5640 PR tree-optimization/84238
5641 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5644 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5646 PR tree-optimization/84265
5647 * tree-vect-stmts.c (vectorizable_store): Don't treat
5648 VMAT_CONTIGUOUS accesses as grouped.
5649 (vectorizable_load): Likewise.
5651 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5653 PR tree-optimization/81635
5654 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5655 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5656 (test_round_for_mask): New functions.
5657 (wide_int_cc_tests): Call test_round_for_mask.
5658 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5659 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5660 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5661 range returned by get_range_info.
5663 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5666 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5667 * symtab.c: Include builtins.h
5668 (symtab_node::output_to_lto_symbol_table_p): Move here
5669 from lto-streamer-out.c:output_symbol_p.
5670 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5671 (output_symbol_p): Move all logic to symtab.c
5672 (produce_symtab): Update.
5674 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5676 * config/s390/s390-opts.h (enum indirect_branch): Define.
5677 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5678 (s390_indirect_branch_via_thunk)
5679 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5680 (enum s390_indirect_branch_type): Define.
5681 * config/s390/s390.c (struct s390_frame_layout, struct
5682 machine_function): Remove.
5683 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5684 (indirect_branch_table_label_no, indirect_branch_table_name):
5686 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5687 (enum s390_indirect_branch_option): Define.
5688 (s390_return_addr_from_memory): New function.
5689 (s390_handle_string_attribute): New function.
5690 (s390_attribute_table): Add new attribute handler.
5691 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5692 (s390_indirect_branch_via_thunk): New function.
5693 (s390_indirect_branch_via_inline_thunk): New function.
5694 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5695 sibling call optimization for non z10 compiles.
5696 (s390_emit_call): Force indirect branch target to be a single
5697 register. Add r1 clobber for non-z10 compiles.
5698 (s390_emit_epilogue): Emit return jump via return_use expander.
5699 (s390_reorg): Handle JUMP_INSNs as execute targets.
5700 (s390_option_override_internal): Perform validity checks for the
5701 new command line options.
5702 (s390_indirect_branch_attrvalue): New function.
5703 (s390_indirect_branch_settings): New function.
5704 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5705 (s390_output_indirect_thunk_function): New function.
5706 (s390_code_end): Implement target hook.
5707 (s390_case_values_threshold): Implement target hook.
5708 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5710 * config/s390/s390.h (struct s390_frame_layout)
5711 (struct machine_function): Move here from s390.c.
5712 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5713 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5714 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5715 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5716 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5717 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5718 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5719 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5720 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5721 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5722 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5723 (mnemonic attribute): Add values which aren't recognized
5725 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5726 pattern for branch conversion. Fix mnemonic attribute.
5727 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5728 indirect branch via thunk if requested.
5729 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5730 ("*indirect_jump"): Disable for branch conversion using out of
5732 ("indirect_jump_via_thunk<mode>_z10")
5733 ("indirect_jump_via_thunk<mode>")
5734 ("indirect_jump_via_inlinethunk<mode>_z10")
5735 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5736 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5737 ("casesi_jump_via_inlinethunk<mode>_z10")
5738 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5739 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5740 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5741 ("*indirect2_jump"): Disable for branch conversion.
5742 ("casesi_jump"): Turn into expander and expand patterns for branch
5744 ("return_use"): New expander.
5745 ("*return"): Emit return via thunk and rename it to ...
5746 ("*return<mode>"): ... this one.
5747 * config/s390/s390.opt: Add new options and and enum for the
5750 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5752 * lra-constraints.c (match_reload): Unconditionally use
5753 gen_lowpart_SUBREG, rather than selecting between that
5754 and equivalent gen_rtx_SUBREG code.
5756 2018-02-08 Richard Biener <rguenther@suse.de>
5758 PR tree-optimization/84233
5759 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5760 changed flag instead of boguously re-using phi_inserted.
5762 2018-02-08 Martin Jambor <mjambor@suse.cz>
5764 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5765 static local variables.
5767 2018-02-08 Richard Biener <rguenther@suse.de>
5769 PR tree-optimization/84278
5770 * tree-vect-stmts.c (vectorizable_store): When looking for
5771 smaller vector types to perform grouped strided loads/stores
5772 make sure the mode is supported by the target.
5773 (vectorizable_load): Likewise.
5775 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5777 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5778 Increase LDP/STP opportunities by adding adjacent callee-saves.
5780 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5782 PR rtl-optimization/84068
5783 PR rtl-optimization/83459
5784 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5786 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5788 PR tree-optimization/84224
5789 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5790 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5793 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5796 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5797 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5798 count, remove one USE.
5800 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5802 * doc/install.texi (Configuration): Document the
5803 --with-long-double-format={ibm,ieee} PowerPC configuration
5807 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5808 Convert from define_expand to be define_insn_and_split. Rework
5809 float/double/_Float128 conversions to QI/HI/SImode to work with
5810 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5811 conversions to QI/HImode types did a store and then a load to
5812 truncate the value. For conversions to VSX registers, don't split
5813 the insn, instead emit the code directly. Use the code iterator
5814 any_fix to combine signed and unsigned conversions.
5815 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5816 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5817 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5818 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5819 (fix_<mode>di2_hw): Likewise.
5820 (fixuns_<mode>di2_hw): Likewise.
5821 (fix_<mode>si2_hw): Likewise.
5822 (fixuns_<mode>si2_hw): Likewise.
5823 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5824 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5825 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5826 fix<uns>_trunc<SFDF:mode>si2_p8.
5827 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5829 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5830 (fix<uns>_<mode>_mem): Likewise.
5831 (fctiw<u>z_<mode>_mem): Likewise.
5832 (fix<uns>_<mode>_mem): Likewise.
5833 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5834 the register allocator from doing a direct move to the GPRs to do
5835 a store, and instead use the ISA 3.0 store byte/half-word from
5836 vector register instruction. For IEEE 128-bit floating point,
5837 also optimize stores of 32-bit ints.
5838 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5840 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5842 * genextract.c (push_pathstr_operand): New function to support
5844 (walk_rtx): Call push_pathstr_operand.
5845 (print_path): Support [a-zA-Z].
5847 2018-02-07 Richard Biener <rguenther@suse.de>
5849 PR tree-optimization/84037
5850 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5851 (cse_and_gimplify_to_preheader): Declare.
5852 (vect_get_place_in_interleaving_chain): Likewise.
5853 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5855 (_loop_vec_info::~_loop_vec_info): Delete it.
5856 (cse_and_gimplify_to_preheader): New function.
5857 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5858 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5859 (vectorizable_load): Likewise. For grouped stores always base
5860 the IV on the first element.
5861 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5862 condition before gimplifying.
5864 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5866 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5867 *DIV_EXPR and *MOD_EXPR.
5869 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5872 * config/i386/i386.c (ix86_option_override_internal): Mask out
5873 the CF_SET bit when checking -fcf-protection.
5875 2018-02-07 Tom de Vries <tom@codesourcery.com>
5878 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5881 2018-02-07 Richard Biener <rguenther@suse.de>
5883 PR tree-optimization/84204
5884 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5887 PR tree-optimization/84205
5888 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5889 special-case isl_ast_op_zdiv_r.
5891 PR tree-optimization/84223
5892 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5893 Only add conditions from within the region.
5894 (gather_bbs::after_dom_children): Adjust.
5896 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5899 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5900 * config/avr/avr.md: Only post-reload split REG-REG moves if
5901 either register is GENERAL_REG_P.
5903 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5905 PR tree-optimization/84235
5906 * tree-ssa-scopedtables.c
5907 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5908 if the subtraction is performed in floating point type where NaNs are
5909 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5910 build 1. Formatting fix.
5912 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5915 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5916 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5917 and skip it regardless of bb boundaries. Use CALL_P macro,
5918 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5919 unnecessarily, formatting fix.
5921 2018-02-06 Michael Collison <michael.collison@arm.com>
5923 * config/arm/thumb2.md:
5924 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5925 (*thumb_mov_notscc): Ditto.
5927 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5930 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5931 unsigned_fix, not "s".
5933 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5935 * configure.ac (gcc_fn_eh_frame_ro): New function.
5936 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5937 correct .eh_frame permissions.
5938 * configure: Regenerate.
5940 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5942 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5945 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5947 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5948 Display warning message for -mno-speculate-indirect-jumps.
5950 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5952 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5954 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5956 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5958 PR tree-optimization/84225
5959 * tree-eh.c (find_trapping_overflow): Only call
5960 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5962 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5965 * config/i386/i386.c: Reimplement the check of possible options
5966 -mibt/-mshstk conbination. Change error messages.
5967 * doc/invoke.texi: Fix a typo: remove extra '='.
5969 2018-02-06 Marek Polacek <polacek@redhat.com>
5971 PR tree-optimization/84228
5972 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5974 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5977 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5978 emitted arch directives.
5979 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5980 __ARM_FEATURE_COPROC before changing architectures.
5982 2018-02-06 Richard Biener <rguenther@suse.de>
5984 * config/i386/i386.c (print_reg): Fix typo.
5985 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5987 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5989 * configure: Regenerate.
5991 2018-02-05 Martin Sebor <msebor@redhat.com>
5993 PR tree-optimization/83369
5994 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5997 2018-02-05 Martin Liska <mliska@suse.cz>
5999 * doc/invoke.texi: Cherry-pick upstream r323995.
6001 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6003 * ira.c (ira_init_register_move_cost): Adjust comment.
6005 2018-02-05 Martin Liska <mliska@suse.cz>
6007 PR gcov-profile/84137
6008 * doc/gcov.texi: Fix typo in documentation.
6010 2018-02-05 Martin Liska <mliska@suse.cz>
6012 PR gcov-profile/83879
6013 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6014 using dlopen functionality.
6016 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6018 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6019 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6020 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6021 _mm_maskz_range_round_ss): New intrinsics.
6022 (__builtin_ia32_rangesd128_round)
6023 (__builtin_ia32_rangess128_round): Remove.
6024 (__builtin_ia32_rangesd128_mask_round,
6025 __builtin_ia32_rangess128_mask_round): New builtins.
6026 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6027 __builtin_ia32_rangess128_round): Remove.
6028 (__builtin_ia32_rangesd128_mask_round,
6029 __builtin_ia32_rangess128_mask_round): New builtins.
6030 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6031 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6032 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6033 "<round_saeonly_constraint>")): Changed to ...
6034 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6035 "<round_saeonly_scalar_constraint>")): ... this.
6036 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6037 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6038 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6039 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6040 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6042 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6044 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6046 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6047 Remove all values except native, 8540 and 8548.
6049 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6051 * config/i386/i386.c (ix86_output_function_return): Pass
6052 INVALID_REGNUM, instead of -1, as invalid register number to
6053 indirect_thunk_name and output_indirect_thunk.
6055 2018-02-02 Julia Koval <julia.koval@intel.com>
6057 * config.gcc: Add -march=icelake.
6058 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6059 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6060 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6061 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6062 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6063 (processor_target_table): Add icelake.
6064 (ix86_option_override_internal): Handle new PTAs.
6065 (get_builtin_code_for_version): Handle icelake.
6066 (M_INTEL_COREI7_ICELAKE): New.
6067 (fold_builtin_cpu): Handle icelake.
6068 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6069 * doc/invoke.texi: Add -march=icelake.
6071 2018-02-02 Julia Koval <julia.koval@intel.com>
6073 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6074 to wide_int_bitmask.
6075 * wide-int-bitmask.h: New.
6077 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6080 * config/i386/i386.md: Replace Pmode with word_mode in
6081 builtin_setjmp_setup and builtin_longjmp to support x32.
6083 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6087 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6089 (rs6000_supported_cpu_names): New static variable.
6090 (linux_cpu_translation_table): Likewise.
6091 (elf_platform) <cpu>: Define new static variable and use it.
6092 Translate kernel AT_PLATFORM name to canonical name if needed.
6093 Error if platform name is unknown.
6095 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6098 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6100 2018-02-01 Jeff Law <law@redhat.com>
6103 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6104 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6105 the scratch if RELEASE_VIA_POP is false.
6106 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6107 If we have to save a temporary register, decrement SIZE appropriately.
6108 Pass new arguments to release_scratch_register_on_entry.
6109 (ix86_adjust_stack_and_probe): Likewise.
6110 (ix86_emit_probe_stack_range): Pass new arguments to
6111 release_scratch_register_on_entry.
6113 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6115 PR rtl-optimization/84157
6116 * combine.c (change_zero_ext): Use REG_P predicate in
6117 front of HARD_REGISTER_P predicate.
6119 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6121 * config/avr/avr.c (avr_option_override): Move disabling of
6122 -fdelete-null-pointer-checks to...
6123 * common/config/avr/avr-common.c (avr_option_optimization_table):
6126 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6128 PR tree-optimization/81635
6129 * tree-data-ref.c (split_constant_offset_1): For types that
6130 wrap on overflow, try to use range info to prove that wrapping
6133 2018-02-01 Renlin Li <renlin.li@arm.com>
6136 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6138 (aarch64_register_move_cost): Likewise.
6139 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6141 (REG_CLASS_NAMES): Likewise.
6142 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6143 TAILCALL_ADDR_REGS. Remove IP registers.
6144 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6146 2018-02-01 Richard Biener <rguenther@suse.de>
6148 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6149 for specifying RPO order and allow NULL for that.
6150 * domwalk.c (dom_walker::dom_walker): Likewise.
6151 (dom_walker::walk): Handle NULL RPO order.
6152 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6154 (rewrite_update_dom_walker): Likewise.
6155 (mark_def_dom_walker): Likewise.
6157 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6159 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6160 (aarch64_maybe_expand_sve_subreg_move): Declare.
6161 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6162 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6164 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6165 that are semantically a reverse operation.
6166 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6167 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6168 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6170 (aarch64_can_change_mode_class): For big-endian, forbid changes
6171 between two SVE modes if they have different element sizes.
6173 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6175 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6176 the TImode handling for big-endian targets.
6178 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6180 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6181 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6183 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6184 Remove BSWAP handing for big-endian targets and use the form of
6185 LD1RQ appropariate for the mode.
6187 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6189 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6190 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6193 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6196 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6197 check for operands that need to go through aarch64_sve_reload_be.
6199 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6201 PR tree-optimization/81661
6202 PR tree-optimization/84117
6203 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6204 * tree-eh.c: Include gimplify.h.
6205 (find_trapping_overflow, replace_trapping_overflow,
6206 rewrite_to_non_trapping_overflow): New functions.
6207 * tree-vect-loop.c: Include tree-eh.h.
6208 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6209 * tree-data-ref.c: Include tree-eh.h.
6210 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6212 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6214 PR rtl-optimization/84123
6215 * combine.c (change_zero_ext): Check if hard register satisfies
6216 can_change_dest_mode before calling gen_lowpart_SUBREG.
6218 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6221 * ira.c (ira_init_register_move_cost): Remove assert.
6223 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6225 PR rtl-optimization/84071
6226 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6227 * doc/tm.texi: Regenerate.
6229 2018-01-31 Richard Biener <rguenther@suse.de>
6231 PR tree-optimization/84132
6232 * tree-data-ref.c (analyze_miv_subscript): Properly
6233 check whether evolution_function_is_affine_multivariate_p
6234 before calling gcd_of_steps_may_divide_p.
6236 2018-01-31 Julia Koval <julia.koval@intel.com>
6239 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6240 * config/i386/i386.md (rdpid_rex64) New.
6241 (rdpid): Make 32bit only.
6243 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6246 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6247 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6249 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6252 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6254 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6256 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6258 PR rtl-optimization/84071
6259 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6260 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6262 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6264 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6265 (arc_attribute_table): Add 'aux' attribute.
6266 (arc_in_small_data_p): Consider aux like variables.
6267 (arc_is_aux_reg_p): New function.
6268 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6269 (arc_get_aux_arg): New function.
6270 (prepare_move_operands): Handle aux-register access.
6271 (arc_handle_aux_attribute): New function.
6272 * doc/extend.texi (ARC Variable attributes): Add subsection.
6274 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6276 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6277 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6278 (arc_attribute_table): Add 'uncached' attribute.
6279 (arc_print_operand): Print '.di' flag for uncached memory
6281 (arc_in_small_data_p): Do not consider for small data the uncached
6283 (arc_is_uncached_mem_p): New function.
6284 * config/arc/predicates.md (compact_store_memory_operand): Check
6285 for uncached memory accesses.
6286 (nonvol_nonimm_operand): Likewise.
6287 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6289 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6292 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6293 falign-loops=): Add Optimization flag.
6295 2018-01-30 Jeff Law <law@redhat.com>
6298 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6299 INT_REGISTERS_SAVED. Check it prior to calling
6300 get_scratch_register_on_entry.
6301 (ix86_adjust_stack_and_probe): Similarly.
6302 (ix86_emit_probe_stack_range): Similarly.
6303 (ix86_expand_prologue): Corresponding changes.
6305 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6308 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6309 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6311 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6314 * lra-constraints.c (curr_insn_transform): Process AND in the
6317 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6319 PR rtl-optimization/83986
6320 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6321 dependence against last_pending_memory_flush in addition to
6324 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6326 PR tree-optimization/81611
6327 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6330 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6333 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6336 2018-01-30 Richard Biener <rguenther@suse.de>
6337 Jakub Jelinek <jakub@redhat.com>
6339 PR tree-optimization/84111
6340 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6341 inner loops added during recursion, as they don't have up-to-date
6344 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6347 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6348 (can_inline_edge_by_limits_p): ... here.
6349 (can_early_inline_edge_p, check_callers,
6350 update_caller_keys, update_callee_keys, recursive_inlining,
6351 add_new_edges_to_heap, speculation_useful_p,
6352 inline_small_functions,
6353 inline_small_functions, flatten_function,
6354 inline_to_all_callers_1): Update.
6356 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6358 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6361 2018-01-30 Richard Biener <rguenther@suse.de>
6363 PR tree-optimization/83008
6364 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6365 invariant and constant vector uses in stmts when they need
6368 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6371 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6372 * configure: Regenerate.
6374 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6376 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6378 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6379 Use gen_rtx_REG rather than gen_lowpart.
6381 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6383 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6384 rather than 0 when creating partial subregs.
6386 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6388 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6391 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6394 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6395 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6396 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6397 flags. This restores the settings used before the 2017-07-24.
6398 Turning off pre increment/decrement/modify allows IVOPTS to
6399 optimize DF/SF loops where the index is an int.
6401 2018-01-29 Richard Biener <rguenther@suse.de>
6402 Kelvin Nilsen <kelvin@gcc.gnu.org>
6405 * tree-vect-stmts.c (vectorizable_call): Don't call
6406 targetm.vectorize_builtin_md_vectorized_function if callee is
6409 2018-01-22 Carl Love <cel@us.ibm.com>
6411 * doc/extend.tex: Fix typo in second arg in
6412 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6414 2018-01-29 Richard Biener <rguenther@suse.de>
6416 PR tree-optimization/84086
6417 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6418 (flush_ssaname_freelist): When SSA names were released reset
6419 the SCEV hash table.
6421 2018-01-29 Richard Biener <rguenther@suse.de>
6423 PR tree-optimization/84057
6424 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6425 removed paths when removing edges.
6427 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6429 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6430 -mfunction-return=@var{choice}.
6432 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6435 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6437 (layout::print_source_line): Likewise.
6438 (test_get_line_width_without_trailing_whitespace): Add test cases.
6440 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6443 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6446 2018-01-26 Jim Wilson <jimw@sifive.com>
6448 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6450 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6453 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6455 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6456 and CMP + SUB-immediate -> SUBS.
6458 2018-01-26 Martin Sebor <msebor@redhat.com>
6460 PR tree-optimization/83896
6461 * tree-ssa-strlen.c (get_string_len): Rename...
6462 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6463 Avoid assuming length is constant.
6464 (handle_char_store): Use HOST_WIDE_INT for string length.
6466 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6469 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6470 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6472 2018-01-26 Richard Biener <rguenther@suse.de>
6474 PR rtl-optimization/84003
6475 * dse.c (record_store): Only record redundant stores when
6476 the earlier store aliases at least all accesses the later one does.
6478 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6480 PR rtl-optimization/83985
6481 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6482 REG_CFA_RESTORE insns.
6483 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6484 REG_CFA_RESTORE insns here.
6487 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6488 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6490 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6492 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6493 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6494 (arc_init): Likewise.
6495 (arc_override_options): Likewise.
6496 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6498 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6500 * config/arc/arc.h (TARGET_DBNZ): Define.
6501 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6502 properly set the tune attribute.
6503 (dbnz): Use TARGET_DBNZ guard.
6504 * config/arc/arc.opt (mtune): Add core3 option.
6506 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6508 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6509 recognize new pic like addresses.
6510 (arc_delegitimize_address): Clean up.
6512 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6514 * config/arc/arc-arches.def: Option mrf16 valid for all
6516 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6517 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6518 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6519 * config/arc/arc-tables.opt: Regenerate.
6520 * config/arc/arc.c (arc_conditional_register_usage): Handle
6521 reduced register file case.
6522 (arc_file_start): Set must have build attributes.
6523 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6525 * config/arc/arc.opt (mrf16): Add new option.
6526 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6527 * config/arc/genmultilib.awk: Handle new mrf16 option.
6528 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6529 * config/arc/t-multilib: Regenerate.
6530 * doc/invoke.texi (ARC Options): Document mrf16 option.
6532 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6534 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6535 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6536 (arc_attribute_table): Add 'secure_call' attribute.
6537 (arc_print_operand): Print secure call operand.
6538 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6540 (arc_is_secure_call_p): New function. * config/arc/arc.md
6541 (call_i): Add support for sjli instruction.
6542 (call_value_i): Likewise.
6543 * config/arc/constraints.md (Csc): New constraint.
6545 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6546 John Eric Martin <John.Martin@emmicro-us.com>
6548 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6549 * config/arc/arc.c (_arc_jli_section): New struct.
6550 (arc_jli_section): New type.
6551 (rc_jli_sections): New static variable.
6552 (arc_handle_jli_attribute): New function.
6553 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6554 (arc_file_end): New function.
6555 (TARGET_ASM_FILE_END): Define.
6556 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6557 (arc_add_jli_section): New function.
6558 (jli_call_scan): Likewise.
6559 (arc_reorg): Call jli_call_scan.
6560 (arc_output_addsi): Remove 'S' from printing asm operand.
6561 (arc_is_jli_call_p): New function.
6562 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6564 (movhi_insn): Likewise.
6565 (movsi_insn): Likewise.
6566 (movsi_set_cc_insn): Likewise.
6567 (loadqi_update): Likewise.
6568 (load_zeroextendqisi_update): Likewise.
6569 (load_signextendqisi_update): Likewise.
6570 (loadhi_update): Likewise.
6571 (load_zeroextendhisi_update): Likewise.
6572 (load_signextendhisi_update): Likewise.
6573 (loadsi_update): Likewise.
6574 (loadsf_update): Likewise.
6575 (movsicc_insn): Likewise.
6576 (bset_insn): Likewise.
6577 (bxor_insn): Likewise.
6578 (bclr_insn): Likewise.
6579 (bmsk_insn): Likewise.
6580 (bicsi3_insn): Likewise.
6581 (cmpsi_cc_c_insn): Likewise.
6582 (movsi_ne): Likewise.
6583 (movsi_cond_exec): Likewise.
6584 (clrsbsi2): Likewise.
6595 (call_i): Remove 'S' asm letter, add jli instruction.
6596 (call_value_i): Likewise.
6597 * config/arc/arc.op (mjli-always): New option.
6598 * config/arc/constraints.md (Cji): New constraint.
6599 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6601 (subsf3_fpx): Likewise.
6602 (mulsf3_fpx): Likewise.
6603 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6605 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6607 * doc/invoke.texi (ARC): Document mjli-always option.
6609 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6611 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6612 avoid addition with 0 and use incw and decw where possible.
6614 2018-01-26 Richard Biener <rguenther@suse.de>
6616 PR tree-optimization/81082
6617 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6618 association if it requires casting to unsigned.
6619 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6620 from fold_plusminus_mult_expr to catch important cases late when
6621 range info is available.
6623 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6625 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6626 * configure.ac (hidden_linkonce): New test.
6627 * configure: Regenerate.
6628 * config.in: Regenerate.
6630 2018-01-26 Julia Koval <julia.koval@intel.com>
6632 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6633 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6634 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6635 _mm_mask_bitshuffle_epi64_mask): Fix type.
6636 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6637 USI_FTYPE_V4DI_V4DI_USI): Remove.
6638 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6639 __builtin_ia32_vpshufbitqmb256_mask,
6640 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6641 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6642 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6644 2018-01-26 Alan Modra <amodra@gmail.com>
6647 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6648 UNSPEC_VBPERMQ. Sort other unspecs.
6650 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6652 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6654 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6657 * predict.c (drop_profile): Do not push/pop cfun; update also
6659 (handle_missing_profiles): Fix logic looking for zero profiles.
6661 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6664 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6665 on functions with #pragma omp declare simd or functions with simd
6667 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6668 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6669 Remove trailing \n from warning_at calls.
6671 2018-01-25 Tom de Vries <tom@codesourcery.com>
6674 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6675 for neutered workers.
6677 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6680 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6681 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6683 2018-01-24 Jeff Law <law@redhat.com>
6686 * i386.c (get_probe_interval): Move to earlier point.
6687 (ix86_compute_frame_layout): If -fstack-clash-protection and
6688 the frame is larger than the probe interval, then use pushes
6689 to save registers rather than reg->mem moves.
6690 (ix86_expand_prologue): Remove conditional for int_registers_saved
6693 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6696 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6697 min/max for never referenced object.
6699 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6702 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6704 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6705 attributes from DECL_ATTRIBUTES (decl) without affecting
6706 DECL_ATTRIBUTES (current_function_decl).
6707 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6708 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6710 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6712 PR tree-optimization/83979
6713 * fold-const.c (fold_comparison): Use constant_boolean_node
6714 instead of boolean_{true,false}_node.
6716 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6718 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6721 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6723 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6724 Simplify the clause that sets the length attribute.
6725 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6726 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6727 clause that sets the length attribute.
6728 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6730 2018-01-24 Tom de Vries <tom@codesourcery.com>
6733 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6734 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6735 Add strict parameter.
6736 (prevent_branch_around_nothing): Insert dummy insn between branch to
6737 label and label with no ptx insn inbetween.
6738 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6740 2018-01-24 Tom de Vries <tom@codesourcery.com>
6743 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6744 for neutered threads in warp.
6745 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6747 2018-01-24 Richard Biener <rguenther@suse.de>
6749 PR tree-optimization/83176
6750 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6753 2018-01-24 Richard Biener <rguenther@suse.de>
6755 PR tree-optimization/82819
6756 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6757 code generating pluses that are no-ops in the target precision.
6759 2018-01-24 Richard Biener <rguenther@suse.de>
6762 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6764 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6766 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6767 to merge probabilities.
6768 * predict.c (probably_never_executed): Also mark as cold functions
6769 with global 0 profile and guessed local profile.
6770 * profile-count.c (profile_probability::combine_with_count): New
6772 * profile-count.h (profile_probability::operator*,
6773 profile_probability::operator*=, profile_probability::operator/,
6774 profile_probability::operator/=): Reduce precision to adjusted
6775 and set value to guessed on contradictory divisions.
6776 (profile_probability::combine_with_freq): Remove.
6777 (profile_probability::combine_wiht_count): Declare.
6778 (profile_count::force_nonzero):: Set to adjusted.
6779 (profile_count::probability_in):: Set quality to adjusted.
6780 * tree-ssa-tail-merge.c (replace_block_by): Use
6783 2018-01-23 Andrew Waterman <andrew@sifive.com>
6784 Jim Wilson <jimw@sifive.com>
6786 * config/riscv/riscv.c (riscv_stack_boundary): New.
6787 (riscv_option_override): Set riscv_stack_boundary. Handle
6788 riscv_preferred_stack_boundary_arg.
6789 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6790 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6791 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6792 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6793 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6794 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6796 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6799 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6800 of struct ix86_frame.
6801 (ix86_expand_epilogue): Likewise. Add a local variable for
6802 the reg_save_offset field in struct ix86_frame.
6804 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6806 PR tree-optimization/82604
6807 * tree-loop-distribution.c (enum partition_kind): New enum item
6808 PKIND_PARTIAL_MEMSET.
6809 (partition_builtin_p): Support above new enum item.
6810 (generate_code_for_partition): Ditto.
6811 (compute_access_range): Differentiate cases that equality can be
6812 proven at all loops, the innermost loops or no loops.
6813 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6814 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6815 (finalize_partitions, distribute_loop): Don't fuse partition of
6816 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6817 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6820 2018-01-23 Martin Liska <mliska@suse.cz>
6822 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6823 order to ignore the predictor.
6824 (PRED_POLYMORPHIC_CALL): Likewise.
6825 (PRED_RECURSIVE_CALL): Likewise.
6827 2018-01-23 Martin Liska <mliska@suse.cz>
6829 * tree-profile.c (tree_profiling): Print function header to
6830 aware reader which function we are working on.
6831 * value-prof.c (gimple_find_values_to_profile): Do not print
6832 not interesting value histograms.
6834 2018-01-23 Martin Liska <mliska@suse.cz>
6836 * profile-count.h (enum profile_quality): Add
6837 profile_uninitialized as the first value. Do not number values
6838 as they are zero based.
6839 (profile_count::verify): Update sanity check.
6840 (profile_probability::verify): Likewise.
6842 2018-01-23 Nathan Sidwell <nathan@acm.org>
6844 * doc/invoke.texi (ffor-scope): Deprecate.
6846 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6848 PR tree-optimization/83510
6849 * domwalk.c (set_all_edges_as_executable): New function.
6850 (dom_walker::dom_walker): Convert bool param
6851 "skip_unreachable_blocks" to enum reachability. Move setup of
6852 edge flags to set_all_edges_as_executable and only do it when
6853 reachability is REACHABLE_BLOCKS.
6854 * domwalk.h (enum dom_walker::reachability): New enum.
6855 (dom_walker::dom_walker): Convert bool param
6856 "skip_unreachable_blocks" to enum reachability.
6857 (set_all_edges_as_executable): New decl.
6858 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6859 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6861 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6862 but converting true to REACHABLE_BLOCKS.
6863 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6865 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6866 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6867 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6869 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6870 if check_all_array_refs will be called.
6872 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6874 * tree.c (selftest::test_location_wrappers): Add more test
6877 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6879 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6880 (selftest::test_bit_in_range): Likewise.
6882 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6885 * doc/sourcebuild.texi (vect_float): Say that the selector
6886 only describes the situation when -funsafe-math-optimizations is on.
6887 (vect_float_strict): Document.
6889 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6891 PR tree-optimization/83965
6892 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6893 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6894 instead of checking only for a reduction.
6895 (vect_recog_widen_sum_pattern): Likewise.
6897 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6899 * predict.c (probably_never_executed): Only use precise profile info.
6900 (compute_function_frequency): Skip after inlining hack since we now
6901 have quality checking.
6903 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6905 * profile-count.h (profile_probability::very_unlikely,
6906 profile_probability::unlikely, profile_probability::even): Set
6907 precision to guessed.
6909 2018-01-23 Richard Biener <rguenther@suse.de>
6911 PR tree-optimization/83963
6912 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6913 Properly terminate dominator walk when crossing the exit edge not
6914 when visiting its source block.
6916 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6919 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6920 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6922 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6924 PR tree-optimization/83957
6925 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6926 semicolon after for body surrounded by braces.
6928 PR tree-optimization/83081
6929 * profile-count.h (profile_probability::split): New method.
6930 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6931 Use profile_probability::split.
6932 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6933 when splitting a single conditional jump into 2.
6935 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6937 PR tree-optimization/69452
6938 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6941 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6943 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6944 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6945 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6947 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6949 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6950 * config/rl78/rl78.md (movdi): New define_expand.
6951 * config/rl78/rl78.c (rl78_split_movdi): New function.
6953 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6956 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6958 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6959 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6960 128-bit to produce an UNSPEC move to get the double word with the
6961 signbit and then a shift directly to do signbit.
6962 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6963 implementation with a new version that just does either a direct
6964 move or a regular move. Move memory interface to separate insns.
6965 Move insns so they are next to the expander.
6966 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6967 with signbit move. Split big and little endian case.
6968 (signbit<mode>2_dm_mem_le): Likewise.
6969 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6970 (signbit<mode>2_dm2): Likewise.
6972 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6974 * config/rl78/rl78.md (anddi3): New define_expand.
6976 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6978 * config/rl78/rl78.md (umindi3): New define_expand.
6980 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6982 * config/rl78/rl78.md (smindi3): New define_expand.
6984 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6986 * config/rl78/rl78.md (smaxdi3): New define_expand.
6988 2018-01-22 Carl Love <cel@us.ibm.com>
6990 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6991 LVX_V1TI): Add macro expansion.
6992 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6993 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6994 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6995 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6996 Change check to determine if the instruction is a byte reversing
6997 entry. Fix typo in comment.
6998 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6999 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7000 Add def_builtin calls for new builtins.
7001 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7002 Add define_insn expansion.
7004 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7006 * config/rl78/rl78.md (umaxdi3): New define_expand.
7008 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7010 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7011 for non-QImode registers.
7013 2018-01-22 Richard Biener <rguenther@suse.de>
7015 PR tree-optimization/83963
7016 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7017 including the loop exit block.
7018 (scop_detection::merge_sese): Likewise.
7019 (scop_detection::add_scop): Do it here instead.
7021 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7023 * doc/sourcebuild.texi (arm_softfloat): Document.
7025 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7028 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7029 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7030 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7032 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7033 David Edelsohn <dje.gcc@gmail.com>
7036 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7037 Change "crset eq" to "crset 2".
7038 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7039 (*call_indirect_aix<mode>_nospec): Likewise.
7040 (*call_value_indirect_aix<mode>_nospec): Likewise.
7041 (*call_indirect_elfv2<mode>_nospec): Likewise.
7042 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7043 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7044 change assembly output from . to $.
7045 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7046 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7047 (*tablejump<mode>_internal1_nospec): Likewise.
7049 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7052 * config/sh/sh_optimize_sett_clrt.cc:
7053 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7055 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7057 PR tree-optimization/83940
7058 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7059 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7060 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7061 use it to pass back the definition type.
7062 (vect_check_store_rhs): Likewise rhs_dt_out.
7063 (vect_build_gather_load_calls): Add a mask_dt argument and use
7064 it instead of a call to vect_is_simple_use.
7065 (vectorizable_store): Update calls to vect_check_load_store_mask
7066 and vect_check_store_rhs. Use the dt returned by the latter instead
7067 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7068 instead of calls to vect_is_simple_use. Pass the scalar rather
7069 than the vector operand to vect_is_simple_use when handling
7070 second and subsequent copies of an rhs value.
7071 (vectorizable_load): Update calls to vect_check_load_store_mask
7072 and vect_build_gather_load_calls. Use the cached mask_dt and
7073 gs_info.offset_dt instead of calls to vect_is_simple_use.
7075 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7078 * tree-emutls.c: Include gimplify.h.
7079 (lower_emutls_2): New function.
7080 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7081 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7082 it before further processing.
7085 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7086 UINTVAL (trueop1) instead of INTVAL (op1).
7088 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7092 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7093 INCOMING_FRAME_SP_OFFSET if not defined.
7094 (scan_trace): Add ENTRY argument. If true and
7095 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7096 emit a note to adjust the CFA offset.
7097 (create_cfi_notes): Adjust scan_trace callers.
7098 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7099 INCOMING_FRAME_SP_OFFSET in the CIE.
7100 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7101 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7103 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7104 * doc/tm.texi: Regenerated.
7106 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7108 PR rtl-optimization/83147
7109 * lra-constraints.c (remove_inheritance_pseudos): Use
7110 lra_substitute_pseudo_within_insn.
7112 2018-01-19 Tom de Vries <tom@codesourcery.com>
7113 Cesar Philippidis <cesar@codesourcery.com>
7116 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7118 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7121 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7122 spaces for function labels.
7124 2018-01-19 Martin Liska <mliska@suse.cz>
7126 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7127 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7128 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7129 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7130 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7131 (PRED_CONST_RETURN): Change from 69 to 65.
7132 (PRED_NULL_RETURN): Change from 91 to 71.
7133 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7134 (PRED_LOOP_GUARD): Change from 66 to 73.
7136 2018-01-19 Martin Liska <mliska@suse.cz>
7138 * predict.c (predict_insn_def): Add new assert.
7139 (struct branch_predictor): Change type to signed integer.
7140 (test_prediction_value_range): Amend test to cover
7142 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7143 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7144 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7145 (PRED_LOOP_IV_COMPARE): Likewise.
7146 * predict.h (PROB_UNINITIALIZED): Define new constant.
7148 2018-01-19 Martin Liska <mliska@suse.cz>
7150 * predict.c (dump_prediction): Add new format for
7151 analyze_brprob.py script which is enabled with -details
7153 * profile-count.h (precise_p): New function.
7155 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7157 PR tree-optimization/83922
7158 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7159 there are no statements that need masking.
7160 (vect_active_double_reduction_p): New function.
7161 (vect_analyze_loop_operations): Use it when handling phis that
7162 are not in the loop header.
7164 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7166 PR tree-optimization/83914
7167 * tree-vect-loop.c (vectorizable_induction): Don't convert
7168 init_expr or apply the peeling adjustment for inductions
7169 that are nested within the vectorized loop.
7171 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7173 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7176 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7180 * function.h (gimplify_parameters): Add gimple_seq * argument.
7181 * function.c: Include gimple.h and options.h.
7182 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7183 for the added local temporaries if needed.
7184 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7185 if there are any parameter cleanups, wrap whole body into a
7186 try/finally with the cleanups.
7188 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7191 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7192 Use GET_MODE_CLASS for scalar floating point.
7194 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7198 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7199 Fix call of call_cgraph_insertion_hooks.
7201 2018-01-18 Martin Sebor <msebor@redhat.com>
7203 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7205 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7208 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7211 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7214 * common.opt: (-ffile-prefix-map): New option.
7215 * opts.c (common_handle_option): Defer it.
7216 * opts-global.c (handle_common_deferred_options): Handle it.
7217 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7218 * file-prefix-map.h: New file.
7219 (remap_debug_filename, add_debug_prefix_map): ...here.
7220 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7221 * final.c (debug_prefix_map, add_debug_prefix_map
7222 remap_debug_filename): Move to...
7223 * file-prefix-map.c: New file.
7224 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7225 generalize, get rid of alloca(), use strrchr() instead of strchr().
7226 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7227 Implement in terms of add_prefix_map().
7228 (remap_macro_filename, remap_debug_filename): Implement in term of
7230 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7231 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7232 * dbxout.c: Include file-prefix-map.h.
7233 * varasm.c: Likewise.
7234 * vmsdbgout.c: Likewise.
7235 * xcoffout.c: Likewise.
7236 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7237 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7238 * doc/invoke.texi (-ffile-prefix-map): Document.
7239 (-fdebug-prefix-map): Update description.
7241 2018-01-18 Martin Liska <mliska@suse.cz>
7243 * config/i386/i386.c (indirect_thunk_name): Document that also
7245 (output_indirect_thunk): Document why both instructions
7246 (pause and lfence) are generated.
7248 2018-01-18 Richard Biener <rguenther@suse.de>
7250 PR tree-optimization/83887
7251 * graphite-scop-detection.c
7252 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7253 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7254 (scop_detection::merge_sese): Re-implement with a flood-fill
7255 algorithm that properly finds a SESE region if it exists.
7257 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7260 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7261 pointer_diff optimizations use view_convert instead of convert.
7263 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7265 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7266 Generate different code for -mno-speculate-indirect-jumps.
7267 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7268 (*call_indirect_aix<mode>): Disable for
7269 -mno-speculate-indirect-jumps.
7270 (*call_indirect_aix<mode>_nospec): New define_insn.
7271 (*call_value_indirect_aix<mode>): Disable for
7272 -mno-speculate-indirect-jumps.
7273 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7274 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7275 -mno-speculate-indirect-jumps.
7276 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7278 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7280 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7281 long double type, set the flags for noting the default long double
7282 type, even if we don't pass or return a long double type.
7284 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7287 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7290 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7292 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7293 support for merge[hl].
7294 (fold_mergehl_helper): New helper function.
7295 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7296 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7297 (altivec_vmrglw_direct): Add xxmrglw insn.
7299 2018-01-17 Andrew Waterman <andrew@sifive.com>
7301 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7302 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7304 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7307 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7308 call the lto_location_cache before reading the
7309 DECL_SOURCE_LOCATION of the types.
7311 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7312 Richard Sandiford <richard.sandiford@linaro.org>
7314 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7315 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7316 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7317 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7318 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7320 * config/aarch64/constraints.md (aarch64_movti_operand):
7322 * config/aarch64/predicates.md (Uti): Add new constraint.
7324 2018-01-17 Carl Love <cel@us.ibm.com>
7326 * config/rs6000/vsx.md (define_expand xl_len_r,
7327 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7328 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7330 (define_expand, define_insn): Move the shift left from the
7331 define_insn to the define_expand for lxvl and stxvl instructions.
7332 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7333 and XL_LEN_R definitions to PURE.
7335 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7337 * config/i386/i386.c (indirect_thunk_name): Declare regno
7338 as unsigned int. Compare regno with INVALID_REGNUM.
7339 (output_indirect_thunk): Ditto.
7340 (output_indirect_thunk_function): Ditto.
7341 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7342 in the call to output_indirect_thunk_function.
7344 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7347 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7348 rather than the size of inner_type to determine the stack slot size
7349 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7351 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7354 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7357 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7359 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7360 endian Linux systems to optionally enable multilibs for selecting
7361 the long double type if the user configured an explicit type.
7362 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7363 have no long double multilibs if not defined.
7364 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7365 warn if the user used -mabi={ieee,ibm}longdouble and we built
7366 multilibs for long double.
7367 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7368 appropriate multilib option.
7369 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7371 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7372 for building long double multilibs.
7373 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7375 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7377 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7380 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7382 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7385 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7388 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7391 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7393 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7394 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7396 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7398 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7399 different rtl trees depending on TARGET_64BIT.
7400 (rs6000_gen_lvx): Likewise.
7402 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7404 * config/visium/visium.md (nop): Tweak comment.
7405 (hazard_nop): Likewise.
7407 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7409 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7410 -mspeculate-indirect-jumps.
7411 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7412 for -mno-speculate-indirect-jumps.
7413 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7414 (*call_value_indirect_elfv2<mode>): Disable for
7415 -mno-speculate-indirect-jumps.
7416 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7417 (indirect_jump): Emit different RTL for
7418 -mno-speculate-indirect-jumps.
7419 (*indirect_jump<mode>): Disable for
7420 -mno-speculate-indirect-jumps.
7421 (*indirect_jump<mode>_nospec): New define_insn.
7422 (tablejump): Emit different RTL for
7423 -mno-speculate-indirect-jumps.
7424 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7425 (tablejumpsi_nospec): New define_expand.
7426 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7427 (tablejumpdi_nospec): New define_expand.
7428 (*tablejump<mode>_internal1): Disable for
7429 -mno-speculate-indirect-jumps.
7430 (*tablejump<mode>_internal1_nospec): New define_insn.
7431 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7434 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7436 * caller-save.c (insert_save): Drop unnecessary parameter. All
7439 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7440 Richard Biener <rguenth@suse.de>
7443 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7444 return early, inline manually is_gimple_sizepos. Make sure if we
7445 call gimplify_expr we don't end up with a gimple constant.
7446 * tree.c (variably_modified_type_p): Don't return true for
7447 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7448 * gimplify.h (is_gimple_sizepos): Remove.
7450 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7452 PR tree-optimization/83857
7453 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7454 vectorizable_live_operation for pure SLP statements.
7455 (vectorizable_live_operation): Handle PHIs.
7457 2018-01-16 Richard Biener <rguenther@suse.de>
7459 PR tree-optimization/83867
7460 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7461 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7463 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7466 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7467 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7468 If off is not INTEGER_CST, issue a may not be aligned warning
7469 rather than isn't aligned. Use isn%'t rather than isn't.
7470 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7472 <case MULT_EXPR>: Improve the case when bottom and one of the
7473 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7474 operand, in that case check if the other operand is multiple of
7475 bottom divided by the INTEGER_CST operand.
7477 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7480 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7481 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7482 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7483 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7484 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7485 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7486 (pa_function_arg_size): New function.
7488 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7490 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7491 in a separate statement.
7493 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7495 PR tree-optimization/83847
7496 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7497 group gathers and scatters.
7499 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7501 PR rtl-optimization/86620
7502 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7504 PR rtl-optimization/83213
7505 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7506 to last if both are JUMP_INSNs.
7508 PR tree-optimization/83843
7509 * gimple-ssa-store-merging.c
7510 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7511 store_immediate_info for bswap/nop orig_stores.
7513 2018-01-15 Andrew Waterman <andrew@sifive.com>
7515 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7517 <UDIV>: Increase cost if !TARGET_DIV.
7519 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7521 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7522 (define_attr "cr_logical_3op"): New.
7523 (cceq_ior_compare): Adjust.
7524 (cceq_ior_compare_complement): Adjust.
7525 (*cceq_rev_compare): Adjust.
7526 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7527 (is_cracked_insn): Adjust.
7528 (insn_must_be_first_in_group): Adjust.
7529 * config/rs6000/40x.md: Adjust.
7530 * config/rs6000/440.md: Adjust.
7531 * config/rs6000/476.md: Adjust.
7532 * config/rs6000/601.md: Adjust.
7533 * config/rs6000/603.md: Adjust.
7534 * config/rs6000/6xx.md: Adjust.
7535 * config/rs6000/7450.md: Adjust.
7536 * config/rs6000/7xx.md: Adjust.
7537 * config/rs6000/8540.md: Adjust.
7538 * config/rs6000/cell.md: Adjust.
7539 * config/rs6000/e300c2c3.md: Adjust.
7540 * config/rs6000/e500mc.md: Adjust.
7541 * config/rs6000/e500mc64.md: Adjust.
7542 * config/rs6000/e5500.md: Adjust.
7543 * config/rs6000/e6500.md: Adjust.
7544 * config/rs6000/mpc.md: Adjust.
7545 * config/rs6000/power4.md: Adjust.
7546 * config/rs6000/power5.md: Adjust.
7547 * config/rs6000/power6.md: Adjust.
7548 * config/rs6000/power7.md: Adjust.
7549 * config/rs6000/power8.md: Adjust.
7550 * config/rs6000/power9.md: Adjust.
7551 * config/rs6000/rs64.md: Adjust.
7552 * config/rs6000/titan.md: Adjust.
7554 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7556 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7557 ix86_indirect_branch_register logic.
7559 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7561 * config/i386/constraints.md (Bs): Update
7562 ix86_indirect_branch_register check. Don't check
7563 ix86_indirect_branch_register with GOT_memory_operand.
7565 * config/i386/predicates.md (GOT_memory_operand): Don't check
7566 ix86_indirect_branch_register here.
7567 (GOT32_symbol_operand): Likewise.
7569 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7571 * config/i386/predicates.md (constant_call_address_operand):
7572 Rewrite ix86_indirect_branch_register logic.
7573 (sibcall_insn_operand): Likewise.
7575 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7577 * config/i386/constraints.md (Bs): Replace
7578 ix86_indirect_branch_thunk_register with
7579 ix86_indirect_branch_register.
7581 * config/i386/i386.md (indirect_jump): Likewise.
7582 (tablejump): Likewise.
7583 (*sibcall_memory): Likewise.
7584 (*sibcall_value_memory): Likewise.
7585 Peepholes of indirect call and jump via memory: Likewise.
7586 * config/i386/i386.opt: Likewise.
7587 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7588 (GOT_memory_operand): Likewise.
7589 (call_insn_operand): Likewise.
7590 (sibcall_insn_operand): Likewise.
7591 (GOT32_symbol_operand): Likewise.
7593 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7596 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7597 type rather than type addr's type points to.
7598 (expand_omp_atomic_mutex): Likewise.
7599 (expand_omp_atomic): Likewise.
7601 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7604 * config/i386/i386.c (output_indirect_thunk_function): Use
7605 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7606 for __x86_return_thunk.
7608 2018-01-15 Richard Biener <rguenther@suse.de>
7611 * expmed.c (extract_bit_field_1): Fix typo.
7613 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7616 * config/arm/iterators.md (VF): New mode iterator.
7617 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7618 Remove integer-related logic from pattern.
7619 (neon_vabd<mode>_3): Likewise.
7621 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7624 * common.opt (fstrict-overflow): No longer an alias.
7625 (fwrapv-pointer): New option.
7626 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7627 also for pointer types based on flag_wrapv_pointer.
7628 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7629 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7630 opts->x_flag_wrapv got set.
7631 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7632 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7633 POINTER_TYPE_OVERFLOW_UNDEFINED.
7634 * match.pd: Likewise in address comparison pattern.
7635 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7637 2018-01-15 Richard Biener <rguenther@suse.de>
7640 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7641 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7642 Reset type names to their identifier if their TYPE_DECL doesn't
7643 have linkage (and thus is used for ODR and devirt).
7644 (save_debug_info_for_decl): Remove.
7645 (save_debug_info_for_type): Likewise.
7646 (add_tree_to_fld_list): Adjust.
7647 * tree-pretty-print.c (dump_generic_node): Make dumping of
7648 type names more robust.
7650 2018-01-15 Richard Biener <rguenther@suse.de>
7652 * BASE-VER: Bump to 8.0.1.
7654 2018-01-14 Martin Sebor <msebor@redhat.com>
7657 * builtins.c (check_access): Avoid warning when the no-warning bit
7660 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7662 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7663 * ira-color (allocno_hard_regs_compare): Likewise.
7665 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7668 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7669 Use .pushsection/.popsection.
7671 2018-01-14 Martin Sebor <msebor@redhat.com>
7674 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7676 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7678 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7679 entry from extra_headers.
7680 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7681 extra_headers, make the list bitwise identical to the i?86-*-* one.
7683 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7685 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7686 -mcmodel=large with -mindirect-branch=thunk,
7687 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7688 -mfunction-return=thunk-extern.
7689 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7690 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7691 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7693 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7695 * config/i386/i386.c (print_reg): Print the name of the full
7696 integer register without '%'.
7697 (ix86_print_operand): Handle 'V'.
7698 * doc/extend.texi: Document 'V' modifier.
7700 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7702 * config/i386/constraints.md (Bs): Disallow memory operand for
7703 -mindirect-branch-register.
7705 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7706 (GOT_memory_operand): Likewise.
7707 (call_insn_operand): Likewise.
7708 (sibcall_insn_operand): Likewise.
7709 (GOT32_symbol_operand): Likewise.
7710 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7711 for -mindirect-branch-register.
7712 (tablejump): Likewise.
7713 (*sibcall_memory): Likewise.
7714 (*sibcall_value_memory): Likewise.
7715 Disallow peepholes of indirect call and jump via memory for
7716 -mindirect-branch-register.
7717 (*call_pop): Replace m with Bw.
7718 (*call_value_pop): Likewise.
7719 (*sibcall_pop_memory): Replace m with Bs.
7720 * config/i386/i386.opt (mindirect-branch-register): New option.
7721 * doc/invoke.texi: Document -mindirect-branch-register option.
7723 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7725 * config/i386/i386-protos.h (ix86_output_function_return): New.
7726 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7727 set function_return_type.
7728 (indirect_thunk_name): Add ret_p to indicate thunk for function
7730 (output_indirect_thunk_function): Pass false to
7731 indirect_thunk_name.
7732 (ix86_output_indirect_branch_via_reg): Likewise.
7733 (ix86_output_indirect_branch_via_push): Likewise.
7734 (output_indirect_thunk_function): Create alias for function
7735 return thunk if regno < 0.
7736 (ix86_output_function_return): New function.
7737 (ix86_handle_fndecl_attribute): Handle function_return.
7738 (ix86_attribute_table): Add function_return.
7739 * config/i386/i386.h (machine_function): Add
7740 function_return_type.
7741 * config/i386/i386.md (simple_return_internal): Use
7742 ix86_output_function_return.
7743 (simple_return_internal_long): Likewise.
7744 * config/i386/i386.opt (mfunction-return=): New option.
7745 (indirect_branch): Mention -mfunction-return=.
7746 * doc/extend.texi: Document function_return function attribute.
7747 * doc/invoke.texi: Document -mfunction-return= option.
7749 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7751 * config/i386/i386-opts.h (indirect_branch): New.
7752 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7753 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7754 with local indirect jump when converting indirect call and jump.
7755 (ix86_set_indirect_branch_type): New.
7756 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7757 (indirectlabelno): New.
7758 (indirect_thunk_needed): Likewise.
7759 (indirect_thunk_bnd_needed): Likewise.
7760 (indirect_thunks_used): Likewise.
7761 (indirect_thunks_bnd_used): Likewise.
7762 (INDIRECT_LABEL): Likewise.
7763 (indirect_thunk_name): Likewise.
7764 (output_indirect_thunk): Likewise.
7765 (output_indirect_thunk_function): Likewise.
7766 (ix86_output_indirect_branch_via_reg): Likewise.
7767 (ix86_output_indirect_branch_via_push): Likewise.
7768 (ix86_output_indirect_branch): Likewise.
7769 (ix86_output_indirect_jmp): Likewise.
7770 (ix86_code_end): Call output_indirect_thunk_function if needed.
7771 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7773 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7774 (ix86_attribute_table): Add indirect_branch.
7775 * config/i386/i386.h (machine_function): Add indirect_branch_type
7776 and has_local_indirect_jump.
7777 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7779 (tablejump): Likewise.
7780 (*indirect_jump): Use ix86_output_indirect_jmp.
7781 (*tablejump_1): Likewise.
7782 (simple_return_indirect_internal): Likewise.
7783 * config/i386/i386.opt (mindirect-branch=): New option.
7784 (indirect_branch): New.
7787 (thunk-inline): Likewise.
7788 (thunk-extern): Likewise.
7789 * doc/extend.texi: Document indirect_branch function attribute.
7790 * doc/invoke.texi: Document -mindirect-branch= option.
7792 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7795 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7797 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7799 * ipa-inline.c (want_inline_small_function_p): Return false if
7800 inlining has already failed with CIF_FINAL_ERROR.
7801 (update_caller_keys): Call want_inline_small_function_p before
7803 (update_callee_keys): Likewise.
7805 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7807 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7809 (rs6000_quadword_masked_address_p): Likewise.
7810 (quad_aligned_load_p): Likewise.
7811 (quad_aligned_store_p): Likewise.
7812 (const_load_sequence_p): Add comment to describe the outer-most loop.
7813 (mimic_memory_attributes_and_flags): New function.
7814 (rs6000_gen_stvx): Likewise.
7815 (replace_swapped_aligned_store): Likewise.
7816 (rs6000_gen_lvx): Likewise.
7817 (replace_swapped_aligned_load): Likewise.
7818 (replace_swapped_load_constant): Capitalize argument name in
7819 comment describing this function.
7820 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7821 and stores that access quad-word aligned addresses and replace
7822 with stvx or lvx instructions when appropriate.
7823 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7824 New function prototype.
7825 (rs6000_quadword_masked_address_p): Likewise.
7826 (rs6000_gen_lvx): Likewise.
7827 (rs6000_gen_stvx): Likewise.
7828 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7829 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7830 when memory address is aligned.
7831 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7832 this split to select lvx instruction when memory address is aligned.
7833 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7834 instruction when memory address is aligned.
7835 (*vsx_le_perm_load_v16qi): Likewise.
7836 (four unnamed splitters): Modify to select the stvx instruction
7837 when memory is aligned.
7839 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7841 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7842 which appears in the queue multiple times.
7844 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7845 Alan Hayward <alan.hayward@arm.com>
7846 David Sherwood <david.sherwood@arm.com>
7848 * tree-vectorizer.h (vec_lower_bound): New structure.
7849 (_loop_vec_info): Add check_nonzero and lower_bounds.
7850 (LOOP_VINFO_CHECK_NONZERO): New macro.
7851 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7852 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7853 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7854 fields. Make seg_len the distance travelled, not including the
7856 (dr_direction_indicator): Declare.
7857 (dr_zero_step_indicator): Likewise.
7858 (dr_known_forward_stride_p): Likewise.
7859 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7861 (runtime_alias_check_p): Allow runtime alias checks with
7863 (operator ==): Compare access_size and align.
7864 (prune_runtime_alias_test_list): Rework for new distinction between
7865 the access_size and seg_len.
7866 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7868 (get_segment_min_max): New function.
7869 (create_intersect_range_checks): Use it.
7870 (dr_step_indicator): New function.
7871 (dr_direction_indicator): Likewise.
7872 (dr_zero_step_indicator): Likewise.
7873 (dr_known_forward_stride_p): Likewise.
7874 * tree-loop-distribution.c (data_ref_segment_size): Return
7875 DR_STEP * (niters - 1).
7876 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7878 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7879 (vect_preserves_scalar_order_p): New function, split out from...
7880 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7881 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7882 (vect_vfa_access_size): New function.
7883 (vect_vfa_align): Likewise.
7884 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7885 (dump_lower_bound): New function.
7886 (vect_check_lower_bound): Likewise.
7887 (vect_small_gap_p): Likewise.
7888 (vectorizable_with_step_bound_p): Likewise.
7889 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7890 depencies if the vectorization factor is 1. Convert the checks
7891 for nonzero steps into checks on the bounds of DR_STEP. Try using
7892 a bunds check for variable steps if the minimum required step is
7893 relatively small. Update calls to the dr_with_seg_len
7894 constructor and to vect_compile_time_alias.
7895 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7897 (vect_loop_versioning): Call it.
7898 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7900 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7902 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7903 Alan Hayward <alan.hayward@arm.com>
7904 David Sherwood <david.sherwood@arm.com>
7906 * doc/sourcebuild.texi (vect_scatter_store): Document.
7907 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7909 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7911 * genopinit.c (main): Add supports_vec_scatter_store and
7912 supports_vec_scatter_store_cached to target_optabs.
7913 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7914 IFN_MASK_SCATTER_STORE.
7915 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7917 * internal-fn.h (internal_store_fn_p): Declare.
7918 (internal_fn_stored_value_index): Likewise.
7919 * internal-fn.c (scatter_store_direct): New macro.
7920 (expand_scatter_store_optab_fn): New function.
7921 (direct_scatter_store_optab_supported_p): New macro.
7922 (internal_store_fn_p): New function.
7923 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7924 IFN_MASK_SCATTER_STORE.
7925 (internal_fn_mask_index): Likewise.
7926 (internal_fn_stored_value_index): New function.
7927 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7929 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7930 * optabs-query.c (supports_vec_scatter_store_p): New function.
7931 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7932 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7933 true for scatter stores.
7934 (vect_gather_scatter_fn_p): Handle scatter stores too.
7935 (vect_check_gather_scatter): Consider using scatter stores if
7936 supports_vec_scatter_store_p.
7937 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7939 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7940 internal_fn_stored_value_index.
7941 (check_load_store_masking): Handle scatter stores too.
7942 (vect_get_store_rhs): Make public.
7943 (vectorizable_call): Use internal_store_fn_p.
7944 (vectorizable_store): Handle scatter store internal functions.
7945 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7946 when deciding whether the end of the group has been reached.
7947 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7948 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7949 (mask_scatter_store<mode>): New insns.
7951 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7952 Alan Hayward <alan.hayward@arm.com>
7953 David Sherwood <david.sherwood@arm.com>
7955 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7956 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7957 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7959 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7960 Use vect_truncate_gather_scatter_offset if we can't treat the
7961 operation as a normal gather load or scatter store.
7962 (get_group_load_store_type): Take the gather_scatter_info
7963 as argument. Try using a gather load or scatter store for
7964 single-element groups.
7965 (get_load_store_type): Update calls to get_group_load_store_type
7966 and vect_use_strided_gather_scatters_p.
7968 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7969 Alan Hayward <alan.hayward@arm.com>
7970 David Sherwood <david.sherwood@arm.com>
7972 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7973 optional tree argument.
7974 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7976 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7977 but continue to use the current value as a fallback.
7978 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7979 to compare the updates.
7980 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7981 (get_load_store_type): Use it when handling a strided access.
7982 (vect_get_strided_load_store_ops): New function.
7983 (vect_get_data_ptr_increment): Likewise.
7984 (vectorizable_load): Handle strided gather loads. Always pass
7985 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7987 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7988 Alan Hayward <alan.hayward@arm.com>
7989 David Sherwood <david.sherwood@arm.com>
7991 * doc/md.texi (gather_load@var{m}): Document.
7992 (mask_gather_load@var{m}): Likewise.
7993 * genopinit.c (main): Add supports_vec_gather_load and
7994 supports_vec_gather_load_cached to target_optabs.
7995 * optabs-tree.c (init_tree_optimization_optabs): Use
7996 ggc_cleared_alloc to allocate target_optabs.
7997 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7998 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8000 * internal-fn.h (internal_load_fn_p): Declare.
8001 (internal_gather_scatter_fn_p): Likewise.
8002 (internal_fn_mask_index): Likewise.
8003 (internal_gather_scatter_fn_supported_p): Likewise.
8004 * internal-fn.c (gather_load_direct): New macro.
8005 (expand_gather_load_optab_fn): New function.
8006 (direct_gather_load_optab_supported_p): New macro.
8007 (direct_internal_fn_optab): New function.
8008 (internal_load_fn_p): Likewise.
8009 (internal_gather_scatter_fn_p): Likewise.
8010 (internal_fn_mask_index): Likewise.
8011 (internal_gather_scatter_fn_supported_p): Likewise.
8012 * optabs-query.c (supports_at_least_one_mode_p): New function.
8013 (supports_vec_gather_load_p): Likewise.
8014 * optabs-query.h (supports_vec_gather_load_p): Declare.
8015 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8016 and memory_type field.
8017 (NUM_PATTERNS): Bump to 15.
8018 * tree-vect-data-refs.c: Include internal-fn.h.
8019 (vect_gather_scatter_fn_p): New function.
8020 (vect_describe_gather_scatter_call): Likewise.
8021 (vect_check_gather_scatter): Try using internal functions for
8022 gather loads. Recognize existing calls to a gather load function.
8023 (vect_analyze_data_refs): Consider using gather loads if
8024 supports_vec_gather_load_p.
8025 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8026 (vect_get_gather_scatter_offset_type): Likewise.
8027 (vect_convert_mask_for_vectype): Likewise.
8028 (vect_add_conversion_to_patterm): Likewise.
8029 (vect_try_gather_scatter_pattern): Likewise.
8030 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8031 (vect_vect_recog_func_ptrs): Add it.
8032 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8033 internal_fn_mask_index and internal_gather_scatter_fn_p.
8034 (check_load_store_masking): Take the gather_scatter_info as an
8035 argument and handle gather loads.
8036 (vect_get_gather_scatter_ops): New function.
8037 (vectorizable_call): Check internal_load_fn_p.
8038 (vectorizable_load): Likewise. Handle gather load internal
8040 (vectorizable_store): Update call to check_load_store_masking.
8041 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8042 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8043 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8044 (aarch64_gather_scale_operand_d): New predicates.
8045 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8046 (mask_gather_load<mode>): New insns.
8048 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8049 Alan Hayward <alan.hayward@arm.com>
8050 David Sherwood <david.sherwood@arm.com>
8052 * optabs.def (fold_left_plus_optab): New optab.
8053 * doc/md.texi (fold_left_plus_@var{m}): Document.
8054 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8055 * internal-fn.c (fold_left_direct): Define.
8056 (expand_fold_left_optab_fn): Likewise.
8057 (direct_fold_left_optab_supported_p): Likewise.
8058 * fold-const-call.c (fold_const_fold_left): New function.
8059 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8060 * tree-parloops.c (valid_reduction_p): New function.
8061 (gather_scalar_reductions): Use it.
8062 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8063 (vect_finish_replace_stmt): Declare.
8064 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8065 (needs_fold_left_reduction_p): New function, split out from...
8066 (vect_is_simple_reduction): ...here. Accept reductions that
8067 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8068 (vect_force_simple_reduction): Also store the reduction type in
8069 the assignment's STMT_VINFO_REDUC_TYPE.
8070 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8071 (merge_with_identity): New function.
8072 (vect_expand_fold_left): Likewise.
8073 (vectorize_fold_left_reduction): Likewise.
8074 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8075 scalar phi in place for it. Check for target support and reject
8076 cases that would reassociate the operation. Defer the transform
8077 phase to vectorize_fold_left_reduction.
8078 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8079 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8080 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8082 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8084 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8085 call to ifc_temp_var.
8087 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8088 Alan Hayward <alan.hayward@arm.com>
8089 David Sherwood <david.sherwood@arm.com>
8091 * target.def (legitimize_address_displacement): Take the original
8092 offset as a poly_int.
8093 * targhooks.h (default_legitimize_address_displacement): Update
8095 * targhooks.c (default_legitimize_address_displacement): Likewise.
8096 * doc/tm.texi: Regenerate.
8097 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8098 as an argument, moving assert of ad->disp == ad->disp_term to...
8099 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8100 Try calling targetm.legitimize_address_displacement before expanding
8101 the address rather than afterwards, and adjust for the new interface.
8102 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8103 Match the new hook interface. Handle SVE addresses.
8104 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8107 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8109 * Makefile.in (OBJS): Add early-remat.o.
8110 * target.def (select_early_remat_modes): New hook.
8111 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8112 * doc/tm.texi: Regenerate.
8113 * targhooks.h (default_select_early_remat_modes): Declare.
8114 * targhooks.c (default_select_early_remat_modes): New function.
8115 * timevar.def (TV_EARLY_REMAT): New timevar.
8116 * passes.def (pass_early_remat): New pass.
8117 * tree-pass.h (make_pass_early_remat): Declare.
8118 * early-remat.c: New file.
8119 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8121 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8123 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8124 Alan Hayward <alan.hayward@arm.com>
8125 David Sherwood <david.sherwood@arm.com>
8127 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8128 vfm1 with a bound_epilog parameter.
8129 (vect_do_peeling): Update calls accordingly, and move the prologue
8130 call earlier in the function. Treat the base bound_epilog as 0 for
8131 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8132 this base when peeling for gaps.
8133 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8134 with fully-masked loops.
8135 (vect_estimate_min_profitable_iters): Handle the single peeled
8136 iteration in that case.
8138 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8139 Alan Hayward <alan.hayward@arm.com>
8140 David Sherwood <david.sherwood@arm.com>
8142 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8143 single-element interleaving even if the size is not a power of 2.
8144 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8145 accesses for single-element interleaving if the group size is
8148 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8149 Alan Hayward <alan.hayward@arm.com>
8150 David Sherwood <david.sherwood@arm.com>
8152 * doc/md.texi (fold_extract_last_@var{m}): Document.
8153 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8154 * optabs.def (fold_extract_last_optab): New optab.
8155 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8156 * internal-fn.c (fold_extract_direct): New macro.
8157 (expand_fold_extract_optab_fn): Likewise.
8158 (direct_fold_extract_optab_supported_p): Likewise.
8159 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8160 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8161 EXTRACT_LAST_REDUCTION.
8162 (get_initial_def_for_reduction): Do not create an initial vector
8163 for EXTRACT_LAST_REDUCTION reductions.
8164 (vectorizable_reduction): Leave the scalar phi in place for
8165 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8166 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8167 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8168 transform phase to vectorizable_condition.
8169 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8171 (vect_finish_stmt_generation): ...here.
8172 (vect_finish_replace_stmt): New function.
8173 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8174 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8176 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8178 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8179 Alan Hayward <alan.hayward@arm.com>
8180 David Sherwood <david.sherwood@arm.com>
8182 * doc/md.texi (extract_last_@var{m}): Document.
8183 * optabs.def (extract_last_optab): New optab.
8184 * internal-fn.def (EXTRACT_LAST): New internal function.
8185 * internal-fn.c (cond_unary_direct): New macro.
8186 (expand_cond_unary_optab_fn): Likewise.
8187 (direct_cond_unary_optab_supported_p): Likewise.
8188 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8189 loops using EXTRACT_LAST.
8190 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8191 (extract_last_<mode>): ...this optab.
8192 (vec_extract<mode><Vel>): Update accordingly.
8194 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8195 Alan Hayward <alan.hayward@arm.com>
8196 David Sherwood <david.sherwood@arm.com>
8198 * target.def (empty_mask_is_expensive): New hook.
8199 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8200 * doc/tm.texi: Regenerate.
8201 * targhooks.h (default_empty_mask_is_expensive): Declare.
8202 * targhooks.c (default_empty_mask_is_expensive): New function.
8203 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8204 if the target says that empty masks are expensive.
8205 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8207 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8209 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8210 Alan Hayward <alan.hayward@arm.com>
8211 David Sherwood <david.sherwood@arm.com>
8213 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8214 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8215 (vect_use_loop_mask_for_alignment_p): New function.
8216 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8217 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8218 niters_skip argument. Make sure that the first niters_skip elements
8219 of the first iteration are inactive.
8220 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8221 Update call to vect_set_loop_masks_directly.
8222 (get_misalign_in_elems): New function, split out from...
8223 (vect_gen_prolog_loop_niters): ...here.
8224 (vect_update_init_of_dr): Take a code argument that specifies whether
8225 the adjustment should be added or subtracted.
8226 (vect_update_init_of_drs): Likewise.
8227 (vect_prepare_for_masked_peels): New function.
8228 (vect_do_peeling): Skip prologue peeling if we're using a mask
8229 instead. Update call to vect_update_inits_of_drs.
8230 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8232 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8233 alignment. Do not include the number of peeled iterations in
8234 the minimum threshold in that case.
8235 (vectorizable_induction): Adjust the start value down by
8236 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8237 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8238 Take the number of skipped iterations into account when calculating
8240 * tree-vect-stmts.c (vect_gen_while_not): New function.
8242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8243 Alan Hayward <alan.hayward@arm.com>
8244 David Sherwood <david.sherwood@arm.com>
8246 * doc/sourcebuild.texi (vect_fully_masked): Document.
8247 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8249 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8251 (vect_analyze_loop_2): ...here. Don't check the vectorization
8252 factor against the number of loop iterations if the loop is
8255 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8256 Alan Hayward <alan.hayward@arm.com>
8257 David Sherwood <david.sherwood@arm.com>
8259 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8260 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8261 (dump_groups): Update accordingly.
8262 (iv_use::mem_type): New member variable.
8263 (address_p): New function.
8264 (record_use): Add a mem_type argument and initialize the new
8266 (record_group_use): Add a mem_type argument. Use address_p.
8267 Remove obsolete null checks of base_object. Update call to record_use.
8268 (find_interesting_uses_op): Update call to record_group_use.
8269 (find_interesting_uses_cond): Likewise.
8270 (find_interesting_uses_address): Likewise.
8271 (get_mem_type_for_internal_fn): New function.
8272 (find_address_like_use): Likewise.
8273 (find_interesting_uses_stmt): Try find_address_like_use before
8274 calling find_interesting_uses_op.
8275 (addr_offset_valid_p): Use the iv mem_type field as the type
8276 of the addressed memory.
8277 (add_autoinc_candidates): Likewise.
8278 (get_address_cost): Likewise.
8279 (split_small_address_groups_p): Use address_p.
8280 (split_address_groups): Likewise.
8281 (add_iv_candidate_for_use): Likewise.
8282 (autoinc_possible_for_pair): Likewise.
8283 (rewrite_groups): Likewise.
8284 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8285 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8286 (get_alias_ptr_type_for_ptr_address): New function.
8287 (rewrite_use_address): Rewrite address uses in calls that were
8288 identified by find_address_like_use.
8290 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8291 Alan Hayward <alan.hayward@arm.com>
8292 David Sherwood <david.sherwood@arm.com>
8294 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8296 * gimple-expr.h (is_gimple_addressable: Likewise.
8297 * gimple-expr.c (is_gimple_address): Likewise.
8298 * internal-fn.c (expand_call_mem_ref): New function.
8299 (expand_mask_load_optab_fn): Use it.
8300 (expand_mask_store_optab_fn): Likewise.
8302 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8303 Alan Hayward <alan.hayward@arm.com>
8304 David Sherwood <david.sherwood@arm.com>
8306 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8307 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8308 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8309 (cond_umax@var{mode}): Document.
8310 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8311 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8312 (cond_umin_optab, cond_umax_optab): New optabs.
8313 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8314 (COND_IOR, COND_XOR): New internal functions.
8315 * internal-fn.h (get_conditional_internal_fn): Declare.
8316 * internal-fn.c (cond_binary_direct): New macro.
8317 (expand_cond_binary_optab_fn): Likewise.
8318 (direct_cond_binary_optab_supported_p): Likewise.
8319 (get_conditional_internal_fn): New function.
8320 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8321 Cope with reduction statements that are vectorized as calls rather
8323 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8324 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8325 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8326 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8327 (UNSPEC_COND_EOR): New unspecs.
8328 (optab): Add mappings for them.
8329 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8330 (sve_int_op, sve_fp_op): New int attributes.
8332 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8333 Alan Hayward <alan.hayward@arm.com>
8334 David Sherwood <david.sherwood@arm.com>
8336 * optabs.def (while_ult_optab): New optab.
8337 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8338 * internal-fn.def (WHILE_ULT): New internal function.
8339 * internal-fn.h (direct_internal_fn_supported_p): New override
8340 that takes two types as argument.
8341 * internal-fn.c (while_direct): New macro.
8342 (expand_while_optab_fn): New function.
8343 (convert_optab_supported_p): Likewise.
8344 (direct_while_optab_supported_p): New macro.
8345 * wide-int.h (wi::udiv_ceil): New function.
8346 * tree-vectorizer.h (rgroup_masks): New structure.
8347 (vec_loop_masks): New typedef.
8348 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8350 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8351 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8352 (vect_max_vf): New function.
8353 (slpeel_make_loop_iterate_ntimes): Delete.
8354 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8355 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8356 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8357 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8358 internal-fn.h, stor-layout.h and optabs-query.h.
8359 (vect_set_loop_mask): New function.
8360 (add_preheader_seq): Likewise.
8361 (add_header_seq): Likewise.
8362 (interleave_supported_p): Likewise.
8363 (vect_maybe_permute_loop_masks): Likewise.
8364 (vect_set_loop_masks_directly): Likewise.
8365 (vect_set_loop_condition_masked): Likewise.
8366 (vect_set_loop_condition_unmasked): New function, split out from
8367 slpeel_make_loop_iterate_ntimes.
8368 (slpeel_make_loop_iterate_ntimes): Rename to..
8369 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8370 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8371 (vect_do_peeling): Update call accordingly.
8372 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8374 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8375 mask_compare_type, can_fully_mask_p and fully_masked_p.
8376 (release_vec_loop_masks): New function.
8377 (_loop_vec_info): Use it to free the loop masks.
8378 (can_produce_all_loop_masks_p): New function.
8379 (vect_get_max_nscalars_per_iter): Likewise.
8380 (vect_verify_full_masking): Likewise.
8381 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8382 retries, and free the mask rgroups before retrying. Check loop-wide
8383 reasons for disallowing fully-masked loops. Make the final decision
8384 about whether use a fully-masked loop or not.
8385 (vect_estimate_min_profitable_iters): Do not assume that peeling
8386 for the number of iterations will be needed for fully-masked loops.
8387 (vectorizable_reduction): Disable fully-masked loops.
8388 (vectorizable_live_operation): Likewise.
8389 (vect_halve_mask_nunits): New function.
8390 (vect_double_mask_nunits): Likewise.
8391 (vect_record_loop_mask): Likewise.
8392 (vect_get_loop_mask): Likewise.
8393 (vect_transform_loop): Handle the case in which the final loop
8394 iteration might handle a partial vector. Call vect_set_loop_condition
8395 instead of slpeel_make_loop_iterate_ntimes.
8396 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8397 (check_load_store_masking): New function.
8398 (prepare_load_store_mask): Likewise.
8399 (vectorizable_store): Handle fully-masked loops.
8400 (vectorizable_load): Likewise.
8401 (supportable_widening_operation): Use vect_halve_mask_nunits for
8403 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8404 (vect_gen_while): New function.
8405 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8406 (aarch64_uqdec<mode>): New insn.
8408 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8409 Alan Hayward <alan.hayward@arm.com>
8410 David Sherwood <david.sherwood@arm.com>
8412 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8413 (reduc_xor_scal_optab): New optabs.
8414 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8415 (reduc_xor_scal_@var{m}): Document.
8416 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8417 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8419 * fold-const-call.c (fold_const_call): Handle them.
8420 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8421 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8422 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8423 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8424 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8425 (UNSPEC_XORV): New unspecs.
8426 (optab): Add entries for them.
8427 (BITWISEV): New int iterator.
8428 (bit_reduc_op): New int attributes.
8430 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8431 Alan Hayward <alan.hayward@arm.com>
8432 David Sherwood <david.sherwood@arm.com>
8434 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8435 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8436 * optabs.def (vec_shl_insert_optab): New optab.
8437 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8438 (duplicate_and_interleave): Likewise.
8439 * tree-vect-loop.c: Include internal-fn.h.
8440 (neutral_op_for_slp_reduction): New function, split out from
8441 get_initial_defs_for_reduction.
8442 (get_initial_def_for_reduction): Handle option 2 for variable-length
8443 vectors by loading the neutral value into a vector and then shifting
8444 the initial value into element 0.
8445 (get_initial_defs_for_reduction): Replace the code argument with
8446 the neutral value calculated by neutral_op_for_slp_reduction.
8447 Use gimple_build_vector for constant-length vectors.
8448 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8449 but the first group_size elements have a neutral value.
8450 Use duplicate_and_interleave otherwise.
8451 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8452 Update call to get_initial_defs_for_reduction. Handle SLP
8453 reductions for variable-length vectors by creating one vector
8454 result for each scalar result, with the elements associated
8455 with other scalar results stubbed out with the neutral value.
8456 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8457 Require IFN_VEC_SHL_INSERT for double reductions on
8458 variable-length vectors, or SLP reductions that have
8459 a neutral value. Require can_duplicate_and_interleave_p
8460 support for variable-length unchained SLP reductions if there
8461 is no neutral value, such as for MIN/MAX reductions. Also require
8462 the number of vector elements to be a multiple of the number of
8463 SLP statements when doing variable-length unchained SLP reductions.
8464 Update call to vect_create_epilog_for_reduction.
8465 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8466 and remove initial values.
8467 (duplicate_and_interleave): Make public.
8468 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8469 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8471 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8472 Alan Hayward <alan.hayward@arm.com>
8473 David Sherwood <david.sherwood@arm.com>
8475 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8476 (can_duplicate_and_interleave_p): New function.
8477 (vect_get_and_check_slp_defs): Take the vector of statements
8478 rather than just the current one. Remove excess parentheses.
8479 Restriction rejectinon of vect_constant_def and vect_external_def
8480 for variable-length vectors to boolean types, or types for which
8481 can_duplicate_and_interleave_p is false.
8482 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8483 (duplicate_and_interleave): New function.
8484 (vect_get_constant_vectors): Use gimple_build_vector for
8485 constant-length vectors and suitable variable-length constant
8486 vectors. Use duplicate_and_interleave for other variable-length
8487 vectors. Don't defer the update when inserting new statements.
8489 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8490 Alan Hayward <alan.hayward@arm.com>
8491 David Sherwood <david.sherwood@arm.com>
8493 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8494 min_profitable_iters doesn't go negative.
8496 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8497 Alan Hayward <alan.hayward@arm.com>
8498 David Sherwood <david.sherwood@arm.com>
8500 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8501 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8502 * optabs.def (vec_mask_load_lanes_optab): New optab.
8503 (vec_mask_store_lanes_optab): Likewise.
8504 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8505 (MASK_STORE_LANES): Likewise.
8506 * internal-fn.c (mask_load_lanes_direct): New macro.
8507 (mask_store_lanes_direct): Likewise.
8508 (expand_mask_load_optab_fn): Handle masked operations.
8509 (expand_mask_load_lanes_optab_fn): New macro.
8510 (expand_mask_store_optab_fn): Handle masked operations.
8511 (expand_mask_store_lanes_optab_fn): New macro.
8512 (direct_mask_load_lanes_optab_supported_p): Likewise.
8513 (direct_mask_store_lanes_optab_supported_p): Likewise.
8514 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8516 (vect_load_lanes_supported): Likewise.
8517 * tree-vect-data-refs.c (strip_conversion): New function.
8518 (can_group_stmts_p): Likewise.
8519 (vect_analyze_data_ref_accesses): Use it instead of checking
8520 for a pair of assignments.
8521 (vect_store_lanes_supported): Take a masked_p parameter.
8522 (vect_load_lanes_supported): Likewise.
8523 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8524 vect_store_lanes_supported and vect_load_lanes_supported.
8525 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8526 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8527 parameter. Don't allow gaps for masked accesses.
8528 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8529 and vect_load_lanes_supported.
8530 (get_load_store_type): Take a masked_p parameter and update
8531 call to get_group_load_store_type.
8532 (vectorizable_store): Update call to get_load_store_type.
8533 Handle IFN_MASK_STORE_LANES.
8534 (vectorizable_load): Update call to get_load_store_type.
8535 Handle IFN_MASK_LOAD_LANES.
8537 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8538 Alan Hayward <alan.hayward@arm.com>
8539 David Sherwood <david.sherwood@arm.com>
8541 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8543 * config/aarch64/aarch64-protos.h
8544 (aarch64_sve_struct_memory_operand_p): Declare.
8545 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8546 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8547 (VPRED, vpred): Handle SVE structure modes.
8548 * config/aarch64/constraints.md (Utx): New constraint.
8549 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8550 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8551 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8552 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8553 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8554 structure modes. Split into pieces after RA.
8555 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8556 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8558 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8559 SVE structure modes.
8560 (aarch64_classify_address): Likewise.
8561 (sizetochar): Move earlier in file.
8562 (aarch64_print_operand): Handle SVE register lists.
8563 (aarch64_array_mode): New function.
8564 (aarch64_sve_struct_memory_operand_p): Likewise.
8565 (TARGET_ARRAY_MODE): Redefine.
8567 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8568 Alan Hayward <alan.hayward@arm.com>
8569 David Sherwood <david.sherwood@arm.com>
8571 * target.def (array_mode): New target hook.
8572 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8573 * doc/tm.texi: Regenerate.
8574 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8575 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8576 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8578 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8581 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8582 Alan Hayward <alan.hayward@arm.com>
8583 David Sherwood <david.sherwood@arm.com>
8585 * fold-const.c (fold_binary_loc): Check the argument types
8586 rather than the result type when testing for a vector operation.
8588 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8590 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8591 * doc/tm.texi: Regenerate.
8593 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8594 Alan Hayward <alan.hayward@arm.com>
8595 David Sherwood <david.sherwood@arm.com>
8597 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8598 (sve): Document new AArch64 extension.
8599 * doc/md.texi (w): Extend the description of the AArch64
8600 constraint to include SVE vectors.
8601 (Upl, Upa): Document new AArch64 predicate constraints.
8602 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8604 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8605 (msve-vector-bits=): New option.
8606 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8607 SVE when these are disabled.
8608 (sve): New extension.
8609 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8610 modes. Adjust their number of units based on aarch64_sve_vg.
8611 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8612 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8613 aarch64_addr_query_type.
8614 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8615 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8616 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8617 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8618 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8619 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8620 (aarch64_simd_imm_zero_p): Delete.
8621 (aarch64_check_zero_based_sve_index_immediate): Declare.
8622 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8623 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8624 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8625 (aarch64_sve_float_mul_immediate_p): Likewise.
8626 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8628 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8629 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8630 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8631 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8632 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8633 (aarch64_regmode_natural_size): Likewise.
8634 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8635 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8637 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8638 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8639 for VG and the SVE predicate registers.
8640 (V_ALIASES): Add a "z"-prefixed alias.
8641 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8642 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8643 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8644 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8645 (REG_CLASS_NAMES): Add entries for them.
8646 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8647 and the predicate registers.
8648 (aarch64_sve_vg): Declare.
8649 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8650 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8651 (REGMODE_NATURAL_SIZE): Define.
8652 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8654 * config/aarch64/aarch64.c: Include cfgrtl.h.
8655 (simd_immediate_info): Add a constructor for series vectors,
8656 and an associated step field.
8657 (aarch64_sve_vg): New variable.
8658 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8659 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8660 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8661 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8662 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8663 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8664 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8665 (aarch64_get_mask_mode): New functions.
8666 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8667 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8668 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8669 predicate modes and predicate registers. Explicitly restrict
8670 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8671 to store a vector mode if it is recognized by
8672 aarch64_classify_vector_mode.
8673 (aarch64_regmode_natural_size): New function.
8674 (aarch64_hard_regno_caller_save_mode): Return the original mode
8676 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8677 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8678 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8679 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8681 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8682 does not overlap dest if the function is frame-related. Handle
8684 (aarch64_split_add_offset): New function.
8685 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8686 them aarch64_add_offset.
8687 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8688 and update call to aarch64_sub_sp.
8689 (aarch64_add_cfa_expression): New function.
8690 (aarch64_expand_prologue): Pass extra temporary registers to the
8691 functions above. Handle the case in which we need to emit new
8692 DW_CFA_expressions for registers that were originally saved
8693 relative to the stack pointer, but now have to be expressed
8694 relative to the frame pointer.
8695 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8697 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8698 IP0 and IP1 values for SVE frames.
8699 (aarch64_expand_vec_series): New function.
8700 (aarch64_expand_sve_widened_duplicate): Likewise.
8701 (aarch64_expand_sve_const_vector): Likewise.
8702 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8703 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8704 into the register, rather than emitting a SET directly.
8705 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8706 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8707 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8708 (offset_9bit_signed_scaled_p): New functions.
8709 (aarch64_replicate_bitmask_imm): New function.
8710 (aarch64_bitmask_imm): Use it.
8711 (aarch64_cannot_force_const_mem): Reject expressions involving
8712 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8713 (aarch64_classify_index): Handle SVE indices, by requiring
8714 a plain register index with a scale that matches the element size.
8715 (aarch64_classify_address): Handle SVE addresses. Assert that
8716 the mode of the address is VOIDmode or an integer mode.
8717 Update call to aarch64_classify_symbol.
8718 (aarch64_classify_symbolic_expression): Update call to
8719 aarch64_classify_symbol.
8720 (aarch64_const_vec_all_in_range_p): New function.
8721 (aarch64_print_vector_float_operand): Likewise.
8722 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8723 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8724 and the FP immediates 1.0 and 0.5.
8725 (aarch64_print_address_internal): Handle SVE addresses.
8726 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8727 (aarch64_regno_regclass): Handle predicate registers.
8728 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8730 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8731 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8732 (aarch64_convert_sve_vector_bits): New function.
8733 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8734 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8736 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8737 Handle SVE vector and predicate modes. Accept VL-based constants
8738 that need only one temporary register, and VL offsets that require
8739 no temporary registers.
8740 (aarch64_conditional_register_usage): Mark the predicate registers
8741 as fixed if SVE isn't available.
8742 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8743 Return true for SVE vector and predicate modes.
8744 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8745 rather than an unsigned int. Handle SVE modes.
8746 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8748 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8750 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8751 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8752 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8753 (aarch64_sve_float_mul_immediate_p): New functions.
8754 (aarch64_sve_valid_immediate): New function.
8755 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8756 Explicitly reject structure modes. Check for INDEX constants.
8757 Handle PTRUE and PFALSE constants.
8758 (aarch64_check_zero_based_sve_index_immediate): New function.
8759 (aarch64_simd_imm_zero_p): Delete.
8760 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8761 vector modes. Accept constants in the range of CNT[BHWD].
8762 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8763 ask for an Advanced SIMD mode.
8764 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8765 (aarch64_simd_vector_alignment): Handle SVE predicates.
8766 (aarch64_vectorize_preferred_vector_alignment): New function.
8767 (aarch64_simd_vector_alignment_reachable): Use it instead of
8769 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8770 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8772 (MAX_VECT_LEN): Delete.
8773 (expand_vec_perm_d): Add a vec_flags field.
8774 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8775 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8776 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8778 (aarch64_evpc_rev): Rename to...
8779 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8780 (aarch64_evpc_rev_global): New function.
8781 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8782 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8784 (aarch64_evpc_sve_tbl): New function.
8785 (aarch64_expand_vec_perm_const_1): Update after rename of
8786 aarch64_evpc_rev. Handle SVE permutes too, trying
8787 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8788 than aarch64_evpc_tbl.
8789 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8790 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8791 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8792 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8793 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8794 (aarch64_expand_sve_vcond): New functions.
8795 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8796 of aarch64_vector_mode_p.
8797 (aarch64_dwarf_poly_indeterminate_value): New function.
8798 (aarch64_compute_pressure_classes): Likewise.
8799 (aarch64_can_change_mode_class): Likewise.
8800 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8801 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8802 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8803 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8804 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8805 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8806 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8807 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8809 (Dn, Dl, Dr): Accept const as well as const_vector.
8810 (Dz): Likewise. Compare against CONST0_RTX.
8811 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8812 of "vector" where appropriate.
8813 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8814 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8815 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8816 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8817 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8818 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8819 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8820 (v_int_equiv): Extend to SVE modes.
8821 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8823 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8824 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8825 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8826 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8827 (SVE_COND_FP_CMP): New int iterators.
8828 (perm_hilo): Handle the new unpack unspecs.
8829 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8831 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8832 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8833 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8834 (aarch64_equality_operator, aarch64_constant_vector_operand)
8835 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8836 (aarch64_sve_nonimmediate_operand): Likewise.
8837 (aarch64_sve_general_operand): Likewise.
8838 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8839 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8840 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8841 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8842 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8843 (aarch64_sve_float_arith_immediate): Likewise.
8844 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8845 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8846 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8847 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8848 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8849 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8850 (aarch64_sve_float_arith_operand): Likewise.
8851 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8852 (aarch64_sve_float_mul_operand): Likewise.
8853 (aarch64_sve_vec_perm_operand): Likewise.
8854 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8855 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8856 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8857 as well as const_vector.
8858 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8859 in file. Use CONST0_RTX and CONSTM1_RTX.
8860 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8861 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8862 Use aarch64_simd_imm_zero.
8863 * config/aarch64/aarch64-sve.md: New file.
8864 * config/aarch64/aarch64.md: Include it.
8865 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8866 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8867 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8868 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8869 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8870 (sve): New attribute.
8871 (enabled): Disable instructions with the sve attribute unless
8873 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8874 aarch64_expand_mov_immediate.
8875 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8876 CNT[BHSD] immediates.
8877 (movti): Split CONST_POLY_INT moves into two halves.
8878 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8879 Split additions that need a temporary here if the destination
8880 is the stack pointer.
8881 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8882 (*add<mode>3_poly_1): New instruction.
8883 (set_clobber_cc): New expander.
8885 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8887 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8888 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8889 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8890 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8891 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8892 Change innermode from fixed_mode_size to machine_mode.
8893 (simplify_subreg): Update call accordingly. Handle a constant-sized
8894 subreg of a variable-length CONST_VECTOR.
8896 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8897 Alan Hayward <alan.hayward@arm.com>
8898 David Sherwood <david.sherwood@arm.com>
8900 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8901 (add_offset_to_base): New function, split out from...
8902 (create_mem_ref): ...here. When handling a scale other than 1,
8903 check first whether the address is valid without the offset.
8904 Add it into the base if so, leaving the index and scale as-is.
8906 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8909 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8910 fold_for_warn before checking if arg2 is INTEGER_CST.
8912 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8914 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8915 (store_multiple_operation): Delete.
8916 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8917 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8918 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8919 guarded by TARGET_STRING.
8920 (rs6000_output_load_multiple): Delete.
8921 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8922 OPTION_MASK_STRING / TARGET_STRING handling.
8923 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8924 (const rs6000_opt_masks) <"string">: Change mask to 0.
8925 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8926 (MASK_STRING): Delete.
8927 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8929 (load_multiple): Delete.
8936 (store_multiple): Delete.
8943 (movmemsi_8reg): Delete.
8944 (corresponding unnamed define_insn): Delete.
8945 (movmemsi_6reg): Delete.
8946 (corresponding unnamed define_insn): Delete.
8947 (movmemsi_4reg): Delete.
8948 (corresponding unnamed define_insn): Delete.
8949 (movmemsi_2reg): Delete.
8950 (corresponding unnamed define_insn): Delete.
8951 (movmemsi_1reg): Delete.
8952 (corresponding unnamed define_insn): Delete.
8953 * config/rs6000/rs6000.opt (mno-string): New.
8954 (mstring): Replace by deprecation warning stub.
8955 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8957 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8959 * regrename.c (regrename_do_replace): If replacing the same
8960 reg multiple times, try to reuse last created gen_raw_REG.
8963 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8964 main to workaround a bug in GDB.
8966 2018-01-12 Tom de Vries <tom@codesourcery.com>
8969 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8971 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8973 PR rtl-optimization/80481
8974 * ira-color.c (get_cap_member): New function.
8975 (allocnos_conflict_by_live_ranges_p): Use it.
8976 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8977 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8979 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8982 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8983 (*saddl_se_1): Ditto.
8985 (*ssubl_se_1): Ditto.
8987 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8989 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8990 rather than wi::to_widest for DR_INITs.
8991 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8992 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8993 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8995 (vect_analyze_group_access_1): Note that here.
8997 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8999 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9000 polynomial type sizes.
9002 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9004 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9005 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9006 (gimple_add_tmp_var): Likewise.
9008 2018-01-12 Martin Liska <mliska@suse.cz>
9010 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9011 (gimple_alloc_sizes): Likewise.
9012 (dump_gimple_statistics): Use PRIu64 in printf format.
9013 * gimple.h: Change uint64_t to int.
9015 2018-01-12 Martin Liska <mliska@suse.cz>
9017 * tree-core.h: Use uint64_t instead of int.
9018 * tree.c (tree_node_counts): Likewise.
9019 (tree_node_sizes): Likewise.
9020 (dump_tree_statistics): Use PRIu64 in printf format.
9022 2018-01-12 Martin Liska <mliska@suse.cz>
9024 * Makefile.in: As qsort_chk is implemented in vec.c, add
9025 vec.o to linkage of gencfn-macros.
9026 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9027 passing the info to record_node_allocation_statistics.
9028 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9030 * ggc-common.c (struct ggc_usage): Add operator== and use
9031 it in operator< and compare function.
9032 * mem-stats.h (struct mem_usage): Likewise.
9033 * vec.c (struct vec_usage): Remove operator< and compare
9034 function. Can be simply inherited.
9036 2018-01-12 Martin Jambor <mjambor@suse.cz>
9039 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9040 * tree-ssa-math-opts.c: Include domwalk.h.
9041 (convert_mult_to_fma_1): New function.
9042 (fma_transformation_info): New type.
9043 (fma_deferring_state): Likewise.
9044 (cancel_fma_deferring): New function.
9045 (result_of_phi): Likewise.
9046 (last_fma_candidate_feeds_initial_phi): Likewise.
9047 (convert_mult_to_fma): Added deferring logic, split actual
9048 transformation to convert_mult_to_fma_1.
9049 (math_opts_dom_walker): New type.
9050 (math_opts_dom_walker::after_dom_children): New method, body moved
9051 here from pass_optimize_widening_mul::execute, added deferring logic
9053 (pass_optimize_widening_mul::execute): Moved most of code to
9054 math_opts_dom_walker::after_dom_children.
9055 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9056 * config/i386/i386.c (ix86_option_override_internal): Added
9057 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9059 2018-01-12 Richard Biener <rguenther@suse.de>
9062 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9063 inline instance vars.
9065 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9068 * config/rx/rx.c (rx_is_restricted_memory_address):
9071 2018-01-12 Richard Biener <rguenther@suse.de>
9073 PR tree-optimization/80846
9074 * target.def (split_reduction): New target hook.
9075 * targhooks.c (default_split_reduction): New function.
9076 * targhooks.h (default_split_reduction): Declare.
9077 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9078 target requests first reduce vectors by combining low and high
9080 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9081 (get_vectype_for_scalar_type_and_size): Export.
9082 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9083 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9084 * doc/tm.texi: Regenerate.
9085 * config/i386/i386.c (ix86_split_reduction): Implement
9086 TARGET_VECTORIZE_SPLIT_REDUCTION.
9088 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9091 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9092 in PIC mode except for TARGET_VXWORKS_RTP.
9093 * config/sparc/sparc.c: Include cfgrtl.h.
9094 (TARGET_INIT_PIC_REG): Define.
9095 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9096 (sparc_pic_register_p): New predicate.
9097 (sparc_legitimate_address_p): Use it.
9098 (sparc_legitimize_pic_address): Likewise.
9099 (sparc_delegitimize_address): Likewise.
9100 (sparc_mode_dependent_address_p): Likewise.
9101 (gen_load_pcrel_sym): Remove 4th parameter.
9102 (load_got_register): Adjust call to above. Remove obsolete stuff.
9103 (sparc_expand_prologue): Do not call load_got_register here.
9104 (sparc_flat_expand_prologue): Likewise.
9105 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9106 (sparc_use_pseudo_pic_reg): New function.
9107 (sparc_init_pic_reg): Likewise.
9108 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9109 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9111 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9113 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9114 Add item for branch_cost.
9116 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9118 PR rtl-optimization/83565
9119 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9120 not extend the result to a larger mode for rotate operations.
9121 (num_sign_bit_copies1): Likewise.
9123 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9126 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9128 Use values-Xc.o for -pedantic.
9129 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9131 2018-01-12 Martin Liska <mliska@suse.cz>
9134 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9136 (possible_polymorphic_call_targets): Use it.
9137 (ipa_devirt): Likewise.
9139 2018-01-12 Martin Liska <mliska@suse.cz>
9141 * profile-count.h (enum profile_quality): Use 0 as invalid
9142 enum value of profile_quality.
9144 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9146 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9147 -mext-string options.
9149 2018-01-12 Richard Biener <rguenther@suse.de>
9151 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9152 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9153 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9155 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9157 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9159 * configure.ac (--with-long-double-format): Add support for the
9160 configuration option to change the default long double format on
9162 * config.gcc (powerpc*-linux*-*): Likewise.
9163 * configure: Regenerate.
9164 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9165 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9166 used without modification.
9168 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9170 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9171 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9172 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9173 MISC_BUILTIN_SPEC_BARRIER.
9174 (rs6000_init_builtins): Likewise.
9175 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9177 (speculation_barrier): New define_insn.
9178 * doc/extend.texi: Document __builtin_speculation_barrier.
9180 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9183 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9184 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9185 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9187 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9188 integral modes instead of "ss" and "sd".
9189 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9190 vectors with 32-bit and 64-bit elements.
9191 (vecdupssescalarmodesuffix): New mode attribute.
9192 (vec_dup<mode>): Use it.
9194 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9197 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9198 frame if argument is passed on stack.
9200 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9203 * ree.c (combine_reaching_defs): Optimize also
9204 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9205 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9207 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9210 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9212 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9215 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9216 after they are computed.
9218 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9220 PR tree-optimization/83695
9221 * gimple-loop-linterchange.cc
9222 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9223 reset cached scev information after interchange.
9224 (pass_linterchange::execute): Remove call to scev_reset_htab.
9226 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9228 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9229 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9230 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9231 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9232 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9233 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9234 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9235 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9236 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9237 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9238 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9239 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9240 (V_lane_reg): Likewise.
9241 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9243 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9244 (vfmal_lane_low<mode>_intrinsic,
9245 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9246 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9247 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9248 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9249 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9250 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9252 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9254 * config/arm/arm-cpus.in (fp16fml): New feature.
9255 (ALL_SIMD): Add fp16fml.
9256 (armv8.2-a): Add fp16fml as an option.
9257 (armv8.3-a): Likewise.
9258 (armv8.4-a): Add fp16fml as part of fp16.
9259 * config/arm/arm.h (TARGET_FP16FML): Define.
9260 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9262 * config/arm/arm-modes.def (V2HF): Define.
9263 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9264 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9265 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9266 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9267 vfmsl_low, vfmsl_high): New set of builtins.
9268 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9269 (vfml_op): New code attribute.
9270 (VFMLHALVES): New int iterator.
9271 (VFML, VFMLSEL): New mode attributes.
9272 (V_reg): Define mapping for V2HF.
9273 (V_hi, V_lo): New mode attributes.
9274 (VF_constraint): Likewise.
9275 (vfml_half, vfml_half_selector): New int attributes.
9276 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9278 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9279 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9281 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9282 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9283 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9284 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9286 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9287 Document new effective target and option set.
9289 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9291 * config/arm/arm-cpus.in (armv8_4): New feature.
9292 (ARMv8_4a): New fgroup.
9293 (armv8.4-a): New arch.
9294 * config/arm/arm-tables.opt: Regenerate.
9295 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9296 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9297 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9298 Add matching rules for -march=armv8.4-a and extensions.
9299 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9301 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9304 * config/rx/rx.md (BW): New mode attribute.
9305 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9307 2018-01-11 Richard Biener <rguenther@suse.de>
9309 PR tree-optimization/83435
9310 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9311 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9312 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9314 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9315 Alan Hayward <alan.hayward@arm.com>
9316 David Sherwood <david.sherwood@arm.com>
9318 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9320 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9321 (aarch64_print_address_internal): Use it to check for a zero offset.
9323 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9324 Alan Hayward <alan.hayward@arm.com>
9325 David Sherwood <david.sherwood@arm.com>
9327 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9328 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9329 Return a poly_int64 rather than a HOST_WIDE_INT.
9330 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9331 rather than a HOST_WIDE_INT.
9332 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9333 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9334 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9335 final_offset from HOST_WIDE_INT to poly_int64.
9336 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9337 to_constant when getting the number of units in an Advanced SIMD
9339 (aarch64_builtin_vectorized_function): Check for a constant number
9341 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9343 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9344 attribute instead of GET_MODE_NUNITS.
9345 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9346 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9347 GET_MODE_SIZE for fixed-size registers.
9348 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9349 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9350 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9351 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9352 (aarch64_print_operand, aarch64_print_address_internal)
9353 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9354 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9355 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9356 Handle polynomial GET_MODE_SIZE.
9357 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9358 wider than SImode without modification.
9359 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9360 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9361 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9362 passing and returning SVE modes.
9363 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9364 rather than GEN_INT.
9365 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9366 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9367 (aarch64_allocate_and_probe_stack_space): Likewise.
9368 (aarch64_layout_frame): Cope with polynomial offsets.
9369 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9370 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9372 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9373 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9374 poly_int64 rather than a HOST_WIDE_INT.
9375 (aarch64_get_separate_components, aarch64_process_components)
9376 (aarch64_expand_prologue, aarch64_expand_epilogue)
9377 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9378 (aarch64_anchor_offset): New function, split out from...
9379 (aarch64_legitimize_address): ...here.
9380 (aarch64_builtin_vectorization_cost): Handle polynomial
9381 TYPE_VECTOR_SUBPARTS.
9382 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9384 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9385 number of elements from the PARALLEL rather than the mode.
9386 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9387 rather than GET_MODE_BITSIZE.
9388 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9389 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9390 (aarch64_expand_vec_perm_const_1): Handle polynomial
9391 d->perm.length () and d->perm elements.
9392 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9393 Apply to_constant to d->perm elements.
9394 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9395 polynomial CONST_VECTOR_NUNITS.
9396 (aarch64_move_pointer): Take amount as a poly_int64 rather
9398 (aarch64_progress_pointer): Avoid temporary variable.
9399 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9400 the mode attribute instead of GET_MODE.
9402 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9403 Alan Hayward <alan.hayward@arm.com>
9404 David Sherwood <david.sherwood@arm.com>
9406 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9407 x exists before using it.
9408 (aarch64_add_constant_internal): Rename to...
9409 (aarch64_add_offset_1): ...this. Replace regnum with separate
9410 src and dest rtxes. Handle the case in which they're different,
9411 including when the offset is zero. Replace scratchreg with an rtx.
9412 Use 2 additions if there is no spare register into which we can
9413 move a 16-bit constant.
9414 (aarch64_add_constant): Delete.
9415 (aarch64_add_offset): Replace reg with separate src and dest
9416 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9417 Use aarch64_add_offset_1.
9418 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9419 an rtx rather than an int. Take the delta as a poly_int64
9420 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9421 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9422 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9423 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9424 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9426 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9427 aarch64_add_constant.
9429 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9431 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9432 Use scalar_float_mode.
9434 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9436 * config/aarch64/aarch64-simd.md
9437 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9438 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9439 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9440 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9441 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9442 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9443 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9444 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9445 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9446 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9448 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9451 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9452 targ_options->x_arm_arch_string is non NULL.
9454 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9456 * config/aarch64/aarch64.h
9457 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9459 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9462 * expmed.c (emit_store_flag_force): Swap if const op0
9463 and change VOIDmode to mode of op0.
9465 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9467 PR rtl-optimization/83761
9468 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9469 than bytes to mode_for_size.
9471 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9474 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9475 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9478 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9481 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9482 when in layout mode.
9483 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9484 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9487 2018-01-10 Michael Collison <michael.collison@arm.com>
9489 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9490 * config/aarch64/aarch64-option-extension.def: Add
9491 AARCH64_OPT_EXTENSION of 'fp16fml'.
9492 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9493 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9494 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9495 * config/aarch64/constraints.md (Ui7): New constraint.
9496 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9497 (VFMLA_SEL_W): Ditto.
9500 (VFMLA16_LOW): New int iterator.
9501 (VFMLA16_HIGH): Ditto.
9502 (UNSPEC_FMLAL): New unspec.
9503 (UNSPEC_FMLSL): Ditto.
9504 (UNSPEC_FMLAL2): Ditto.
9505 (UNSPEC_FMLSL2): Ditto.
9506 (f16mac): New code attribute.
9507 * config/aarch64/aarch64-simd-builtins.def
9508 (aarch64_fmlal_lowv2sf): Ditto.
9509 (aarch64_fmlsl_lowv2sf): Ditto.
9510 (aarch64_fmlalq_lowv4sf): Ditto.
9511 (aarch64_fmlslq_lowv4sf): Ditto.
9512 (aarch64_fmlal_highv2sf): Ditto.
9513 (aarch64_fmlsl_highv2sf): Ditto.
9514 (aarch64_fmlalq_highv4sf): Ditto.
9515 (aarch64_fmlslq_highv4sf): Ditto.
9516 (aarch64_fmlal_lane_lowv2sf): Ditto.
9517 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9518 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9519 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9520 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9521 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9522 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9523 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9524 (aarch64_fmlal_lane_highv2sf): Ditto.
9525 (aarch64_fmlsl_lane_highv2sf): Ditto.
9526 (aarch64_fmlal_laneq_highv2sf): Ditto.
9527 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9528 (aarch64_fmlalq_lane_highv4sf): Ditto.
9529 (aarch64_fmlsl_lane_highv4sf): Ditto.
9530 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9531 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9532 * config/aarch64/aarch64-simd.md:
9533 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9534 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9535 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9536 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9537 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9538 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9539 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9540 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9541 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9542 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9543 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9544 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9545 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9546 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9547 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9548 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9549 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9550 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9551 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9552 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9553 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9554 (vfmlsl_low_u32): Ditto.
9555 (vfmlalq_low_u32): Ditto.
9556 (vfmlslq_low_u32): Ditto.
9557 (vfmlal_high_u32): Ditto.
9558 (vfmlsl_high_u32): Ditto.
9559 (vfmlalq_high_u32): Ditto.
9560 (vfmlslq_high_u32): Ditto.
9561 (vfmlal_lane_low_u32): Ditto.
9562 (vfmlsl_lane_low_u32): Ditto.
9563 (vfmlal_laneq_low_u32): Ditto.
9564 (vfmlsl_laneq_low_u32): Ditto.
9565 (vfmlalq_lane_low_u32): Ditto.
9566 (vfmlslq_lane_low_u32): Ditto.
9567 (vfmlalq_laneq_low_u32): Ditto.
9568 (vfmlslq_laneq_low_u32): Ditto.
9569 (vfmlal_lane_high_u32): Ditto.
9570 (vfmlsl_lane_high_u32): Ditto.
9571 (vfmlal_laneq_high_u32): Ditto.
9572 (vfmlsl_laneq_high_u32): Ditto.
9573 (vfmlalq_lane_high_u32): Ditto.
9574 (vfmlslq_lane_high_u32): Ditto.
9575 (vfmlalq_laneq_high_u32): Ditto.
9576 (vfmlslq_laneq_high_u32): Ditto.
9577 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9578 (AARCH64_FL_FOR_ARCH8_4): New.
9579 (AARCH64_ISA_F16FML): New ISA flag.
9580 (TARGET_F16FML): New feature flag for fp16fml.
9581 (doc/invoke.texi): Document new fp16fml option.
9583 2018-01-10 Michael Collison <michael.collison@arm.com>
9585 * config/aarch64/aarch64-builtins.c:
9586 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9587 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9588 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9589 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9590 (AARCH64_ISA_SHA3): New ISA flag.
9591 (TARGET_SHA3): New feature flag for sha3.
9592 * config/aarch64/iterators.md (sha512_op): New int attribute.
9593 (CRYPTO_SHA512): New int iterator.
9594 (UNSPEC_SHA512H): New unspec.
9595 (UNSPEC_SHA512H2): Ditto.
9596 (UNSPEC_SHA512SU0): Ditto.
9597 (UNSPEC_SHA512SU1): Ditto.
9598 * config/aarch64/aarch64-simd-builtins.def
9599 (aarch64_crypto_sha512hqv2di): New builtin.
9600 (aarch64_crypto_sha512h2qv2di): Ditto.
9601 (aarch64_crypto_sha512su0qv2di): Ditto.
9602 (aarch64_crypto_sha512su1qv2di): Ditto.
9603 (aarch64_eor3qv8hi): Ditto.
9604 (aarch64_rax1qv2di): Ditto.
9605 (aarch64_xarqv2di): Ditto.
9606 (aarch64_bcaxqv8hi): Ditto.
9607 * config/aarch64/aarch64-simd.md:
9608 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9609 (aarch64_crypto_sha512su0qv2di): Ditto.
9610 (aarch64_crypto_sha512su1qv2di): Ditto.
9611 (aarch64_eor3qv8hi): Ditto.
9612 (aarch64_rax1qv2di): Ditto.
9613 (aarch64_xarqv2di): Ditto.
9614 (aarch64_bcaxqv8hi): Ditto.
9615 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9616 (vsha512h2q_u64): Ditto.
9617 (vsha512su0q_u64): Ditto.
9618 (vsha512su1q_u64): Ditto.
9619 (veor3q_u16): Ditto.
9620 (vrax1q_u64): Ditto.
9622 (vbcaxq_u16): Ditto.
9623 * config/arm/types.md (crypto_sha512): New type attribute.
9624 (crypto_sha3): Ditto.
9625 (doc/invoke.texi): Document new sha3 option.
9627 2018-01-10 Michael Collison <michael.collison@arm.com>
9629 * config/aarch64/aarch64-builtins.c:
9630 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9631 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9632 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9633 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9634 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9635 (AARCH64_ISA_SM4): New ISA flag.
9636 (TARGET_SM4): New feature flag for sm4.
9637 * config/aarch64/aarch64-simd-builtins.def
9638 (aarch64_sm3ss1qv4si): Ditto.
9639 (aarch64_sm3tt1aq4si): Ditto.
9640 (aarch64_sm3tt1bq4si): Ditto.
9641 (aarch64_sm3tt2aq4si): Ditto.
9642 (aarch64_sm3tt2bq4si): Ditto.
9643 (aarch64_sm3partw1qv4si): Ditto.
9644 (aarch64_sm3partw2qv4si): Ditto.
9645 (aarch64_sm4eqv4si): Ditto.
9646 (aarch64_sm4ekeyqv4si): Ditto.
9647 * config/aarch64/aarch64-simd.md:
9648 (aarch64_sm3ss1qv4si): Ditto.
9649 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9650 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9651 (aarch64_sm4eqv4si): Ditto.
9652 (aarch64_sm4ekeyqv4si): Ditto.
9653 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9654 (sm3part_op): Ditto.
9655 (CRYPTO_SM3TT): Ditto.
9656 (CRYPTO_SM3PART): Ditto.
9657 (UNSPEC_SM3SS1): New unspec.
9658 (UNSPEC_SM3TT1A): Ditto.
9659 (UNSPEC_SM3TT1B): Ditto.
9660 (UNSPEC_SM3TT2A): Ditto.
9661 (UNSPEC_SM3TT2B): Ditto.
9662 (UNSPEC_SM3PARTW1): Ditto.
9663 (UNSPEC_SM3PARTW2): Ditto.
9664 (UNSPEC_SM4E): Ditto.
9665 (UNSPEC_SM4EKEY): Ditto.
9666 * config/aarch64/constraints.md (Ui2): New constraint.
9667 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9668 * config/arm/types.md (crypto_sm3): New type attribute.
9669 (crypto_sm4): Ditto.
9670 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9671 (vsm3tt1aq_u32): Ditto.
9672 (vsm3tt1bq_u32): Ditto.
9673 (vsm3tt2aq_u32): Ditto.
9674 (vsm3tt2bq_u32): Ditto.
9675 (vsm3partw1q_u32): Ditto.
9676 (vsm3partw2q_u32): Ditto.
9677 (vsm4eq_u32): Ditto.
9678 (vsm4ekeyq_u32): Ditto.
9679 (doc/invoke.texi): Document new sm4 option.
9681 2018-01-10 Michael Collison <michael.collison@arm.com>
9683 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9684 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9685 (AARCH64_FL_FOR_ARCH8_4): New.
9686 (AARCH64_FL_V8_4): New flag.
9687 (doc/invoke.texi): Document new armv8.4-a option.
9689 2018-01-10 Michael Collison <michael.collison@arm.com>
9691 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9692 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9693 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9694 * config/aarch64/aarch64-option-extension.def: Add
9695 AARCH64_OPT_EXTENSION of 'sha2'.
9696 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9697 (crypto): Disable sha2 and aes if crypto disabled.
9698 (crypto): Enable aes and sha2 if enabled.
9699 (simd): Disable sha2 and aes if simd disabled.
9700 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9702 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9703 (TARGET_SHA2): New feature flag for sha2.
9704 (TARGET_AES): New feature flag for aes.
9705 * config/aarch64/aarch64-simd.md:
9706 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9707 conditional on TARGET_AES.
9708 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9709 (aarch64_crypto_sha1hsi): Make pattern conditional
9711 (aarch64_crypto_sha1hv4si): Ditto.
9712 (aarch64_be_crypto_sha1hv4si): Ditto.
9713 (aarch64_crypto_sha1su1v4si): Ditto.
9714 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9715 (aarch64_crypto_sha1su0v4si): Ditto.
9716 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9717 (aarch64_crypto_sha256su0v4si): Ditto.
9718 (aarch64_crypto_sha256su1v4si): Ditto.
9719 (doc/invoke.texi): Document new aes and sha2 options.
9721 2018-01-10 Martin Sebor <msebor@redhat.com>
9723 PR tree-optimization/83781
9724 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9727 2018-01-11 Martin Sebor <msebor@gmail.com>
9728 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9730 PR tree-optimization/83501
9731 PR tree-optimization/81703
9733 * tree-ssa-strlen.c (get_string_cst): Rename...
9734 (get_string_len): ...to this. Handle global constants.
9735 (handle_char_store): Adjust.
9737 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9738 Jim Wilson <jimw@sifive.com>
9740 * config/riscv/riscv-protos.h (riscv_output_return): New.
9741 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9742 (riscv_attribute_table, riscv_output_return),
9743 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9744 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9745 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9747 (riscv_expand_prologue): Add early return for naked function.
9748 (riscv_expand_epilogue): Likewise.
9749 (riscv_function_ok_for_sibcall): Return false for naked function.
9750 (riscv_set_current_function): New.
9751 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9752 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9753 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9754 * doc/extend.texi (RISC-V Function Attributes): New.
9756 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9758 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9759 check for 128-bit long double before checking TCmode.
9760 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9761 128-bit long doubles before checking TFmode or TCmode.
9762 (FLOAT128_IBM_P): Likewise.
9764 2018-01-10 Martin Sebor <msebor@redhat.com>
9766 PR tree-optimization/83671
9767 * builtins.c (c_strlen): Unconditionally return zero for the empty
9769 Use -Warray-bounds for warnings.
9770 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9771 for non-constant array indices with COMPONENT_REF, arrays of
9772 arrays, and pointers to arrays.
9773 (gimple_fold_builtin_strlen): Determine and set length range for
9774 non-constant character arrays.
9776 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9779 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9782 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9784 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9786 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9789 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9790 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9791 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9792 indexed_or_indirect_operand predicate.
9793 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9794 (*vsx_le_perm_load_v8hi): Likewise.
9795 (*vsx_le_perm_load_v16qi): Likewise.
9796 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9797 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9798 (*vsx_le_perm_store_v8hi): Likewise.
9799 (*vsx_le_perm_store_v16qi): Likewise.
9800 (eight unnamed splitters): Likewise.
9802 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9804 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9805 * config/rs6000/emmintrin.h: Likewise.
9806 * config/rs6000/mmintrin.h: Likewise.
9807 * config/rs6000/xmmintrin.h: Likewise.
9809 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9812 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9814 * tree.c (tree_nop_conversion): Return true for location wrapper
9816 (maybe_wrap_with_location): New function.
9817 (selftest::check_strip_nops): New function.
9818 (selftest::test_location_wrappers): New function.
9819 (selftest::tree_c_tests): Call it.
9820 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9821 (maybe_wrap_with_location): New decl.
9822 (EXPR_LOCATION_WRAPPER_P): New macro.
9823 (location_wrapper_p): New inline function.
9824 (tree_strip_any_location_wrapper): New inline function.
9826 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9829 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9830 stack_realign_offset for the largest alignment of stack slot
9832 (ix86_find_max_used_stack_alignment): New function.
9833 (ix86_finalize_stack_frame_flags): Use it. Set
9834 max_used_stack_alignment if we don't realign stack.
9835 * config/i386/i386.h (machine_function): Add
9836 max_used_stack_alignment.
9838 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9840 * config/arm/arm.opt (-mbranch-cost): New option.
9841 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9844 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9847 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9848 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9850 2018-01-10 Richard Biener <rguenther@suse.de>
9853 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9854 early out so it also covers the case where we have a non-NULL
9857 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9859 PR tree-optimization/83753
9860 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9861 for non-strided grouped accesses if the number of elements is 1.
9863 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9866 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9867 * i386.h (TARGET_USE_GATHER): Define.
9868 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9870 2018-01-10 Martin Liska <mliska@suse.cz>
9873 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9874 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9876 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9877 CLEANUP_NO_PARTITIONING is not set.
9879 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9881 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9882 for vectors, as a partial revert of r254296.
9883 * rtl.h (const_vec_p): Delete.
9884 (const_vec_duplicate_p): Don't test for vector CONSTs.
9885 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9886 * expmed.c (make_tree): Likewise.
9889 * common.md (E, F): Use CONSTANT_P instead of checking for
9891 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9892 checking for CONST_VECTOR.
9894 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9897 * predict.c (force_edge_cold): Handle in more sane way edges
9900 2018-01-09 Carl Love <cel@us.ibm.com>
9902 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9904 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9905 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9906 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9907 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9908 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9909 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9910 * config/rs6000/rs6000-protos.h: Add extern defition for
9911 rs6000_generate_float2_double_code.
9912 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9914 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9915 (float2_v2df): Add define_expand.
9917 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9920 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9921 op_mode in the force_to_mode call.
9923 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9925 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9926 instead of checking each element individually.
9927 (aarch64_evpc_uzp): Likewise.
9928 (aarch64_evpc_zip): Likewise.
9929 (aarch64_evpc_ext): Likewise.
9930 (aarch64_evpc_rev): Likewise.
9931 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9932 instead of checking each element individually. Return true without
9934 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9935 whether all selected elements come from the same input, instead of
9936 checking each element individually. Remove calls to gen_rtx_REG,
9937 start_sequence and end_sequence and instead assert that no rtl is
9940 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9942 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9943 order of HIGH and CONST checks.
9945 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9947 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9948 if the destination isn't an SSA_NAME.
9950 2018-01-09 Richard Biener <rguenther@suse.de>
9952 PR tree-optimization/83668
9953 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9955 (canonicalize_loop_form): ... here, renamed from ...
9956 (canonicalize_loop_closed_ssa_form): ... this and amended to
9957 swap successor edges for loop exit blocks to make us use
9958 the RPO order we need for initial schedule generation.
9960 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9962 PR tree-optimization/64811
9963 * match.pd: When optimizing comparisons with Inf, avoid
9964 introducing or losing exceptions from comparisons with NaN.
9966 2018-01-09 Martin Liska <mliska@suse.cz>
9969 * asan.c (shadow_mem_size): Add gcc_assert.
9971 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9973 Don't save registers in main().
9976 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9977 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9978 * config/avr/avr.c (avr_set_current_function): Don't error if
9979 naked, OS_task or OS_main are specified at the same time.
9980 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9982 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9984 * common/config/avr/avr-common.c (avr_option_optimization_table):
9985 Switch on -mmain-is-OS_task for optimizing compilations.
9987 2018-01-09 Richard Biener <rguenther@suse.de>
9989 PR tree-optimization/83572
9990 * graphite.c: Include cfganal.h.
9991 (graphite_transform_loops): Connect infinite loops to exit
9992 and remove fake edges at the end.
9994 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9996 * ipa-inline.c (edge_badness): Revert accidental checkin.
9998 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10001 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10002 symbols; not inline clones.
10004 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10007 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10008 hard registers. Formatting fixes.
10010 PR preprocessor/83722
10011 * gcc.c (try_generate_repro): Pass
10012 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10013 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10016 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10017 Kito Cheng <kito.cheng@gmail.com>
10019 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10020 (riscv_leaf_function_p): Delete.
10021 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10023 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10025 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10027 (do_ifelse): New function.
10028 (do_isel): New function.
10029 (do_sub3): New function.
10030 (do_add3): New function.
10031 (do_load_mask_compare): New function.
10032 (do_overlap_load_compare): New function.
10033 (expand_compare_loop): New function.
10034 (expand_block_compare): Call expand_compare_loop() when appropriate.
10035 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10036 option description.
10037 (-mblock-compare-inline-loop-limit): New option.
10039 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10042 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10043 Reverse order of second and third operands in first alternative.
10044 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10045 of first and second elements in UNSPEC_VPERMR vector.
10046 (altivec_expand_vec_perm_le): Likewise.
10048 2018-01-08 Jeff Law <law@redhat.com>
10050 PR rtl-optimizatin/81308
10051 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10052 (process_switch): If group_case_labels makes a change, then set
10054 (pass_convert_switch::execute): If a switch is converted, then
10055 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10057 PR rtl-optimization/81308
10058 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10061 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10063 PR target/83663 - Revert r255946
10064 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10065 generation for cases where splatting a value is not useful.
10066 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10067 across a vec_duplicate and a paradoxical subreg forming a vector
10068 mode to a vec_concat.
10070 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10072 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10073 -march=armv8.3-a variants.
10074 * config/arm/t-multilib: Likewise.
10075 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10077 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10079 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10081 (cceq_ior_compare_complement): Give it a name so I can use it, and
10082 change boolean_or_operator predicate to boolean_operator so it can
10083 be used to generate a crand.
10084 (eqne): New code iterator.
10085 (bd/bd_neg): New code_attrs.
10086 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10087 a single define_insn.
10088 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10089 decrement (bdnzt/bdnzf/bdzt/bdzf).
10090 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10091 with the new names of the branch decrement patterns, and added the
10092 names of the branch decrement conditional patterns.
10094 2018-01-08 Richard Biener <rguenther@suse.de>
10096 PR tree-optimization/83563
10097 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10100 2018-01-08 Richard Biener <rguenther@suse.de>
10102 PR middle-end/83713
10103 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10105 2018-01-08 Richard Biener <rguenther@suse.de>
10107 PR tree-optimization/83685
10108 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10109 references to abnormals.
10111 2018-01-08 Richard Biener <rguenther@suse.de>
10114 * dwarf2out.c (output_indirect_strings): Handle empty
10115 skeleton_debug_str_hash.
10116 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10118 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10120 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10121 (emit_store_direct): Likewise.
10122 (arc_trampoline_adjust_address): Likewise.
10123 (arc_asm_trampoline_template): New function.
10124 (arc_initialize_trampoline): Use asm_trampoline_template.
10125 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10126 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10127 * config/arc/arc.md (flush_icache): Delete pattern.
10129 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10131 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10132 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10135 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10138 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10139 by not USED_FOR_TARGET.
10140 (make_pass_resolve_sw_modes): Likewise.
10142 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10144 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10147 2018-01-08 Richard Biener <rguenther@suse.de>
10149 PR middle-end/83580
10150 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10152 2018-01-08 Richard Biener <rguenther@suse.de>
10154 PR middle-end/83517
10155 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10157 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10159 PR middle-end/81897
10160 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10161 basic blocks with a small number of successors.
10162 (convert_control_dep_chain_into_preds): Improve handling of
10164 (dump_predicates): Split apart into...
10165 (dump_pred_chain): ...here...
10166 (dump_pred_info): ...and here.
10167 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10168 (can_chain_union_be_invalidated_p): Improve check for invalidation
10170 (uninit_uses_cannot_happen): Avoid unnecessary if
10171 convert_control_dep_chain_into_preds yielded nothing.
10173 2018-01-06 Martin Sebor <msebor@redhat.com>
10175 PR tree-optimization/83640
10176 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10177 subtracting negative offset from size.
10178 (builtin_access::overlap): Adjust offset bounds of the access to fall
10179 within the size of the object if possible.
10181 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10183 PR rtl-optimization/83699
10184 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10185 extract_bit_field_as_subreg to cases in which the extracted
10186 value is also a vector.
10188 * lra-constraints.c (process_alt_operands): Test for the equivalence
10189 substitutions when detecting a possible reload cycle.
10191 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10194 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10195 by default if flag_selective_schedling{,2}. Formatting fixes.
10197 PR rtl-optimization/83682
10198 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10199 if it has non-VECTOR_MODE element mode.
10200 (vec_duplicate_p): Likewise.
10202 PR middle-end/83694
10203 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10204 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10206 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10209 * config/i386/i386-builtin.def
10210 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10211 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10212 Require also OPTION_MASK_ISA_AVX512F in addition to
10213 OPTION_MASK_ISA_GFNI.
10214 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10215 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10216 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10217 to OPTION_MASK_ISA_GFNI.
10218 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10219 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10220 OPTION_MASK_ISA_AVX512BW.
10221 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10222 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10223 addition to OPTION_MASK_ISA_GFNI.
10224 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10225 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10226 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10227 to OPTION_MASK_ISA_GFNI.
10228 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10229 a requirement for all ISAs rather than any of them with a few
10231 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10233 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10234 bitmasks to be enabled with 3 exceptions, instead of requiring any
10235 enabled ISA with lots of exceptions.
10236 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10237 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10238 Change avx512bw in isa attribute to avx512f.
10239 * config/i386/sgxintrin.h: Add license boilerplate.
10240 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10241 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10242 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10243 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10245 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10246 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10247 temporarily sse2 rather than sse if not enabled already.
10250 * config/i386/sse.md (VI248_VLBW): Rename to ...
10251 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10252 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10253 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10254 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10255 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10256 mode iterator instead of VI248_VLBW.
10258 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10260 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10261 (record_modified): Skip clobbers; add debug output.
10262 (param_change_prob): Use sreal frequencies.
10264 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10266 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10267 punt for user-aligned variables.
10269 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10271 * tree-chrec.c (chrec_contains_symbols): Return true for
10274 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10277 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10278 of (x|y) == x for BICS pattern.
10280 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10282 PR tree-optimization/83605
10283 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10284 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10287 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10289 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10290 * config/epiphany/rtems.h: New file.
10292 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10293 Uros Bizjak <ubizjak@gmail.com>
10296 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10297 QIreg_operand instead of register_operand predicate.
10298 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10299 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10300 comments instead of -fmitigate[-_]rop.
10302 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10305 * cgraphunit.c (symbol_table::compile): Switch to text_section
10306 before calling assembly_start debug hook.
10307 * run-rtl-passes.c (run_rtl_passes): Likewise.
10310 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10312 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10313 range_int_cst_p rather than !symbolic_range_p before calling
10314 extract_range_from_multiplicative_op_1.
10316 2018-01-04 Jeff Law <law@redhat.com>
10318 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10319 redundant test in assertion.
10321 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10323 * doc/rtl.texi: Document machine_mode wrapper classes.
10325 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10327 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10328 using tree_to_uhwi.
10330 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10332 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10333 the VEC_PERM_EXPR fold to fail.
10335 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10338 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10339 to switched_sections.
10341 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10344 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10345 test for d.testing.
10347 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10350 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10351 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10353 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10356 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10357 is BLKmode and bitpos not zero or mode change is needed.
10359 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10362 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10365 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10368 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10369 instead of MULT rtx. Update all corresponding splitters.
10370 (*saddl_se): Ditto.
10371 (*ssub<modesuffix>): Ditto.
10372 (*ssubl_se): Ditto.
10373 (*cmp_sadd_di): Update split patterns.
10374 (*cmp_sadd_si): Ditto.
10375 (*cmp_sadd_sidi): Ditto.
10376 (*cmp_ssub_di): Ditto.
10377 (*cmp_ssub_si): Ditto.
10378 (*cmp_ssub_sidi): Ditto.
10379 * config/alpha/predicates.md (const23_operand): New predicate.
10380 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10381 Look for ASHIFT, not MULT inner operand.
10382 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10384 2018-01-04 Martin Liska <mliska@suse.cz>
10386 PR gcov-profile/83669
10387 * gcov.c (output_intermediate_file): Add version to intermediate
10389 * doc/gcov.texi: Document new field 'version' in intermediate
10390 file format. Fix location of '-k' option of gcov command.
10392 2018-01-04 Martin Liska <mliska@suse.cz>
10395 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10397 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10399 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10401 2018-01-03 Martin Sebor <msebor@redhat.com>
10403 PR tree-optimization/83655
10404 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10405 checking calls with invalid arguments.
10407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10409 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10410 (vectorizable_mask_load_store): Delete.
10411 (vectorizable_call): Return false for masked loads and stores.
10412 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10413 instead of gimple_assign_rhs1.
10414 (vectorizable_load): Handle IFN_MASK_LOAD.
10415 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10419 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10421 (vectorizable_mask_load_store): ...here.
10422 (vectorizable_load): ...and here.
10424 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10426 * tree-vect-stmts.c (vect_build_all_ones_mask)
10427 (vect_build_zero_merge_argument): New functions, split out from...
10428 (vectorizable_load): ...here.
10430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10432 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10434 (vectorizable_mask_load_store): ...here.
10435 (vectorizable_store): ...and here.
10437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10439 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10441 (vectorizable_mask_load_store): ...here.
10443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10445 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10446 (vect_model_store_cost): Take a vec_load_store_type instead of a
10448 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10449 (vect_model_store_cost): Take a vec_load_store_type instead of a
10451 (vectorizable_mask_load_store): Update accordingly.
10452 (vectorizable_store): Likewise.
10453 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10457 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10458 IFN_MASK_LOAD calls here rather than...
10459 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10462 Alan Hayward <alan.hayward@arm.com>
10463 David Sherwood <david.sherwood@arm.com>
10465 * expmed.c (extract_bit_field_1): For vector extracts,
10466 fall back to extract_bit_field_as_subreg if vec_extract
10469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10470 Alan Hayward <alan.hayward@arm.com>
10471 David Sherwood <david.sherwood@arm.com>
10473 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10474 they are variable or constant sized.
10475 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10476 slots for constant-sized data.
10478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10479 Alan Hayward <alan.hayward@arm.com>
10480 David Sherwood <david.sherwood@arm.com>
10482 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10483 handling COND_EXPRs with boolean comparisons, try to find a better
10484 basis for the mask type than the boolean itself.
10486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10488 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10489 is calculated and how it can be overridden.
10490 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10491 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10493 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10496 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10497 Alan Hayward <alan.hayward@arm.com>
10498 David Sherwood <david.sherwood@arm.com>
10500 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10501 Remove the mode argument.
10502 (aarch64_simd_valid_immediate): Remove the mode and inverse
10504 * config/aarch64/iterators.md (bitsize): New iterator.
10505 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10506 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10507 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10508 aarch64_simd_valid_immediate.
10509 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10510 (aarch64_reg_or_bic_imm): Likewise.
10511 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10512 with an insn_type enum and msl with a modifier_type enum.
10513 Replace element_width with a scalar_mode. Change the shift
10514 to unsigned int. Add constructors for scalar_float_mode and
10515 scalar_int_mode elements.
10516 (aarch64_vect_float_const_representable_p): Delete.
10517 (aarch64_can_const_movi_rtx_p)
10518 (aarch64_simd_scalar_immediate_valid_for_move)
10519 (aarch64_simd_make_constant): Update call to
10520 aarch64_simd_valid_immediate.
10521 (aarch64_advsimd_valid_immediate_hs): New function.
10522 (aarch64_advsimd_valid_immediate): Likewise.
10523 (aarch64_simd_valid_immediate): Remove mode and inverse
10524 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10525 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10526 and aarch64_float_const_representable_p on the result.
10527 (aarch64_output_simd_mov_immediate): Remove mode argument.
10528 Update call to aarch64_simd_valid_immediate and use of
10529 simd_immediate_info.
10530 (aarch64_output_scalar_simd_mov_immediate): Update call
10533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10534 Alan Hayward <alan.hayward@arm.com>
10535 David Sherwood <david.sherwood@arm.com>
10537 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10538 (mode_nunits): Likewise CONST_MODE_NUNITS.
10539 * machmode.def (ADJUST_NUNITS): Document.
10540 * genmodes.c (mode_data::need_nunits_adj): New field.
10541 (blank_mode): Update accordingly.
10542 (adj_nunits): New variable.
10543 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10545 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10546 listed in adj_nunits.
10547 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10548 listed in adj_nunits. Don't emit case statements for such modes.
10549 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10550 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10551 nothing if adj_nunits is nonnull.
10552 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10553 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10554 (emit_mode_fbit): Update use of print_maybe_const_decl.
10555 (emit_move_size): Likewise. Treat the array as non-const
10557 (emit_mode_adjustments): Handle adj_nunits.
10559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10561 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10562 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10563 (VECTOR_MODES): Use it.
10564 (make_vector_modes): Take the prefix as an argument.
10566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10567 Alan Hayward <alan.hayward@arm.com>
10568 David Sherwood <david.sherwood@arm.com>
10570 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10571 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10572 for MODE_VECTOR_BOOL.
10573 * machmode.def (VECTOR_BOOL_MODE): Document.
10574 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10575 (make_vector_bool_mode): New function.
10576 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10578 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10579 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10581 * stor-layout.c (int_mode_for_mode): Likewise.
10582 * tree.c (build_vector_type_for_mode): Likewise.
10583 * varasm.c (output_constant_pool_2): Likewise.
10584 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10585 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10586 for MODE_VECTOR_BOOL.
10587 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10588 of mode class checks.
10589 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10590 instead of a list of mode class checks.
10591 (expand_vector_scalar_condition): Likewise.
10592 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10595 Alan Hayward <alan.hayward@arm.com>
10596 David Sherwood <david.sherwood@arm.com>
10598 * machmode.h (mode_size): Change from unsigned short to
10600 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10601 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10602 or if measurement_type is not polynomial.
10603 (fixed_size_mode::includes_p): Check for constant-sized modes.
10604 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10605 return a poly_uint16 rather than an unsigned short.
10606 (emit_mode_size): Change the type of mode_size from unsigned short
10607 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10608 (emit_mode_adjustments): Cope with polynomial vector sizes.
10609 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10611 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10613 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10614 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10615 * caller-save.c (setup_save_areas): Likewise.
10616 (replace_reg_with_saved_mem): Likewise.
10617 * calls.c (emit_library_call_value_1): Likewise.
10618 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10619 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10620 (gen_lowpart_for_combine): Likewise.
10621 * convert.c (convert_to_integer_1): Likewise.
10622 * cse.c (equiv_constant, cse_insn): Likewise.
10623 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10624 (cselib_subst_to_values): Likewise.
10625 * dce.c (word_dce_process_block): Likewise.
10626 * df-problems.c (df_word_lr_mark_ref): Likewise.
10627 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10628 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10629 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10630 (rtl_for_decl_location): Likewise.
10631 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10632 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10633 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10634 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10635 (expand_expr_real_1): Likewise.
10636 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10637 (pad_below): Likewise.
10638 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10639 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10640 * ira.c (get_subreg_tracking_sizes): Likewise.
10641 * ira-build.c (ira_create_allocno_objects): Likewise.
10642 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10643 (ira_sort_regnos_for_alter_reg): Likewise.
10644 * ira-costs.c (record_operand_costs): Likewise.
10645 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10646 (resolve_simple_move): Likewise.
10647 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10648 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10649 (lra_constraints): Likewise.
10650 (CONST_POOL_OK_P): Reject variable-sized modes.
10651 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10652 (add_pseudo_to_slot, lra_spill): Likewise.
10653 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10654 * optabs-query.c (get_best_extraction_insn): Likewise.
10655 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10656 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10657 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10658 * recog.c (offsettable_address_addr_space_p): Likewise.
10659 * regcprop.c (maybe_mode_change): Likewise.
10660 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10661 * regrename.c (build_def_use): Likewise.
10662 * regstat.c (dump_reg_info): Likewise.
10663 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10664 (find_reloads, find_reloads_subreg_address): Likewise.
10665 * reload1.c (eliminate_regs_1): Likewise.
10666 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10667 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10668 (simplify_binary_operation_1, simplify_subreg): Likewise.
10669 * targhooks.c (default_function_arg_padding): Likewise.
10670 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10671 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10672 (verify_gimple_assign_ternary): Likewise.
10673 * tree-inline.c (estimate_move_cost): Likewise.
10674 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10675 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10676 (get_address_cost_ainc): Likewise.
10677 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10678 (vect_supportable_dr_alignment): Likewise.
10679 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10680 (vectorizable_reduction): Likewise.
10681 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10682 (vectorizable_operation, vectorizable_load): Likewise.
10683 * tree.c (build_same_sized_truth_vector_type): Likewise.
10684 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10685 * var-tracking.c (emit_note_insn_var_location): Likewise.
10686 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10687 (ADDR_VEC_ALIGN): Likewise.
10689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10690 Alan Hayward <alan.hayward@arm.com>
10691 David Sherwood <david.sherwood@arm.com>
10693 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10695 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10696 or if measurement_type is polynomial.
10697 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10698 * combine.c (make_extraction): Likewise.
10699 * dse.c (find_shift_sequence): Likewise.
10700 * dwarf2out.c (mem_loc_descriptor): Likewise.
10701 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10702 (extract_bit_field, extract_low_bits): Likewise.
10703 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10704 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10705 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10706 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10707 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10708 * reload.c (find_reloads): Likewise.
10709 * reload1.c (alter_reg): Likewise.
10710 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10711 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10712 * tree-if-conv.c (predicate_mem_writes): Likewise.
10713 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10714 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10715 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10716 * valtrack.c (dead_debug_insert_temp): Likewise.
10717 * varasm.c (mergeable_constant_section): Likewise.
10718 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10721 Alan Hayward <alan.hayward@arm.com>
10722 David Sherwood <david.sherwood@arm.com>
10724 * expr.c (expand_assignment): Cope with polynomial mode sizes
10725 when assigning to a CONCAT.
10727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10728 Alan Hayward <alan.hayward@arm.com>
10729 David Sherwood <david.sherwood@arm.com>
10731 * machmode.h (mode_precision): Change from unsigned short to
10733 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10735 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10736 or if measurement_type is not polynomial.
10737 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10738 in which the mode is already known to be a scalar_int_mode.
10739 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10740 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10742 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10743 for GET_MODE_PRECISION.
10744 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10745 for GET_MODE_PRECISION.
10746 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10748 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10749 (expand_field_assignment, make_extraction): Likewise.
10750 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10751 (get_last_value): Likewise.
10752 * convert.c (convert_to_integer_1): Likewise.
10753 * cse.c (cse_insn): Likewise.
10754 * expr.c (expand_expr_real_1): Likewise.
10755 * lra-constraints.c (simplify_operand_subreg): Likewise.
10756 * optabs-query.c (can_atomic_load_p): Likewise.
10757 * optabs.c (expand_atomic_load): Likewise.
10758 (expand_atomic_store): Likewise.
10759 * ree.c (combine_reaching_defs): Likewise.
10760 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10761 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10762 * tree.h (type_has_mode_precision_p): Likewise.
10763 * ubsan.c (instrument_si_overflow): Likewise.
10765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10766 Alan Hayward <alan.hayward@arm.com>
10767 David Sherwood <david.sherwood@arm.com>
10769 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10770 polynomial numbers of units.
10771 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10772 (valid_vector_subparts_p): New function.
10773 (build_vector_type): Remove temporary shim and take the number
10774 of units as a poly_uint64 rather than an int.
10775 (build_opaque_vector_type): Take the number of units as a
10776 poly_uint64 rather than an int.
10777 * tree.c (build_vector_from_ctor): Handle polynomial
10778 TYPE_VECTOR_SUBPARTS.
10779 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10780 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10781 (build_vector_from_val): If the number of units is variable,
10782 use build_vec_duplicate_cst for constant operands and
10783 VEC_DUPLICATE_EXPR otherwise.
10784 (make_vector_type): Remove temporary is_constant ().
10785 (build_vector_type, build_opaque_vector_type): Take the number of
10786 units as a poly_uint64 rather than an int.
10787 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10789 * cfgexpand.c (expand_debug_expr): Likewise.
10790 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10791 (store_constructor, expand_expr_real_1): Likewise.
10792 (const_scalar_mask_from_tree): Likewise.
10793 * fold-const-call.c (fold_const_reduction): Likewise.
10794 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10795 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10796 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10797 (fold_relational_const): Likewise.
10798 (native_interpret_vector): Likewise. Change the size from an
10799 int to an unsigned int.
10800 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10801 TYPE_VECTOR_SUBPARTS.
10802 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10803 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10804 duplicating a non-constant operand into a variable-length vector.
10805 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10806 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10807 * ipa-icf.c (sem_variable::equals): Likewise.
10808 * match.pd: Likewise.
10809 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10810 * print-tree.c (print_node): Likewise.
10811 * stor-layout.c (layout_type): Likewise.
10812 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10813 * tree-cfg.c (verify_gimple_comparison): Likewise.
10814 (verify_gimple_assign_binary): Likewise.
10815 (verify_gimple_assign_ternary): Likewise.
10816 (verify_gimple_assign_single): Likewise.
10817 * tree-pretty-print.c (dump_generic_node): Likewise.
10818 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10819 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10820 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10821 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10822 (vect_shift_permute_load_chain): Likewise.
10823 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10824 (expand_vector_condition, optimize_vector_constructor): Likewise.
10825 (lower_vec_perm, get_compute_type): Likewise.
10826 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10827 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10828 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10829 (vect_recog_mask_conversion_pattern): Likewise.
10830 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10831 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10832 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10833 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10834 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10835 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10836 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10837 (supportable_widening_operation): Likewise.
10838 (supportable_narrowing_operation): Likewise.
10839 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10841 * varasm.c (output_constant): Likewise.
10843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10844 Alan Hayward <alan.hayward@arm.com>
10845 David Sherwood <david.sherwood@arm.com>
10847 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10848 so that both the length == 3 and length != 3 cases set up their
10849 own permute vectors. Add comments explaining why we know the
10850 number of elements is constant.
10851 (vect_permute_load_chain): Likewise.
10853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10854 Alan Hayward <alan.hayward@arm.com>
10855 David Sherwood <david.sherwood@arm.com>
10857 * machmode.h (mode_nunits): Change from unsigned char to
10859 (ONLY_FIXED_SIZE_MODES): New macro.
10860 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10861 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10862 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10864 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10865 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10866 or if measurement_type is not polynomial.
10867 * genmodes.c (ZERO_COEFFS): New macro.
10868 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10870 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10871 Use ZERO_COEFFS when emitting initializers.
10872 * data-streamer.h (bp_pack_poly_value): New function.
10873 (bp_unpack_poly_value): Likewise.
10874 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10875 for GET_MODE_NUNITS.
10876 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10877 for GET_MODE_NUNITS.
10878 * tree.c (make_vector_type): Remove temporary shim and make
10879 the real function take the number of units as a poly_uint64
10880 rather than an int.
10881 (build_vector_type_for_mode): Handle polynomial nunits.
10882 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10883 * emit-rtl.c (const_vec_series_p_1): Likewise.
10884 (gen_rtx_CONST_VECTOR): Likewise.
10885 * fold-const.c (test_vec_duplicate_folding): Likewise.
10886 * genrecog.c (validate_pattern): Likewise.
10887 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10888 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10889 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10890 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10891 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10892 * rtlanal.c (subreg_get_info): Likewise.
10893 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10894 (vect_grouped_load_supported): Likewise.
10895 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10896 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10897 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10898 (simplify_const_unary_operation, simplify_binary_operation_1)
10899 (simplify_const_binary_operation, simplify_ternary_operation)
10900 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10901 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10902 instead of CONST_VECTOR_NUNITS.
10903 * varasm.c (output_constant_pool_2): Likewise.
10904 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10905 explicit-encoded elements in the XVEC for variable-length vectors.
10907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10909 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10912 Alan Hayward <alan.hayward@arm.com>
10913 David Sherwood <david.sherwood@arm.com>
10915 * coretypes.h (fixed_size_mode): Declare.
10916 (fixed_size_mode_pod): New typedef.
10917 * builtins.h (target_builtins::x_apply_args_mode)
10918 (target_builtins::x_apply_result_mode): Change type to
10919 fixed_size_mode_pod.
10920 * builtins.c (apply_args_size, apply_result_size, result_vector)
10921 (expand_builtin_apply_args_1, expand_builtin_apply)
10922 (expand_builtin_return): Update accordingly.
10924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10926 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10927 * cselib.c (cselib_hash_rtx): Likewise.
10928 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10929 CONST_VECTOR encoding.
10931 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10932 Jeff Law <law@redhat.com>
10935 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10936 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10937 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10938 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10941 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10942 explicitly probe *sp in a noreturn function if there were any callee
10943 register saves or frame pointer is needed.
10945 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10948 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10949 BLKmode for ternary, binary or unary expressions.
10952 * var-tracking.c (delete_vta_debug_insn): New inline function.
10953 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10954 insns from get_insns () to NULL instead of each bb separately.
10955 Use delete_vta_debug_insn. No longer static.
10956 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10957 delete_vta_debug_insns callers.
10958 * rtl.h (delete_vta_debug_insns): Declare.
10959 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10960 instead of variable_tracking_main.
10962 2018-01-03 Martin Sebor <msebor@redhat.com>
10964 PR tree-optimization/83603
10965 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10966 arguments past the endof the argument list in functions declared
10967 without a prototype.
10968 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10969 Avoid checking when arguments are null.
10971 2018-01-03 Martin Sebor <msebor@redhat.com>
10974 * doc/extend.texi (attribute const): Fix a typo.
10975 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10976 issuing -Wsuggest-attribute for void functions.
10978 2018-01-03 Martin Sebor <msebor@redhat.com>
10980 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10981 offset_int::from instead of wide_int::to_shwi.
10982 (maybe_diag_overlap): Remove assertion.
10983 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10984 * gimple-ssa-sprintf.c (format_directive): Same.
10985 (parse_directive): Same.
10986 (sprintf_dom_walker::compute_format_length): Same.
10987 (try_substitute_return_value): Same.
10989 2018-01-03 Jeff Law <law@redhat.com>
10991 PR middle-end/83654
10992 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10993 non-constant residual for zero at runtime and avoid probing in
10994 that case. Reorganize code for trailing problem to mirror handling
10997 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10999 PR tree-optimization/83501
11000 * tree-ssa-strlen.c (get_string_cst): New.
11001 (handle_char_store): Call get_string_cst.
11003 2018-01-03 Martin Liska <mliska@suse.cz>
11005 PR tree-optimization/83593
11006 * tree-ssa-strlen.c: Include tree-cfg.h.
11007 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11008 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11009 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11011 (strlen_dom_walker::before_dom_children): Call
11012 gimple_purge_dead_eh_edges. Dump tranformation with details
11014 (strlen_dom_walker::before_dom_children): Update call by adding
11015 new argument cleanup_eh.
11016 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11018 2018-01-03 Martin Liska <mliska@suse.cz>
11021 * cif-code.def (VARIADIC_THUNK): New enum value.
11022 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11025 2018-01-03 Jan Beulich <jbeulich@suse.com>
11027 * sse.md (mov<mode>_internal): Tighten condition for when to use
11028 vmovdqu<ssescalarsize> for TI and OI modes.
11030 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11032 Update copyright years.
11034 2018-01-03 Martin Liska <mliska@suse.cz>
11037 * ipa-visibility.c (function_and_variable_visibility): Skip
11038 functions with noipa attribure.
11040 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11042 * gcc.c (process_command): Update copyright notice dates.
11043 * gcov-dump.c (print_version): Ditto.
11044 * gcov.c (print_version): Ditto.
11045 * gcov-tool.c (print_version): Ditto.
11046 * gengtype.c (create_file): Ditto.
11047 * doc/cpp.texi: Bump @copying's copyright year.
11048 * doc/cppinternals.texi: Ditto.
11049 * doc/gcc.texi: Ditto.
11050 * doc/gccint.texi: Ditto.
11051 * doc/gcov.texi: Ditto.
11052 * doc/install.texi: Ditto.
11053 * doc/invoke.texi: Ditto.
11055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11057 * vector-builder.h (vector_builder::m_full_nelts): Change from
11058 unsigned int to poly_uint64.
11059 (vector_builder::full_nelts): Update prototype accordingly.
11060 (vector_builder::new_vector): Likewise.
11061 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11062 (vector_builder::operator ==): Likewise.
11063 (vector_builder::finalize): Likewise.
11064 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11065 Take the number of elements as a poly_uint64 rather than an
11067 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11068 from unsigned int to poly_uint64.
11069 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11070 (vec_perm_indices::new_vector): Likewise.
11071 (vec_perm_indices::length): Likewise.
11072 (vec_perm_indices::nelts_per_input): Likewise.
11073 (vec_perm_indices::input_nelts): Likewise.
11074 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11075 number of elements per input as a poly_uint64 rather than an
11076 unsigned int. Use the original encoding for variable-length
11077 vectors, rather than clamping each individual element.
11078 For the second and subsequent elements in each pattern,
11079 clamp the step and base before clamping their sum.
11080 (vec_perm_indices::series_p): Handle polynomial element counts.
11081 (vec_perm_indices::all_in_range_p): Likewise.
11082 (vec_perm_indices_to_tree): Likewise.
11083 (vec_perm_indices_to_rtx): Likewise.
11084 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11085 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11086 (tree_vector_builder::new_binary_operation): Handle polynomial
11087 element counts. Return false if we need to know the number
11088 of elements at compile time.
11089 * fold-const.c (fold_vec_perm): Punt if the number of elements
11090 isn't known at compile time.
11092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11094 * vec-perm-indices.h (vec_perm_builder): Change element type
11095 from HOST_WIDE_INT to poly_int64.
11096 (vec_perm_indices::element_type): Update accordingly.
11097 (vec_perm_indices::clamp): Handle polynomial element_types.
11098 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11099 (vec_perm_indices::all_in_range_p): Likewise.
11100 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11102 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11103 polynomial vec_perm_indices element types.
11104 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11105 * fold-const.c (fold_vec_perm): Likewise.
11106 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11107 * tree-vect-generic.c (lower_vec_perm): Likewise.
11108 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11109 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11110 element type to HOST_WIDE_INT.
11112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11113 Alan Hayward <alan.hayward@arm.com>
11114 David Sherwood <david.sherwood@arm.com>
11116 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11117 rather than an int. Use plus_constant.
11118 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11119 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11122 Alan Hayward <alan.hayward@arm.com>
11123 David Sherwood <david.sherwood@arm.com>
11125 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11126 a HOST_WIDE_INT to a poly_int64.
11128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11129 Alan Hayward <alan.hayward@arm.com>
11130 David Sherwood <david.sherwood@arm.com>
11132 * calls.c (load_register_parameters): Cope with polynomial
11133 mode sizes. Require a constant size for BLKmode parameters
11134 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11135 forces a parameter to be padded at the lsb end in order to
11136 fill a complete number of words, require the parameter size
11137 to be ordered wrt UNITS_PER_WORD.
11139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11140 Alan Hayward <alan.hayward@arm.com>
11141 David Sherwood <david.sherwood@arm.com>
11143 * reload1.c (spill_stack_slot_width): Change element type
11144 from unsigned int to poly_uint64_pod.
11145 (alter_reg): Treat mode sizes as polynomial.
11147 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11148 Alan Hayward <alan.hayward@arm.com>
11149 David Sherwood <david.sherwood@arm.com>
11151 * reload.c (complex_word_subreg_p): New function.
11152 (reload_inner_reg_of_subreg, push_reload): Use it.
11154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11155 Alan Hayward <alan.hayward@arm.com>
11156 David Sherwood <david.sherwood@arm.com>
11158 * lra-constraints.c (process_alt_operands): Reject matched
11159 operands whose sizes aren't ordered.
11160 (match_reload): Refer to this check here.
11162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11163 Alan Hayward <alan.hayward@arm.com>
11164 David Sherwood <david.sherwood@arm.com>
11166 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11167 that the mode size is in the set {1, 2, 4, 8, 16}.
11169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11170 Alan Hayward <alan.hayward@arm.com>
11171 David Sherwood <david.sherwood@arm.com>
11173 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11174 Use plus_constant instead of gen_rtx_PLUS.
11176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11177 Alan Hayward <alan.hayward@arm.com>
11178 David Sherwood <david.sherwood@arm.com>
11180 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11181 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11182 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11183 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11184 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11185 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11186 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11187 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11188 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11189 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11191 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11192 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11193 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11194 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11195 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11196 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11197 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11198 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11199 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11200 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11202 * expr.c (emit_move_resolve_push): Treat the input and result
11203 of PUSH_ROUNDING as a poly_int64.
11204 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11205 (emit_push_insn): Likewise.
11206 * lra-eliminations.c (mark_not_eliminable): Likewise.
11207 * recog.c (push_operand): Likewise.
11208 * reload1.c (elimination_effects): Likewise.
11209 * rtlanal.c (nonzero_bits1): Likewise.
11210 * calls.c (store_one_arg): Likewise. Require the padding to be
11211 known at compile time.
11213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11214 Alan Hayward <alan.hayward@arm.com>
11215 David Sherwood <david.sherwood@arm.com>
11217 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11218 Use plus_constant instead of gen_rtx_PLUS.
11220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11221 Alan Hayward <alan.hayward@arm.com>
11222 David Sherwood <david.sherwood@arm.com>
11224 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11225 rather than an int.
11227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11228 Alan Hayward <alan.hayward@arm.com>
11229 David Sherwood <david.sherwood@arm.com>
11231 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11232 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11233 via stack temporaries. Treat the mode size as polynomial too.
11235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11236 Alan Hayward <alan.hayward@arm.com>
11237 David Sherwood <david.sherwood@arm.com>
11239 * expr.c (expand_expr_real_2): When handling conversions involving
11240 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11241 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11242 as a poly_uint64 too.
11244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11245 Alan Hayward <alan.hayward@arm.com>
11246 David Sherwood <david.sherwood@arm.com>
11248 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11251 Alan Hayward <alan.hayward@arm.com>
11252 David Sherwood <david.sherwood@arm.com>
11254 * combine.c (can_change_dest_mode): Handle polynomial
11255 REGMODE_NATURAL_SIZE.
11256 * expmed.c (store_bit_field_1): Likewise.
11257 * expr.c (store_constructor): Likewise.
11258 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11259 and polynomial REGMODE_NATURAL_SIZE.
11260 (gen_lowpart_common): Likewise.
11261 * reginfo.c (record_subregs_of_mode): Likewise.
11262 * rtlanal.c (read_modify_subreg_p): Likewise.
11264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11265 Alan Hayward <alan.hayward@arm.com>
11266 David Sherwood <david.sherwood@arm.com>
11268 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11269 numbers of elements.
11271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11272 Alan Hayward <alan.hayward@arm.com>
11273 David Sherwood <david.sherwood@arm.com>
11275 * match.pd: Cope with polynomial numbers of vector elements.
11277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11278 Alan Hayward <alan.hayward@arm.com>
11279 David Sherwood <david.sherwood@arm.com>
11281 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11282 in a POINTER_PLUS_EXPR.
11284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11285 Alan Hayward <alan.hayward@arm.com>
11286 David Sherwood <david.sherwood@arm.com>
11288 * omp-simd-clone.c (simd_clone_subparts): New function.
11289 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11290 (ipa_simd_modify_function_body): Likewise.
11292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11293 Alan Hayward <alan.hayward@arm.com>
11294 David Sherwood <david.sherwood@arm.com>
11296 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11297 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11298 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11299 (expand_vector_condition, vector_element): Likewise.
11300 (subparts_gt): New function.
11301 (get_compute_type): Use subparts_gt.
11302 (count_type_subparts): Delete.
11303 (expand_vector_operations_1): Use subparts_gt instead of
11304 count_type_subparts.
11306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11307 Alan Hayward <alan.hayward@arm.com>
11308 David Sherwood <david.sherwood@arm.com>
11310 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11311 (vect_compile_time_alias): ...this new function. Do the calculation
11312 on poly_ints rather than trees.
11313 (vect_prune_runtime_alias_test_list): Update call accordingly.
11315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11316 Alan Hayward <alan.hayward@arm.com>
11317 David Sherwood <david.sherwood@arm.com>
11319 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11321 (vect_schedule_slp_instance): Likewise.
11323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11324 Alan Hayward <alan.hayward@arm.com>
11325 David Sherwood <david.sherwood@arm.com>
11327 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11328 constant and extern definitions for variable-length vectors.
11329 (vect_get_constant_vectors): Note that the number of units
11330 is known to be constant.
11332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11333 Alan Hayward <alan.hayward@arm.com>
11334 David Sherwood <david.sherwood@arm.com>
11336 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11337 of units as polynomial. Choose between WIDE and NARROW based
11340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11341 Alan Hayward <alan.hayward@arm.com>
11342 David Sherwood <david.sherwood@arm.com>
11344 * tree-vect-stmts.c (simd_clone_subparts): New function.
11345 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11348 Alan Hayward <alan.hayward@arm.com>
11349 David Sherwood <david.sherwood@arm.com>
11351 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11352 vectors as polynomial. Use build_index_vector for
11353 IFN_GOMP_SIMD_LANE.
11355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11356 Alan Hayward <alan.hayward@arm.com>
11357 David Sherwood <david.sherwood@arm.com>
11359 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11360 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11361 for variable-length vectors.
11362 (vectorizable_mask_load_store): Treat the number of units as
11363 polynomial, asserting that it is constant if the condition has
11364 already been enforced.
11365 (vectorizable_store, vectorizable_load): Likewise.
11367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11368 Alan Hayward <alan.hayward@arm.com>
11369 David Sherwood <david.sherwood@arm.com>
11371 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11372 of units as polynomial. Punt if we can't tell at compile time
11373 which vector contains the final result.
11375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11376 Alan Hayward <alan.hayward@arm.com>
11377 David Sherwood <david.sherwood@arm.com>
11379 * tree-vect-loop.c (vectorizable_induction): Treat the number
11380 of units as polynomial. Punt on SLP inductions. Use an integer
11381 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11382 cast of such a series for variable-length floating-point
11385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11386 Alan Hayward <alan.hayward@arm.com>
11387 David Sherwood <david.sherwood@arm.com>
11389 * tree.h (build_index_vector): Declare.
11390 * tree.c (build_index_vector): New function.
11391 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11392 of units as polynomial, forcibly converting it to a constant if
11393 vectorizable_reduction has already enforced the condition.
11394 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11395 to create a {1,2,3,...} vector.
11396 (vectorizable_reduction): Treat the number of units as polynomial.
11397 Choose vectype_in based on the largest scalar element size rather
11398 than the smallest number of units. Enforce the restrictions
11401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11402 Alan Hayward <alan.hayward@arm.com>
11403 David Sherwood <david.sherwood@arm.com>
11405 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11406 number of units as polynomial.
11408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11409 Alan Hayward <alan.hayward@arm.com>
11410 David Sherwood <david.sherwood@arm.com>
11412 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11413 * target.def (autovectorize_vector_sizes): Return the vector sizes
11414 by pointer, using vector_sizes rather than a bitmask.
11415 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11416 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11417 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11419 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11420 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11421 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11422 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11423 * omp-general.c (omp_max_vf): Likewise.
11424 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11425 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11426 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11427 * tree-vect-slp.c (vect_slp_bb): Likewise.
11428 * doc/tm.texi: Regenerate.
11429 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11431 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11432 the vector size as a poly_uint64 rather than an unsigned int.
11433 (current_vector_size): Change from an unsigned int to a poly_uint64.
11434 (get_vectype_for_scalar_type): Update accordingly.
11435 * tree.h (build_truth_vector_type): Take the size and number of
11436 units as a poly_uint64 rather than an unsigned int.
11437 (build_vector_type): Add a temporary overload that takes
11438 the number of units as a poly_uint64 rather than an unsigned int.
11439 * tree.c (make_vector_type): Likewise.
11440 (build_truth_vector_type): Take the number of units as a poly_uint64
11441 rather than an unsigned int.
11443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11444 Alan Hayward <alan.hayward@arm.com>
11445 David Sherwood <david.sherwood@arm.com>
11447 * target.def (get_mask_mode): Take the number of units and length
11448 as poly_uint64s rather than unsigned ints.
11449 * targhooks.h (default_get_mask_mode): Update accordingly.
11450 * targhooks.c (default_get_mask_mode): Likewise.
11451 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11452 * doc/tm.texi: Regenerate.
11454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11455 Alan Hayward <alan.hayward@arm.com>
11456 David Sherwood <david.sherwood@arm.com>
11458 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11459 * omp-general.c (omp_max_vf): Likewise.
11460 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11461 (expand_omp_simd): Handle polynomial safelen.
11462 * omp-low.c (omplow_simd_context): Add a default constructor.
11463 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11464 (lower_rec_simd_input_clauses): Update accordingly.
11465 (lower_rec_input_clauses): Likewise.
11467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11468 Alan Hayward <alan.hayward@arm.com>
11469 David Sherwood <david.sherwood@arm.com>
11471 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11472 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11473 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11474 (vect_analyze_slp_cost): Likewise.
11475 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11476 (vect_model_load_cost): Likewise.
11478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11479 Alan Hayward <alan.hayward@arm.com>
11480 David Sherwood <david.sherwood@arm.com>
11482 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11483 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11484 from an unsigned int * to a poly_uint64_pod *.
11485 (calculate_unrolling_factor): New function.
11486 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11489 Alan Hayward <alan.hayward@arm.com>
11490 David Sherwood <david.sherwood@arm.com>
11492 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11493 from an unsigned int to a poly_uint64.
11494 (_loop_vec_info::slp_unrolling_factor): Likewise.
11495 (_loop_vec_info::vectorization_factor): Change from an int
11497 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11498 (vect_get_num_vectors): New function.
11499 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11500 (vect_get_num_copies): Use vect_get_num_vectors.
11501 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11502 to an unsigned int *.
11503 (vect_analyze_data_refs): Change min_vf from an int * to a
11505 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11506 than an unsigned HOST_WIDE_INT.
11507 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11508 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11509 to an unsigned int *.
11510 (vect_analyze_data_ref_dependences): Likewise.
11511 (vect_compute_data_ref_alignment): Handle polynomial vf.
11512 (vect_enhance_data_refs_alignment): Likewise.
11513 (vect_prune_runtime_alias_test_list): Likewise.
11514 (vect_shift_permute_load_chain): Likewise.
11515 (vect_supportable_dr_alignment): Likewise.
11516 (dependence_distance_ge_vf): Take the vectorization factor as a
11517 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11518 (vect_analyze_data_refs): Change min_vf from an int * to a
11520 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11521 vfm1 as a poly_uint64 rather than an int. Make the same change
11522 for the returned bound_scalar.
11523 (vect_gen_vector_loop_niters): Handle polynomial vf.
11524 (vect_do_peeling): Likewise. Update call to
11525 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11526 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11528 * tree-vect-loop.c (vect_determine_vectorization_factor)
11529 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11530 (vect_get_known_peeling_cost): Likewise.
11531 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11532 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11533 (vect_transform_loop): Likewise. Use the lowest possible VF when
11534 updating the upper bounds of the loop.
11535 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11536 rather than an int.
11537 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11538 polynomial unroll factors.
11539 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11540 (vect_make_slp_decision): Likewise.
11541 (vect_supported_load_permutation_p): Likewise, and polynomial
11543 (vect_analyze_slp_cost): Handle polynomial vf.
11544 (vect_slp_analyze_node_operations): Likewise.
11545 (vect_slp_analyze_bb_1): Likewise.
11546 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11547 than an unsigned HOST_WIDE_INT.
11548 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11549 (vectorizable_load): Handle polynomial vf.
11550 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11552 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11555 Alan Hayward <alan.hayward@arm.com>
11556 David Sherwood <david.sherwood@arm.com>
11558 * match.pd: Handle bit operations involving three constants
11559 and try to fold one pair.
11561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11563 * tree-vect-loop-manip.c: Include gimple-fold.h.
11564 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11565 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11566 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11567 Add a path that uses a step of VF instead of 1, but disable it
11569 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11570 and niters_no_overflow parameters. Update calls to
11571 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11572 Create a new SSA name if the latter choses to use a ste other
11573 than zero, and return it via niters_vector_mult_vf_var.
11574 * tree-vect-loop.c (vect_transform_loop): Update calls to
11575 vect_do_peeling, vect_gen_vector_loop_niters and
11576 slpeel_make_loop_iterate_ntimes.
11577 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11578 (vect_gen_vector_loop_niters): Update declarations after above changes.
11580 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11582 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11583 128-bit round to integer instructions.
11584 (ceil<mode>2): Likewise.
11585 (btrunc<mode>2): Likewise.
11586 (round<mode>2): Likewise.
11588 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11590 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11591 unaligned VSX load/store on P8/P9.
11592 (expand_block_clear): Allow the use of unaligned VSX
11593 load/store on P8/P9.
11595 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11597 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11599 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11600 swap associated with both a load and a store.
11602 2018-01-02 Andrew Waterman <andrew@sifive.com>
11604 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11605 * config/riscv/riscv.md (clear_cache): Use it.
11607 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11609 * web.c: Remove out-of-date comment.
11611 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11613 * expr.c (fixup_args_size_notes): Check that any existing
11614 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11615 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11616 (emit_single_push_insn): ...here.
11618 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11620 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11621 (const_vector_encoded_nelts): New function.
11622 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11623 (const_vector_int_elt, const_vector_elt): Declare.
11624 * emit-rtl.c (const_vector_int_elt_1): New function.
11625 (const_vector_elt): Likewise.
11626 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11627 of CONST_VECTOR_ELT.
11629 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11631 * expr.c: Include rtx-vector-builder.h.
11632 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11633 directly on the tree encoding.
11634 (const_vector_from_tree): Likewise.
11635 * optabs.c: Include rtx-vector-builder.h.
11636 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11637 sequence of "u" values.
11638 * vec-perm-indices.c: Include rtx-vector-builder.h.
11639 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11640 directly on the vec_perm_indices encoding.
11642 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11644 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11645 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11646 * rtx-vector-builder.h: New file.
11647 * rtx-vector-builder.c: Likewise.
11648 * rtl.h (rtx_def::u2): Add a const_vector field.
11649 (CONST_VECTOR_NPATTERNS): New macro.
11650 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11651 (CONST_VECTOR_DUPLICATE_P): Likewise.
11652 (CONST_VECTOR_STEPPED_P): Likewise.
11653 (CONST_VECTOR_ENCODED_ELT): Likewise.
11654 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11655 (unwrap_const_vec_duplicate): Likewise.
11656 (const_vec_series_p): Check for a non-duplicated vector encoding.
11657 Say that the function only returns true for integer vectors.
11658 * emit-rtl.c: Include rtx-vector-builder.h.
11659 (gen_const_vec_duplicate_1): Delete.
11660 (gen_const_vector): Call gen_const_vec_duplicate instead of
11661 gen_const_vec_duplicate_1.
11662 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11663 (gen_const_vec_duplicate): Use rtx_vector_builder.
11664 (gen_const_vec_series): Likewise.
11665 (gen_rtx_CONST_VECTOR): Likewise.
11666 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11667 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11668 Build a new vector rather than modifying a CONST_VECTOR in-place.
11669 (handle_special_swappables): Update call accordingly.
11670 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11671 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11672 Build a new vector rather than modifying a CONST_VECTOR in-place.
11673 (handle_special_swappables): Update call accordingly.
11675 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11677 * simplify-rtx.c (simplify_const_binary_operation): Use
11678 CONST_VECTOR_ELT instead of XVECEXP.
11680 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11682 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11683 the selector elements to be different from the data elements
11684 if the selector is a VECTOR_CST.
11685 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11686 ssizetype for the selector.
11688 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11690 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11691 before testing each element individually.
11692 * tree-vect-generic.c (lower_vec_perm): Likewise.
11694 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11696 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11697 * selftest-run-tests.c (selftest::run_tests): Call it.
11698 * vector-builder.h (vector_builder::operator ==): New function.
11699 (vector_builder::operator !=): Likewise.
11700 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11701 (vec_perm_indices::all_from_input_p): New function.
11702 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11703 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11704 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11705 instead of reading the VECTOR_CST directly. Detect whether both
11706 vector inputs are the same before constructing the vec_perm_indices,
11707 and update the number of inputs argument accordingly. Use the
11708 utility functions added above. Only construct sel2 if we need to.
11710 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11712 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11713 the broadcast of the low byte.
11714 (expand_mult_highpart): Use an explicit encoding for the permutes.
11715 * optabs-query.c (can_mult_highpart_p): Likewise.
11716 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11717 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11718 (vectorizable_bswap): Likewise.
11719 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11720 explicit encoding for the power-of-2 permutes.
11721 (vect_permute_store_chain): Likewise.
11722 (vect_grouped_load_supported): Likewise.
11723 (vect_permute_load_chain): Likewise.
11725 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11727 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11728 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11729 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11730 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11731 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11732 (vect_gen_perm_mask_any): Likewise.
11734 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11736 * int-vector-builder.h: New file.
11737 * vec-perm-indices.h: Include int-vector-builder.h.
11738 (vec_perm_indices): Redefine as an int_vector_builder.
11739 (auto_vec_perm_indices): Delete.
11740 (vec_perm_builder): Redefine as a stand-alone class.
11741 (vec_perm_indices::vec_perm_indices): New function.
11742 (vec_perm_indices::clamp): Likewise.
11743 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11744 (vec_perm_indices::new_vector): New function.
11745 (vec_perm_indices::new_expanded_vector): Update for new
11746 vec_perm_indices class.
11747 (vec_perm_indices::rotate_inputs): New function.
11748 (vec_perm_indices::all_in_range_p): Operate directly on the
11749 encoded form, without computing elided elements.
11750 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11751 encoding. Update for new vec_perm_indices class.
11752 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11753 the given vec_perm_builder.
11754 (expand_vec_perm_var): Update vec_perm_builder constructor.
11755 (expand_mult_highpart): Use vec_perm_builder instead of
11756 auto_vec_perm_indices.
11757 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11758 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11759 or double series encoding as appropriate.
11760 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11761 vec_perm_indices instead of auto_vec_perm_indices.
11762 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11763 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11764 (vect_permute_store_chain): Likewise.
11765 (vect_grouped_load_supported): Likewise.
11766 (vect_permute_load_chain): Likewise.
11767 (vect_shift_permute_load_chain): Likewise.
11768 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11769 (vect_transform_slp_perm_load): Likewise.
11770 (vect_schedule_slp_instance): Likewise.
11771 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11772 (vectorizable_mask_load_store): Likewise.
11773 (vectorizable_bswap): Likewise.
11774 (vectorizable_store): Likewise.
11775 (vectorizable_load): Likewise.
11776 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11777 vec_perm_indices instead of auto_vec_perm_indices. Use
11778 tree_to_vec_perm_builder to read the vector from a tree.
11779 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11780 vec_perm_builder instead of a vec_perm_indices.
11781 (have_whole_vector_shift): Use vec_perm_builder and
11782 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11783 truncation to calc_vec_perm_mask_for_shift.
11784 (vect_create_epilog_for_reduction): Likewise.
11785 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11786 from auto_vec_perm_indices to vec_perm_indices.
11787 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11788 instead of changing individual elements.
11789 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11790 the vector in d.perm.
11791 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11792 from auto_vec_perm_indices to vec_perm_indices.
11793 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11794 instead of changing individual elements.
11795 (arm_vectorize_vec_perm_const): Use new_vector to install
11796 the vector in d.perm.
11797 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11798 Update vec_perm_builder constructor.
11799 (rs6000_expand_interleave): Likewise.
11800 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11801 (rs6000_expand_interleave): Likewise.
11803 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11805 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11806 to qimode could truncate the indices.
11807 * optabs.c (expand_vec_perm_var): Likewise.
11809 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11811 * Makefile.in (OBJS): Add vec-perm-indices.o.
11812 * vec-perm-indices.h: New file.
11813 * vec-perm-indices.c: Likewise.
11814 * target.h (vec_perm_indices): Replace with a forward class
11816 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11817 * optabs.h: Include vec-perm-indices.h.
11818 (expand_vec_perm): Delete.
11819 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11820 (expand_vec_perm_const): Declare.
11821 * target.def (vec_perm_const_ok): Replace with...
11822 (vec_perm_const): ...this new hook.
11823 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11824 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11825 * doc/tm.texi: Regenerate.
11826 * optabs.def (vec_perm_const): Delete.
11827 * doc/md.texi (vec_perm_const): Likewise.
11828 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11829 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11830 expand_vec_perm for constant permutation vectors. Assert that
11831 the mode of variable permutation vectors is the integer equivalent
11832 of the mode that is being permuted.
11833 * optabs-query.h (selector_fits_mode_p): Declare.
11834 * optabs-query.c: Include vec-perm-indices.h.
11835 (selector_fits_mode_p): New function.
11836 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11837 is defined, instead of checking whether the vec_perm_const_optab
11838 exists. Use targetm.vectorize.vec_perm_const instead of
11839 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11840 fit in the vector mode before using a variable permute.
11841 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11842 vec_perm_indices instead of an rtx.
11843 (expand_vec_perm): Replace with...
11844 (expand_vec_perm_const): ...this new function. Take the selector
11845 as a vec_perm_indices rather than an rtx. Also take the mode of
11846 the selector. Update call to shift_amt_for_vec_perm_mask.
11847 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11848 Use vec_perm_indices::new_expanded_vector to expand the original
11849 selector into bytes. Check whether the indices fit in the vector
11850 mode before using a variable permute.
11851 (expand_vec_perm_var): Make global.
11852 (expand_mult_highpart): Use expand_vec_perm_const.
11853 * fold-const.c: Includes vec-perm-indices.h.
11854 * tree-ssa-forwprop.c: Likewise.
11855 * tree-vect-data-refs.c: Likewise.
11856 * tree-vect-generic.c: Likewise.
11857 * tree-vect-loop.c: Likewise.
11858 * tree-vect-slp.c: Likewise.
11859 * tree-vect-stmts.c: Likewise.
11860 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11862 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11863 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11864 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11865 (aarch64_vectorize_vec_perm_const): ...this new function.
11866 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11867 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11868 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11869 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11870 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11871 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11872 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11874 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11875 check for NEON modes.
11876 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11877 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11878 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11879 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11881 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11882 the old VEC_PERM_CONST conditions.
11883 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11884 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11885 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11886 (ia64_vectorize_vec_perm_const_ok): Merge into...
11887 (ia64_vectorize_vec_perm_const): ...this new function.
11888 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11889 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11890 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11891 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11892 * config/mips/mips.c (mips_expand_vec_perm_const)
11893 (mips_vectorize_vec_perm_const_ok): Merge into...
11894 (mips_vectorize_vec_perm_const): ...this new function.
11895 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11896 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11897 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11898 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11899 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11900 (rs6000_expand_vec_perm_const): Delete.
11901 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11903 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11904 (altivec_expand_vec_perm_const_le): Take each operand individually.
11905 Operate on constant selectors rather than rtxes.
11906 (altivec_expand_vec_perm_const): Likewise. Update call to
11907 altivec_expand_vec_perm_const_le.
11908 (rs6000_expand_vec_perm_const): Delete.
11909 (rs6000_vectorize_vec_perm_const_ok): Delete.
11910 (rs6000_vectorize_vec_perm_const): New function.
11911 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11912 an element count and rtx array.
11913 (rs6000_expand_extract_even): Update call accordingly.
11914 (rs6000_expand_interleave): Likewise.
11915 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11916 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11917 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11918 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11919 (rs6000_expand_vec_perm_const): Delete.
11920 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11921 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11922 (altivec_expand_vec_perm_const_le): Take each operand individually.
11923 Operate on constant selectors rather than rtxes.
11924 (altivec_expand_vec_perm_const): Likewise. Update call to
11925 altivec_expand_vec_perm_const_le.
11926 (rs6000_expand_vec_perm_const): Delete.
11927 (rs6000_vectorize_vec_perm_const_ok): Delete.
11928 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11929 reference to the SPE evmerge intructions.
11930 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11931 an element count and rtx array.
11932 (rs6000_expand_extract_even): Update call accordingly.
11933 (rs6000_expand_interleave): Likewise.
11934 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11935 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11937 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11939 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11941 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11942 vector mode and that that mode matches the mode of the data
11944 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11945 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11946 directly using expand_vec_perm_1 when forcing selectors into
11948 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11950 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11952 * optabs-query.h (can_vec_perm_p): Delete.
11953 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11954 * optabs-query.c (can_vec_perm_p): Split into...
11955 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11956 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11957 particular selector is valid.
11958 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11959 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11960 (vect_grouped_load_supported): Likewise.
11961 (vect_shift_permute_load_chain): Likewise.
11962 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11963 (vect_transform_slp_perm_load): Likewise.
11964 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11965 (vectorizable_bswap): Likewise.
11966 (vect_gen_perm_mask_checked): Likewise.
11967 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11968 implementations of variable permutation vectors into account
11969 when deciding which selector to use.
11970 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11971 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11972 with a false third argument.
11973 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11974 to test whether the constant selector is valid and can_vec_perm_var_p
11975 to test whether a variable selector is valid.
11977 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11979 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11980 * optabs-query.c (can_vec_perm_p): Likewise.
11981 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11982 instead of vec_perm_indices.
11983 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11984 (vect_gen_perm_mask_checked): Likewise,
11985 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11986 (vect_gen_perm_mask_checked): Likewise,
11988 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11990 * optabs-query.h (qimode_for_vec_perm): Declare.
11991 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11992 (qimode_for_vec_perm): ...this new function.
11993 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11995 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11997 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11998 does not have a conditional at the top.
12000 2018-01-02 Richard Biener <rguenther@suse.de>
12002 * ipa-inline.c (big_speedup_p): Fix expression.
12004 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12007 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12010 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12014 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12015 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12016 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12017 cond_taken_branch_cost 3->4.
12019 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12021 PR tree-optimization/83581
12022 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12023 TODO_cleanup_cfg if any changes have been made.
12025 PR middle-end/83608
12026 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12027 convert_modes if target mode has the right side, but different mode
12030 PR middle-end/83609
12031 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12032 last argument when extracting from CONCAT. If either from_real or
12033 from_imag is NULL, use expansion through memory. If result is not
12034 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12035 the parts directly to inner mode, if even that fails, use expansion
12038 PR middle-end/83623
12039 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12040 check for bswap in mode rather than HImode and use that in expand_unop
12043 Copyright (C) 2018 Free Software Foundation, Inc.
12045 Copying and distribution of this file, with or without modification,
12046 are permitted in any medium without royalty provided the copyright
12047 notice and this notice are preserved.