1 2018-05-02 Richard Biener <rguenther@suse.de>
3 PR tree-optimization/85597
4 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5 do not use split vect_get_vec_defs call but call vect_get_slp_defs
8 2018-05-02 Tom de Vries <tom@codesourcery.com>
11 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
12 dump files): Add ltrans-tree.
14 2018-05-02 Tom de Vries <tom@codesourcery.com>
17 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18 dump files): Add wpa-ipa.
20 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
22 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
23 powerpc*-*-linux*paired* target.
24 * config/rs6000/750cl.h: Delete.
25 * config/rs6000/paired.h: Delete.
26 * config/rs6000/paired.md: Delete.
27 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
29 * config/rs6000/rs6000-builtin.def: Remove paired float support.
30 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
31 comment. Remove paired float support.
32 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
33 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
35 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
36 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
38 * config/rs6000/rs6000.c: Remove paired float support.
39 (paired_expand_vector_init, paired_expand_vector_move,
40 paired_emit_vector_compare, paired_emit_vector_cond_expr,
41 (paired_expand_lv_builtin, paired_expand_stv_builtin,
42 paired_expand_builtin, paired_expand_predicate_builtin,
43 paired_init_builtins): Delete.
44 * config/rs6000/rs6000.h: Remove paired float support.
45 * config/rs6000/rs6000.md: Remove paired float support.
46 (move_from_CR_ov_bit): Delete.
47 * config/rs6000/rs6000.opt (mpaired): Delete.
48 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
49 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
51 2018-05-02 Richard Biener <rguenther@suse.de>
54 * gimplify.c (gimplify_save_expr): When in SSA form allow
55 SAVE_EXPRs to compute to SSA vars.
57 2018-05-02 Jakub Jelinek <jakub@redhat.com>
60 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
61 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
62 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
63 clobber operands[2], instead use a new pseudo. Formatting fixes.
65 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
67 PR tree-optimization/85586
68 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
69 exit early for statements in the same group if the accesses are
72 2018-05-02 Tom de Vries <tom@codesourcery.com>
75 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
78 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
80 PR tree-optimization/85143
81 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
83 2018-05-01 Tom de Vries <tom@codesourcery.com>
86 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
87 not found" error message.
89 2018-05-01 Tom de Vries <tom@codesourcery.com>
92 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
93 * vec.c (test_ordered_remove_if): New function.
94 (vec_c_tests): Call test_ordered_remove_if.
95 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
96 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
97 * tree-vect-patterns.c (vect_pattern_recog_1): Use
98 VEC_ORDERED_REMOVE_IF.
100 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
102 PR tree-optimization/82665
103 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
104 pointer subtraction where arguments come from a memchr call.
106 2018-05-01 Jakub Jelinek <jakub@redhat.com>
108 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
109 --push-state --as-needed and --pop-state instead of --as-needed and
110 --no-as-needed if ld supports it.
111 * configure: Regenerated.
114 * doc/install.texi2html: Replace _002d with - and _002a with * in
115 generated html files using sed.
117 2018-04-30 David Malcolm <dmalcolm@redhat.com>
120 * gcc-rich-location.c (blank_line_before_p): New function.
121 (use_new_line): New function.
122 (gcc_rich_location::add_fixit_insert_formatted): New function.
123 * gcc-rich-location.h
124 (gcc_rich_location::add_fixit_insert_formatted): New function.
126 2018-04-30 David Malcolm <dmalcolm@redhat.com>
128 * selftest.c (assert_streq): Rename "expected" and "actual" to
129 "val1" and "val2". Extend NULL-handling to cover both inputs
130 symmetrically, while still requiring both to be non-NULL for a pass.
131 * selftest.h (assert_streq): Rename "expected" and "actual" to
133 (ASSERT_EQ): Likewise.
134 (ASSERT_EQ_AT): Likewise.
135 (ASSERT_KNOWN_EQ): Likewise.
136 (ASSERT_KNOWN_EQ_AT): Likewise.
137 (ASSERT_NE): Likewise.
138 (ASSERT_MAYBE_NE): Likewise.
139 (ASSERT_MAYBE_NE_AT): Likewise.
140 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
141 the assertion to pass.
142 (ASSERT_STREQ_AT): Likewise.
144 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
146 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
147 interaction with -pie.
149 2018-04-30 David Malcolm <dmalcolm@redhat.com>
151 * selftest.h: Fix alphabetization of per-source-file selftest
154 2018-04-30 Jason Merrill <jason@redhat.com>
156 PR c++/61982 - dead stores to destroyed objects.
157 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
160 2018-04-30 Jason Merrill <jason@redhat.com>
162 * tree.c (build_clobber): New.
163 * tree.h: Declare it.
164 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
166 2018-04-30 David Malcolm <dmalcolm@redhat.com>
168 * diagnostic-show-locus.c (layout::layout): Update for
169 location_get_source_line returning a char_span.
170 (struct char_span): Move to input.h.
171 (struct correction): Update for fields in char_span becoming
173 (struct source_line): Update for location_get_source_line
174 returning a char_span.
175 (layout::print_line): Likewise.
176 * edit-context.c (edited_file::print_content): Likewise.
177 (edited_file::print_diff_hunk): Likewise.
178 (edited_file::print_run_of_changed_lines): Likewise.
179 (edited_file::get_num_lines): Likewise.
180 (edited_line::edited_line): Likewise.
181 * final.c (asm_show_source): Likewise.
182 * input.c (location_get_source_line): Convert return type
183 from const char * to char_span, losing the final "line_len"
185 (dump_location_info): Update for the above.
186 (get_substring_ranges_for_loc): Likewise. Use a char_span
187 when handling the literal within the line.
188 (test_reading_source_line): Update for location_get_source_line
189 returning a char_span.
190 * input.h (class char_span): Move here from
191 diagnostic-show-locus.c, converting from a struct to a class.
192 Make data members private.
193 (char_span::operator bool): New.
194 (char_span::length): New.
195 (char_span::get_buffer): New.
196 (char_span::operator[]): New.
197 (char_span::subspan): Make const.
198 (char_span::xstrdup): New.
199 (location_get_source_line): Convert return type from const char *
200 to char_span, losing the final "line_size" param.
202 2018-04-30 Jan Hubicka <jh@suse.cz>
204 * lto-wrapper.c (ltrans_priorities): New static var.
206 (run_gcc): Read priorities and if doing parallel build order
207 the Makefile by them.
209 2018-04-30 David Malcolm <dmalcolm@redhat.com>
211 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
213 2018-04-30 Richard Biener <rguenther@suse.de>
215 * tree-cfg.c (verify_address): Remove base argument, add
216 flag whether to check TREE_ADDRESSABLE and do that.
217 (verify_expr): Remove.
218 (verify_types_in_gimple_reference): Add pieces from verify_expr.
219 (verify_gimple_assign_single): Likewise.
220 (verify_gimple_switch): Likewise.
221 (verify_expr_location_1): Dereference tp once. Add (disabled)
222 piece from verify_expr.
223 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
225 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
227 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
229 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
231 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
232 (small_data_pattern): Likewise.
233 (arc_rewrite_small_data): Likewise.
234 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
235 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
236 (get_symbol_alignment): New function.
237 (legitimate_small_data_address_p): Likewise.
238 (legitimate_scaled_address): Update, call
239 legitimate_small_data_address_p.
240 (output_sdata): New static variable.
241 (arc_print_operand): Update how we handle small data operands.
242 (arc_print_operand_address): Likewise.
243 (arc_legitimate_address_p): Update, use
244 legitimate_small_data_address_p.
245 (arc_rewrite_small_data_p): Remove.
246 (arc_rewrite_small_data_1): Likewise.
247 (arc_rewrite_small_data): Likewise.
248 (small_data_pattern): Likewise.
249 (compact_sda_memory_operand): Update to use
250 legitimate_small_data_address_p and get_symbol_alignment.
251 (prepare_move_operands): Don't rewite sdata pattern.
252 (prepare_extend_operands): Remove.
253 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
255 (zero_extendqisi2): Likewise.
256 (zero_extendhisi2): Likewise.
257 (extendqihi2): Likewise.
258 (extendqisi2): Likewise.
259 (extendhisi2): Likewise.
263 * config/arc/constraints.md (Usd): Change it to memory constraint.
265 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
267 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
268 as source of std instructions.
269 * config/arc/arc.md (movsi_insn): Update pattern predicate to
270 allow 6-bit constants as source for store instructions.
271 (movdi_insn): Update instruction pattern to allow 6-bit constants
272 as source for store instructions.
274 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
276 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
278 2018-04-30 Nathan Sidwell <nathan@acm.org>
279 Sandra Loosemore <sandra@codesourcery.com>
281 * dumpfile.c (dump_open): Allow '-' for stdout.
282 * doc/invoke.texi (Developer Options): Document dump filename
283 determination early. Document stdin/stdout selection.
285 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
287 Microblaze Target: PIC data text relative
289 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
290 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
292 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
293 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
294 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
295 New addressing mode for data-text relative position indepenedent code.
296 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
297 'ADDRESS_SYMBOLIC_TXT_REL'.
298 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
299 (microblaze_legitimate_pic_operand): Exclude function calls from
300 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
301 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
303 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
304 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
305 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
306 for 'address + offset'.
307 (microblaze_expand_prologue): Add new function prologue call for
309 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
310 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
311 table in case of TARGET_PIC_DATA_TEXT_REL.
312 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
313 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
314 Add new macros 'UNSPEC_TEXT',
315 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
316 + exclude function calls from 'UNSPEC_PLT' in case of data text
318 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
319 new target hook for generating address diff vector tables in case of
321 * doc/tm.texi : Regenerate.
322 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
323 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
324 of addr diff vector generation.
325 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
326 target hook definition.
327 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
328 Add default function for generate_pic_addr_diff_vec -> flag_pic.
329 * doc/invoke.texi (Add new pic option): Add new microblaze pic
330 option for data text relative.
332 2018-04-30 Richard Biener <rguenther@suse.de>
334 * tree-chrec.h (evolution_function_is_constant_p): Remove
336 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
338 2018-04-30 Richard Biener <rguenther@suse.de>
341 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
343 2018-04-30 Richard Biener <rguenther@suse.de>
345 PR tree-optimization/28364
346 PR tree-optimization/85275
347 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
348 copying first exit test.
350 2018-04-28 Mark Wielaard <mark@klomp.org>
352 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
354 (dwarf_AT): Handle DW_AT_addr_base.
355 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
357 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
360 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
361 (*ashl<dwi>3_doubleword_mask_1): Ditto.
362 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
363 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
365 2018-04-28 Richard Biener <rguenther@suse.de>
367 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
368 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
369 to reflect use. Only add interesting stmts.
371 2018-04-27 Martin Jambor <mjambor@suse.cz>
374 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
375 the jump function allows for passing through aggregate values.
377 2018-04-27 David Malcolm <dmalcolm@redhat.com>
379 * input.h (in_system_header_at): Convert from macro to inline
381 (from_macro_expansion_at): Likewise.
382 (from_macro_definition_at): Likewise.
384 2018-04-27 Jeff Law <law@redhat.com>
386 * config.gcc: Mark tile* targets as deprecated/obsolete.
388 2018-04-27 Richard Biener <rguenther@suse.de>
390 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
393 2018-04-27 Richard Biener <rguenther@suse.de>
395 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
397 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
399 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
400 with Yd constraint. Set "preferred_for_speed" attribute from
401 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
403 (*movdi_internal): Ditto.
404 (movti_interunit splitters): Remove
405 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
406 (movdi_interunit splitters): Ditto.
407 * config/i386/constraints.md (Ye): Remove.
408 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
410 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
413 * config/aarch64/constraints.md (Usg): Limit to 31.
416 2018-04-27 Jakub Jelinek <jakub@redhat.com>
418 PR tree-optimization/85529
419 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
420 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
421 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
422 zero extension or masking of the MSB bit.
423 (optimize_range_tests): Add FIRST_BB argument, pass it through
424 to optimize_range_tests_var_bound.
425 (maybe_optimize_range_tests, reassociate_bb): Adjust
426 optimize_range_tests callers.
428 2018-04-26 Richard Biener <rguenther@suse.de>
429 Jakub Jelinek <jakub@redhat.com>
431 * cgraph.h (symbol_table): Just declare debug method here.
432 * symtab.c (symbol_table::debug): Define.
434 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
436 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
438 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
440 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
441 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
442 (*movdi_internal): Substitute Yi and Yj constraint with x
443 and Ym and Yn constraint with y constraint. Update "isa"
444 attribute and set "preferred_for_speed" attribute from
445 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
446 (*movsi_internal): Ditto.
447 (*movdf_internal): Ditto.
448 (*movsf_internal): Ditto.
449 (*zero_extendsidi2): Ditto.
450 * config/i386/sse.md (vec_set<mode>_0): Ditto.
451 (sse2_loadld): Ditto.
452 (*vec_extract<ssevecmodelower>_0): Ditto.
453 (*vec_extractv4si_0_zext_sse4): Ditto.
454 (vec_concatv2di): Ditto.
455 (*vec_dup<mode>): Ditto.
456 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
457 * config/i386/constraints.md (Yi): Remove.
462 2018-04-26 Nathan Sidwell <nathan@acm.org>
464 * dumpfile.c (dump_open): New.
465 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
466 (dump_finish): Detect stdio/stderr by value not name.
468 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
470 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
472 2018-04-26 Tom de Vries <tom@codesourcery.com>
475 * config/nvptx/nvptx.c (verify_neutering_jumps)
476 (verify_neutering_labels): New function
477 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
479 2018-04-26 Tom de Vries <tom@codesourcery.com>
482 * config/nvptx/nvptx.c (needs_neutering_p): New function.
483 (nvptx_single): Use needs_neutering_p to skip over insns that do not
486 2018-04-26 Richard Biener <rguenther@suse.de>
487 Tom de Vries <tom@codesourcery.com>
490 * lto-streamer-out.c (output_function): Fixup loops if required to match
491 discovery done in the reader.
493 2018-04-26 Richard Biener <rguenther@suse.de>
495 PR tree-optimization/85116
496 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
497 have a loop exit from the single latch predecessor. Remove
498 case of header with just condition.
499 (ch_base::copy_headers): Exclude infinite loops from any
501 (pass_ch::execute): Record exits.
503 2018-04-26 Richard Biener <rguenther@suse.de>
505 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
506 prologue cost vector and pass it to vect_get_load_cost.
507 (vect_get_peeling_costs_all_drs): Likewise.
508 (vect_peeling_hash_get_lowest_cost): Likewise.
509 (vect_enhance_data_refs_alignment): Likewise.
511 2018-04-26 Richard Biener <rguenther@suse.de>
514 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
515 checking of integer<->pointer conversions.
516 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
517 sign-/zero-extending pointer types.
518 (expand_omp_for_static_chunk): Likewise.
520 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
521 Jean Lee <xiaoyur347@gmail.com>
523 * config/mips/mips.c (mips_asan_shadow_offset): New function.
524 (TARGET_ASAN_SHADOW_OFFSET): Define.
525 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
526 true for -fsanitize=address.
528 2018-04-25 Mark Wielaard <mark@klomp.org>
530 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
533 2018-04-25 Jakub Jelinek <jakub@redhat.com>
535 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
536 than "alu", remove explicit "memory" and "imm_disp" attributes.
537 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
540 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
541 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
544 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
547 * config/i386/i386.c (ix86_expand_builtin): Change memory
548 operand to XI, extend p0 to Pmode.
549 * config/i386/i386.md: Change unspec volatile and operand
550 1 mode to XI, change operand 0 mode to P.
552 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
554 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
555 GET_MODE_MASK before any checking.
556 (nds32_can_use_bset_p): Likewise.
557 (nds32_can_use_btgl_p): Likewise.
559 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
561 * config/nds32/nds32-doubleword.md: New define_split pattern for
562 illegal register number.
564 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
566 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
568 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
570 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
572 2018-04-25 Richard Biener <rguenther@suse.de>
574 * lto-streamer.h (LTO_major_version): Bump to 8.
576 2018-04-25 Jakub Jelinek <jakub@redhat.com>
578 * BASE-VER: Set to 9.0.0.
580 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
582 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
583 in __abskf2 and __powikf2.
585 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
588 * config/aarch64/constraints.md (Usg, Usj): New constraints.
589 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
590 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
591 Use the above on operand 2. Reindent.
592 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
594 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
597 * common/config/i386/i386-common.c (ix86_handle_option): Don't
599 * config/i386/i386.opt (mcet): Removed.
600 * doc/install.texi: Remove -mcet documentation.
601 * doc/invoke.texi: Likewise.
603 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
606 * doc/install.texi: Remove -mcet from bootstrap-cet.
608 2018-04-24 Jakub Jelinek <jakub@redhat.com>
611 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
612 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
616 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
617 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
618 containing a CONST_VECTOR.
620 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
622 * doc/install.texi: Update newlib dependency for nvptx.
624 2018-04-24 Jakub Jelinek <jakub@redhat.com>
627 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
628 instead of INTVAL when shifting x left.
630 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
632 PR tree-optimization/85478
633 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
634 vect_grouped_store_supported for single element vectors.
636 2018-04-24 Richard Biener <rguenther@suse.de>
639 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
640 load cost increase to the case of non-constant step.
642 2018-04-24 Jakub Jelinek <jakub@redhat.com>
645 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
646 destination if any_malformed_asm.
648 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
651 * expr.c (store_field): In the bitfield case, if the value comes from
652 a function call and is returned in registers by means of a PARALLEL,
653 do not change the mode of the temporary unless BLKmode and VOIDmode.
655 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
657 PR rtl-optimization/85423
658 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
659 dependencies to debug insns when the previous insn is non-debug.
661 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
663 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
664 enums into a single definition.
665 (fls): Fix predicates and printing.
668 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
670 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
671 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
672 and short u6 immediate.
673 (check_if_valid_sleep_operand): Remove.
674 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
676 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
678 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
679 flag_always_save_lp condition.
680 * config/nds32/nds32.opt (malways-save-lp): New option.
682 2018-04-22 Shiva Chen <shiva0217@gmail.com>
684 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
685 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
686 * config/nds32/nds32.h
687 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
688 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
690 2018-04-22 Shiva Chen <shiva0217@gmail.com>
692 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
693 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
695 2018-04-22 Shiva Chen <shiva0217@gmail.com>
696 Chung-Ju Wu <jasonwucj@gmail.com>
698 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
700 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
701 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
703 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
705 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
707 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
709 * config/nds32/nds32-protos.h (nds32_data_alignment,
710 nds32_local_alignment): Declare.
711 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
712 nds32_local_alignment): New functions.
713 (TARGET_CONSTANT_ALIGNMENT): Define.
714 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
716 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
718 * config/nds32/nds32.c
719 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
720 (TARGET_MODES_TIEABLE_P): Likewise.
722 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
724 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
727 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
728 Chung-Ju Wu <jasonwucj@gmail.com>
730 * config/nds32/constants.md (unspec_volatile_element): Add enum values
731 for unaligned access.
732 * config/nds32/nds32-intrinsic.c: Implementation of expanding
734 * config/nds32/nds32-intrinsic.md: Likewise.
735 * config/nds32/nds32_intrinsic.h: Likewise.
736 * config/nds32/nds32.h (nds32_builtins): Likewise.
737 * config/nds32/nds32.opt (munaligned-access): New option.
738 * config/nds32/nds32.c (nds32_asm_file_start): Display
739 flag_unaligned_access status.
741 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
743 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
744 -mno-relax is present.
745 * config/riscv/linux.h (LINK_SPEC): Ditto.
747 2018-04-20 Martin Sebor <msebor@redhat.com>
750 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
752 (gimple_fold_builtin_stxcpy_chk): Same.
753 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
755 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
758 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
759 __powikf2 when long double is IEEE 128-bit.
761 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
763 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
764 step to make sure stack always aligned.
766 2018-04-20 Carl Love <cel@us.ibm.com>
769 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
772 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
773 Tom de Vries <tom@codesourcery.com>
776 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
777 Emit insns for calls too.
778 (nvptx_find_par): Always look for worker-level predecessor insn.
779 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
781 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
782 (nvptx_process_pars): Propagate frames for calls.
784 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
787 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
789 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
790 (ix86_handle_option): Don't handle OPT_mibt.
791 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
793 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
795 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
796 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
797 (ix86_target_macros): Define __CET__ with flag_cf_protection
799 * config/i386/i386.c (isa2_opts): Remove -mibt.
800 * config/i386/i386.h (TARGET_IBT): Removed.
801 (TARGET_IBT_P): Likewise.
802 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
803 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
804 * config/i386/i386.opt (mcet): Update help message.
807 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
808 -mcet as an alias for -mshstk.
810 2018-04-20 Richard Biener <rguenther@suse.de>
813 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
814 complexity by forcing a single use of the multiply operand.
816 2018-04-20 Martin Jambor <mjambor@suse.cz>
819 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
820 recursion dependency to only apply to non-clones.
822 2018-04-20 Martin Jambor <mjambor@suse.cz>
825 * ipa-cp.c (create_specialized_node): Check that clones of
826 self-recursive edges exist during IPA-CP.
828 2018-04-19 Toon Moene <toon@moene.org>
830 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
833 2018-04-19 Jakub Jelinek <jakub@redhat.com>
835 PR tree-optimization/85467
836 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
837 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
838 VECTOR_CST element to type.
840 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
843 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
844 * config/i386/i386.md (builtin_setjmp_setup): Removed.
845 (builtin_longjmp): Likewise.
846 (save_stack_nonlocal): New pattern.
847 (restore_stack_nonlocal): Likewise.
849 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
852 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
853 Replace ASM_OUTPUT_LABEL with fprintf.
855 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
858 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
859 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
860 * config/i386/i386-c.c (ix86_target_macros_internal): Also
861 define __IBT__ and __SHSTK__ for -fcf-protection.
862 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
864 (ix86_trampoline_init): Likewise.
865 (x86_output_mi_thunk): Likewise.
866 (ix86_notrack_prefixed_insn_p): Likewise.
867 (ix86_option_override_internal): Don't disallow -fcf-protection.
868 * config/i386/i386.md (rdssp<mode>): Also enable for
870 (incssp<mode>): Likewise.
871 (nop_endbr): Likewise.
872 * config/i386/i386.opt (mcet): Change help message to built-in
876 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
877 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
878 enable CET built-in functions.
880 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
882 * common/config/i386/i386-common.c
883 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
884 OPTION_MASK_ISA_MOVDIRI_UNSET,
885 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
886 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
887 * config.gcc (movdirintrin.h): New header.
888 * config/i386/cpuid.h (bit_MOVDIRI,
889 bit_MOVDIR64B): New bits.
890 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
892 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
893 (VOID, PVOID, PCVOID)): New function types.
894 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
895 __builtin_ia32_directstoreu_u64,
896 __builtin_ia32_movdir64b): New builtins.
897 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
898 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
900 (ix86_valid_target_attribute_inner_p): Ditto.
901 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
902 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
903 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
904 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
905 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
906 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
907 (movdiri<mode>, movdir64b_<mode>): New.
908 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
909 * config/i386/immintrin.h: Include movdirintrin.h.
910 * config/i386/movdirintrin.h: New file.
911 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
913 2018-04-19 Richard Biener <rguenther@suse.de>
916 * cfg.c (clear_bb_flags): When loop state says we have
917 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
919 2018-04-19 Richard Biener <rguenther@suse.de>
921 PR tree-optimization/84737
922 * tree-vect-data-refs.c (vect_copy_ref_info): New function
923 copying restrict info.
924 (vect_setup_realignment): Use it.
925 * tree-vectorizer.h (vect_copy_ref_info): Declare.
926 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
927 the first DR to all generated stores.
928 (vectorizable_load): Likewise for loads.
930 2018-04-19 Jakub Jelinek <jakub@redhat.com>
932 PR tree-optimization/85446
933 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
934 the integral and pointer types to have the same precision.
936 * doc/install.texi: Document --disable-cet being the default and
939 2018-04-18 Martin Liska <mliska@suse.cz>
941 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
944 2018-04-18 Martin Liska <mliska@suse.cz>
947 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
950 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
951 arguments if they are comparable.
953 2018-04-18 Martin Liska <mliska@suse.cz>
956 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
959 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
962 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
965 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
966 ENDBR after calling __morestack.
968 2018-04-18 David Malcolm <dmalcolm@redhat.com>
971 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
972 by using gcc_base_ver to generate a gcc_driver_version, and use
973 it when generating GCC_DRIVER_NAME.
974 * configure: Regenerate.
976 2018-04-18 Jakub Jelinek <jakub@redhat.com>
979 * config.gcc: Obsolete powerpc*-*-*spe*.
981 2018-04-17 Jakub Jelinek <jakub@redhat.com>
984 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
985 (stabstr_D): Change type of unum from unsigned int to
986 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
989 2018-04-17 Jim Wilson <jimw@sifive.com>
992 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
993 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
994 Set arg_pointer_offset after using pretend_args_size.
996 2018-04-17 Jakub Jelinek <jakub@redhat.com>
998 PR rtl-optimization/85431
999 * dse.c (record_store): Ignore zero width stores.
1002 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1003 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1004 __builtin_stack_restore rather than after it.
1005 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1006 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1007 argument instead of virtual_dynamic_stack_rtx.
1009 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1011 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1013 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1014 Add note to error message to explain internal mapping of overloaded
1015 built-in function name to non-overloaded built-in function name.
1016 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1019 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1022 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1023 where the inputs overlap with the output.
1025 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1027 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1028 (=v, v) alternative and explicit "memory" attribute.
1029 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1030 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1032 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1033 "sselog1" type instead of "sselog".
1034 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1035 "sselog". Remove explicit "memory" attribute.
1036 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1037 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1039 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1040 "sselog1" type instead of "sselog". Remove explicit "memory"
1042 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1043 use "sselog1" type instead of "sselog". Remove explicit "memory"
1045 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1046 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1048 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1049 "sselog1" type instead of "sselog". Remove explicit "memory"
1051 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1052 use "sselog1" type instead of "sselog". Remove explicit "memory"
1056 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1059 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1062 2018-04-17 Martin Jambor <mjambor@suse.cz>
1065 * ipa-cp.c (create_specialized_node): Call
1066 expand_all_artificial_thunks if necessary.
1068 2018-04-17 Martin Liska <mliska@suse.cz>
1071 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1072 in message, remote space in between '_G' and '('.
1074 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1077 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1078 avx512f_vmcmp<mode>3<round_saeonly_name>,
1079 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1080 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1081 avx512f_rndscale<mode><round_saeonly_name>,
1082 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1083 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1084 Use %<iptr>2 instead of %2 for -masm=intel.
1085 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1086 avx512f_vcvttss2usi<round_saeonly_name>,
1087 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1089 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1090 avx512f_vcvttsd2usi<round_saeonly_name>,
1091 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1092 Use %q1 instead of %1 for -masm=intel.
1093 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1094 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1095 of %3 for -masm=intel.
1096 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1098 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1100 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1102 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1103 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1105 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1107 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1108 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1109 %g1 and one with %0 and %1.
1110 (avx512er_vmrcp28<mode><round_saeonly_name>,
1111 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1113 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1114 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1115 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1116 of %0 and %{%4%} for -masm=intel.
1117 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1118 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1119 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1120 order of %0 and %{%5%}%{z%} for -masm=intel.
1122 2018-04-17 Jan Hubicka <jh@suse.cz>
1125 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1127 2018-04-17 Martin Liska <mliska@suse.cz>
1130 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1131 for target_clone error message. Make default implementation
1132 clone to be a local declaration.
1133 (separate_attrs): Add new argument and check for an empty
1135 (expand_target_clones): Handle it.
1136 (ipa_target_clone): Make redirection just for target_clones
1139 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1140 Tom de Vries <tom@codesourcery.com>
1143 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1144 tiled basic blocks without omp continue statements.
1146 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1149 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1150 vec_extract expression as having side effects to make sure it gets
1153 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1156 * config/i386/i386.c (get_builtin_code_for_version): Check
1159 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1162 * gcc/config.gcc: Support "skylake".
1163 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1165 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1166 (processor_target_table): Add "skylake".
1167 (ix86_option_override_internal): Add "skylake".
1168 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1169 PROCESSOR_CANNONLAKE.
1170 (get_builtin_code_for_version): Fix priority for
1171 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1172 PROCESSOR_SKYLAKE-AVX512.
1173 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1174 (processor_type): Add PROCESSOR_SKYLAKE.
1176 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1177 Jason Merrill <jason@redhat.com>
1180 * convert.c (convert_to_integer_1): Use direct recursion for
1181 enumeral types and types with a precision less than the number
1182 of bits in their mode.
1184 2018-04-16 Julia Koval <julia.koval@intel.com>
1187 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1188 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1190 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1193 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1194 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1195 and -mno-direct-move.
1197 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1200 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1201 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1202 Ensure negative shifts result in {0}.
1204 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1206 PR rtl-optimization/79916
1207 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1208 regs (if any) to define how to gnerate SD moves when LRA is in
1211 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1213 PR rtl-optimization/85393
1214 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1215 * except.c (expand_dw2_landing_pad_for_region): Make static.
1216 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1217 a label and unconditional jump to old_bb, rather than
1218 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1221 PR rtl-optimization/85376
1222 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1223 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1224 instead of a specific value.
1226 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1227 Bin Cheng <bin.cheng@arm.com>
1229 PR tree-optimization/82965
1230 PR tree-optimization/83991
1231 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1232 by_profile_only parameter.
1233 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1234 information if the loop was predicted to iterate too many times.
1235 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1237 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1240 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1243 2018-04-13 Martin Liska <mliska@suse.cz>
1244 Jakub Jelinek <jakub@redhat.com>
1247 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1248 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1249 * builtins.c (expand_builtin_memory_copy_args): Use
1250 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1251 handle dest_addr == pc_rtx.
1253 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1256 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1257 asked to not generate direct moves.
1258 (fix_trunc<mode>si2_stfiwx): Similar.
1259 (fix_trunc<mode>si2_internal): Similar.
1261 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1264 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1265 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1266 lookup if dest in some wider mode is known to be const0_rtx and
1267 if so, record permanent equivalence for it to be ZERO_EXTEND of
1268 the narrower mode destination.
1270 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1272 * lto-streamer-out.c (output_function): Revert 259346.
1273 * omp-expand.c (expand_oacc_for): Likewise.
1275 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1277 PR rtl-optimization/85354
1278 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1279 * sel-sched.c (sel_global_init): ... here.
1281 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1284 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1285 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1286 mode for PE-COFF targets.
1287 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1288 (i386_pe_asm_lto_end): Likewise.
1289 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1290 (TARGET_ASM_LTO_END): Likewise.
1291 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1292 (i386_pe_asm_lto_start): New function.
1293 (i386_pe_asm_lto_end): Likewise.
1295 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1296 Richard Biener <rguenther@suse.de>
1299 * lto-streamer-out.c (output_function): Fix CFG loop state before
1301 * omp-expand.c (expand_oacc_for): Handle calls to internal
1302 functions like regular functions.
1304 2018-04-12 Richard Biener <rguenther@suse.de>
1307 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1308 for the early LTO debug to properly generate references to it
1309 during DIE emission. Do not re-use that for the skeleton for
1311 (dwarf2out_early_finish): Likewise.
1313 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1316 * config/i386/sse.md
1317 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1318 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1319 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1320 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1321 and output is a reg, avoid creating invalid lowpart subreg, but
1322 instead split into a 512-bit move. Don't split if not AVX512VL,
1323 input is xmm16+ reg and output is a mem.
1324 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1325 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1326 xmm16+ reg and output is a mem.
1328 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1330 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1331 also for flag_dwarf2_cfi_asm.
1333 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1335 PR rtl-optimization/85342
1336 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1337 a bool scalar var inside of the loop instead. Don't try to update
1338 recog_data.operand after failed apply_change_group.
1340 2018-04-12 Tom de Vries <tom@codesourcery.com>
1343 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1344 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1345 array with flexible array member as array without given dimension.
1346 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1347 argument for undefined param to true.
1349 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1352 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1353 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1354 from PowerPC section.
1355 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1356 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1357 help text that is too long.
1358 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1359 help text that is too long.
1360 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1361 help text that is too long.
1363 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1365 * config/alpha/alpha.md (stack_probe_internal): Rename
1366 from "probe_stack". Update all callers.
1368 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1370 PR rtl-optimization/84566
1371 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1372 sched_macro_fuse_insns.
1374 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1377 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1378 (compute_block_dependences): ... from here.
1380 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1382 PR tree-optimization/85331
1383 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1384 from int to HOST_WIDE_INT.
1386 2018-04-11 Martin Jambor <mjambor@suse.cz>
1389 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1390 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1391 not the same as the source val.
1392 (cgraph_edge_brings_value_p): New parameter.
1393 (gather_edges_for_value): Pass destination value to
1394 cgraph_edge_brings_value_p.
1395 (perhaps_add_new_callers): Likewise.
1396 (get_info_about_necessary_edges): Likewise and exclude values brought
1397 only by self-recursive edges.
1398 (create_specialized_node): Redirect only clones of self-calling edges.
1399 (+self_recursive_pass_through_p): New function.
1400 (find_more_scalar_values_for_callers_subset): Use it.
1401 (find_aggregate_values_for_callers_subset): Likewise.
1402 (known_aggs_to_agg_replacement_list): Removed.
1403 (decide_whether_version_node): Re-calculate known constants for all
1404 remaining context clones.
1406 2018-04-11 Richard Biener <rguenther@suse.de>
1409 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1410 from early DWARF output.
1411 (dwarf2out_early_finish): Output line info unconditionally into
1412 early DWARF and add reference to it.
1414 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1417 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1418 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1419 other than V2DFmode using iptr mode attribute.
1420 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1422 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1424 PR rtl-optimization/84659
1425 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1427 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1430 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1432 (output_loc_list): Pass address of a dummy size variable even in the
1433 locview handling loop.
1434 (index_location_lists): Add comment on why skip_loc_list_entry can't
1437 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1440 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1443 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1446 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1447 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1448 and -mstring-compare-inline-limit.
1450 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1453 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1454 for stack clash protection in a register whenever we need it to be in
1457 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1459 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1460 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1462 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1465 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1467 (mlong-double-): Ditto.
1468 * config/rs6000/sysv4.opt (msdata=): Ditto.
1469 (mtls-size=): Ditto.
1471 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1473 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1474 erroneous entries for
1475 "vector int vec_ldl (int, long int *)", and
1476 "vector unsigned int vec_ldl (int, unsigned long int *)".
1477 Add comments and entries for
1478 "vector bool char vec_ldl (int, bool char *)",
1479 "vector bool short vec_ldl (int, bool short *)",
1480 "vector bool int vec_ldl (int, bool int *)",
1481 "vector bool long long vec_ldl (int, bool long long *)",
1482 "vector pixel vec_ldl (int, pixel *)",
1483 "vector long long vec_ldl (int, long long *)",
1484 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1485 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1486 type tree bool_long_long_type_node and correct definition of
1487 bool_V2DI_type_node to make reference to this new type tree.
1488 (rs6000_mangle_type): Replace erroneous reference to
1489 bool_long_type_node with bool_long_long_type_node.
1490 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1491 comments to emphasize sign distinctions for char and int types and
1492 replace RS6000_BTI_bool_long constant with
1493 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1494 use of RS6000_BTI_pixel.
1495 (bool_long_type_node): Remove this macro definition.
1496 (bool_long_long_type_node): New macro definition
1498 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1500 PR rtl-optimization/85300
1501 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1502 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1503 simplify_unary_operation fails.
1505 2018-04-10 Martin Liska <mliska@suse.cz>
1507 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1508 cgraph_edge and ipa_ref.
1510 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1514 * config/i386/sse.md
1515 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1516 computation of the VEC_MERGE selector from mask.
1517 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1518 Fix decoding of the VEC_MERGE selector into mask.
1520 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1522 PR tree-optimization/85286
1523 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1525 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1527 * final.c (final_1): Set insn_last_address as well as
1528 insn_current_address.
1530 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1533 * explow.c (emit_stack_probe): Call validize_mem on memory location
1534 before passing it to gen_probe_stack. Create address operand and
1535 legitimize it for the probe_stack_address case.
1537 2018-04-09 Jan Hubicka <jh@suse.cz>
1540 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1541 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1542 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1544 (free_lang_data): Rebuild type inheritance graph.
1546 2018-04-09 Martin Sebor <msebor@redhat.com>
1548 * invoke.texi (-finline-small-functions): Mention other optimization
1550 (-findirect-inlining, -fpartial-inlining): Same.
1551 (-finline-functions-called-once): Same.
1552 (-freorder-blocks-and-partition): Same.
1554 2018-04-09 Jan Hubicka <jh@suse.cz>
1557 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1558 jumps; choose last target that matches the criteria (i.e.
1559 no partition changes for non-crossing jumps).
1560 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1561 support for redirecting crossing jumps to non-crossing.
1563 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1565 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1566 also for naked functions.
1568 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1570 * config/arc/arc.md (add_shift): New pattern.
1571 (add_shift2): Likewise.
1572 (sub_shift): Likewise.
1573 (sub_shift_cmp0_noout): Likewise.
1574 (compare_si_ashiftsi): Likewise.
1575 (xbfu_cmp0_noout): New combine pattern.
1576 (xbfu_cmp0"): Likewise.
1577 (movsi_set_cc_insn): Place the predicable variant first.
1578 (commutative_binary_cmp0_noout): Remove clobber.
1579 (commutative_binary_cmp0): New pattern.
1580 (noncommutative_binary_cmp0): Likewise.
1581 (noncommutative_binary_cmp0_noout): Likewise.
1582 (noncommutative_binary_comparison_result_used): Removed.
1583 (rsub_cmp0): New pattern.
1584 (rsub_cmp0_noout): Likewise.
1585 (extzvsi): Changed, keep only meaningful variants.
1586 (SQH, SEZ): New iterators.
1587 (SQH_postfix): New mode attribute.
1588 (SEZ_prefix): New code attribute.
1589 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1590 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1591 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1593 (noncommutative_operator): Check the availability of barrel
1596 2018-04-09 Richard Biener <rguenther@suse.de>
1598 PR tree-optimization/85284
1599 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1600 Only use the niter constraining form of simple_iv when the exit
1603 2018-04-09 Tom de Vries <tom@codesourcery.com>
1606 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1607 (define_expand "*memory_barrier"): New define_expand.
1608 (define_insn "memory_barrier"): New insn.
1610 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1612 PR rtl-optimization/80463
1613 PR rtl-optimization/83972
1614 PR rtl-optimization/83480
1616 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1617 correct producer for the insn.
1618 (tidy_control_flow): Fixup seqnos in case of debug insns.
1620 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1622 PR rtl-optimization/83913
1624 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1625 different sched-times when merging exprs.
1627 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1629 PR rtl-optimization/83962
1631 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1632 tidy_fallthru_edge and tidy_control_flow.
1634 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1636 PR rtl-optimization/83530
1638 * sel-sched.c (force_next_insn): New global variable.
1639 (remove_insn_for_debug): When force_next_insn is true, also leave only
1640 next insn in the ready list.
1641 (sel_sched_region): When the region wasn't scheduled, make another pass
1642 over it with force_next_insn set to 1.
1644 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1646 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1648 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1649 for interrupt control.
1650 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1651 functions for interrupt control.
1652 * config/nds32/nds32-intrinsic.md: Likewise.
1653 * config/nds32/nds32_intrinsic.h: Likewise.
1654 * config/nds32/nds32.h (nds32_builtins): Likewise.
1656 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1658 * config/nds32/nds32.c (nds32_init_machine_status,
1659 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1660 strict_aligned_p field.
1661 (nds32_expand_to_rtl_hook): New function.
1662 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1663 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1665 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1666 Chung-Ju Wu <jasonwucj@gmail.com>
1668 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1669 * config/nds32/nds32-n7.md: New file.
1670 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1671 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1673 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1674 * config/nds32/nds32.md (pipeline_model): Add n7.
1675 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1676 * config/nds32/pipelines.md: Include n7 settings.
1678 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1679 Chung-Ju Wu <jasonwucj@gmail.com>
1681 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1682 * config/nds32/nds32-e8.md: New file.
1683 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1684 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1686 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1687 * config/nds32/nds32.md (pipeline_model): Add e8.
1688 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1689 * config/nds32/pipelines.md: Include e8 settings.
1691 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1692 Chung-Ju Wu <jasonwucj@gmail.com>
1694 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1695 * config/nds32/nds32-n8.md: New file.
1696 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1697 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1699 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1700 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1701 * config/nds32/nds32.md (pipeline_model): Add n8.
1702 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1703 * config/nds32/pipelines.md: Include n8 settings.
1705 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1706 Chung-Ju Wu <jasonwucj@gmail.com>
1708 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1709 * config/nds32/nds32-n9-2r1w.md: New file.
1710 * config/nds32/nds32-n9-3r2w.md: New file.
1711 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1712 nds32_register_ports): New or modify for cpu n9.
1713 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1715 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1716 * config/nds32/nds32-utils.c: New file.
1717 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1718 TARGET_MUL_SLOW): Define.
1719 * config/nds32/nds32.md (pipeline_model): New attribute.
1720 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1721 New options that support cpu n9.
1722 * config/nds32/pipelines.md: Include n9 settings.
1723 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1725 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1727 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1728 information if necessary.
1729 (output_cond_branch_compare_zero): Likewise.
1730 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1731 (nds32_target_alignment): Refine for alignment.
1732 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1733 (FUNCTION_BOUNDARY): Modify.
1734 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1736 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1738 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1740 * config/nds32/constants.md (unspec_volatile_element): Add values for
1741 TLB operation and data prefetch.
1742 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1743 functions for TLB operation and data prefetch.
1744 * config/nds32/nds32-intrinsic.md: Likewise.
1745 * config/nds32/nds32_intrinsic.h: Likewise.
1746 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1747 (nds32_print_operand): Likewise.
1748 * config/nds32/nds32.h (nds32_builtins): Likewise.
1750 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1751 Andrew Pinski <pinsika@gcc.gnu.org>
1754 * match.pd: Use constant_boolean_node of correct type instead of
1755 boolean_true_node or boolean_false_node for simplifying
1756 pointer comparisons to zero.
1758 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1760 PR tree-optimization/80021
1761 * tree.c (verify_type_variant): Make error call in verify_variant_match
1762 translatable and remove final full stop.
1764 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1766 * config/nds32/constants.md (unspec_volatile_element): Add
1767 UNSPEC_VOLATILE_EH_RETURN.
1768 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1769 nds32_output_stack_pop): Support dwarf exception handling process.
1770 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1771 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1772 exception handling process.
1773 (nds32_compute_stack_frame): Likewise.
1774 (nds32_return_addr_rtx): Likewise.
1775 (nds32_initial_elimination_offset): Likewise.
1776 (nds32_expand_prologue): Likewise.
1777 (nds32_expand_epilogue): Likewise.
1778 (nds32_dynamic_chain_address): New function.
1779 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1781 (DYNAMIC_CHAIN_ADDRESS): Define.
1782 (EH_RETURN_DATA_REGNO): Define.
1783 (EH_RETURN_STACKADJ_RTX): Define.
1784 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1785 patterns for dwarf exception handling.
1787 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1789 * config/nds32/nds32.h: Clean up obsolete macros.
1791 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1793 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1794 Add enum values for particular instructions.
1795 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1796 particular intrinsic functions.
1797 * config/nds32/nds32-intrinsic.md: Likewise.
1798 * config/nds32/nds32_intrinsic.h: Likewise.
1799 * config/nds32/nds32.h (nds32_builtins): Likewise.
1800 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1801 (btst, ave): New patterns for particular instructions.
1803 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1805 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1806 Add enum values for atomic load/store and memory sync.
1807 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1809 * config/nds32/nds32-intrinsic.md: Likewise.
1810 * config/nds32/nds32_intrinsic.h: Likewise.
1811 * config/nds32/nds32.h (nds32_builtins): Likewise.
1813 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1815 PR tree-optimization/85257
1816 * fold-const.c (native_encode_vector): If not all elts could fit
1817 and off is -1, return 0 rather than offset.
1818 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1819 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1820 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1821 adjust buffer in native_interpret_expr call.
1823 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1825 * config/nds32/constants.md (unspec_volatile_element): Add cache
1826 control enum values.
1827 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1828 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1829 * config/nds32/nds32.c (nds32_cctl_names): New.
1830 (nds32_print_operand): Handle cache control register names.
1831 * config/nds32/nds32.h (nds32_builtins): New enum values.
1832 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1834 * config/nds32/nds32.md (type): Add mmu.
1835 * config/nds32/pipelines.md (simple_insn): Add mmu.
1837 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1839 * config/nds32/nds32.md (type): Remove call.
1840 * config/nds32/pipelines.md (simple_insn): Likewise.
1842 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1844 * config/nds32/constants.md (unspec_volatile_element): Add
1845 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1846 UNSPEC_VOLATILE_FMFCFG.
1847 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1848 description for fmfcfg and fmfcsr.
1849 (bdesc_1arg): Add fmtcsr.
1850 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1851 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1852 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1853 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1854 unspec_fmfcfg): New patterns.
1855 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1856 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1857 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1858 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1859 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1860 __nds32__fmfcfg): Define.
1862 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1864 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1865 intrinsic register names.
1866 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1867 intrinsic register enum values and macros.
1869 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1871 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1872 for load/store addressing form.
1873 (nds32_print_operand_address): Likewise.
1875 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1878 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1879 based on LABEL_REF. Remove useless assertion.
1880 (pic_address_needs_scratch): Fix formatting.
1881 (sparc_legitimize_pic_address): Minor tweaks.
1882 (sparc_delegitimize_address): Adjust assertion accordingly.
1883 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1884 into symbolic_operand.
1885 (movsi_high_pic_label_ref): Likewise.
1886 (movsi_lo_sum_pic_label_ref): Likewise.
1887 (movdi_pic_label_ref): Likewise.
1888 (movdi_high_pic_label_ref): Likewise.
1889 (movdi_lo_sum_pic_label_ref): Likewise.
1891 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1893 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1894 custom LIB_SPEC setup.
1896 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1897 Kito Cheng <kito.cheng@gmail.com>
1899 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1900 * config/riscv/freebsd.h: New.
1902 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1904 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1905 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1908 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1909 Kito Cheng <kito.cheng@gmail.com>
1911 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1912 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1913 * config/nds32/nds32-protos.h (nds32_output_call,
1914 nds32_output_return): Declare.
1915 * config/nds32/nds32.md: Refine all the call and return patterns.
1917 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1920 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1921 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1923 PR rtl-optimization/84872
1924 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1925 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1928 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1930 * expr.c (copy_blkmode_to_reg): Revert 254862.
1931 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1933 2018-04-06 Richard Biener <rguenther@suse.de>
1936 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1937 after seeing a component reference with an adjacent field. Treat
1938 refs to arrays at struct end of external decls similar to
1939 refs to unconstrained commons.
1941 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1944 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1945 look through SAVE_EXPRs with non-side-effects argument. Adjust
1947 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1950 2018-04-06 Richard Biener <rguenther@suse.de>
1953 * alias.c (find_base_term): New wrapper around find_base_term
1954 unwinding CSELIB_VAL_PTR changes.
1955 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1958 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1960 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1962 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1963 constant definitions.
1964 ("nop"): lr 0,0 -> nopr r0
1965 ("nop_lr0", "nop_lr1"): New insn definitions.
1967 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1969 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1970 NDS32_V3PUSH_AVAILABLE_P macro.
1972 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1973 Chung-Ju Wu <jasonwucj@gmail.com>
1975 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1976 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1977 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1978 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1979 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1980 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1981 * config/nds32/constraints.md: New constraints and checking for hard
1982 float configuration.
1983 * config/nds32/iterators.md: New mode iterator and attribute for hard
1984 float configuration.
1985 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1987 * config/nds32/nds32-fpu.md: New file.
1988 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1989 deal with hard float code generation.
1990 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1992 (abi_type, float_reg_number): New enum type.
1993 * config/nds32/nds32-predicates.c: New predicates for hard float.
1994 * config/nds32/nds32-protos.h: Declare functions for hard float.
1995 * config/nds32/nds32.c: Implementation for hard float configuration.
1996 * config/nds32/nds32.h: Definitions for hard float configuration.
1997 * config/nds32/nds32.md: Include hard float machine description and
1998 modify patterns for hard float configuration.
1999 * config/nds32/nds32.opt: New options for hard float configuration.
2000 * config/nds32/predicates.md: New predicates for hard float
2003 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2005 * common/config/nds32/nds32-common.c
2006 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2008 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2011 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2012 CONSTRUCTOR_ELT (ctor, ...)->value.
2014 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2017 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2019 2018-04-05 Tom de Vries <tom@codesourcery.com>
2022 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2025 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2026 Kito Cheng <kito.cheng@gmail.com>
2028 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2029 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2030 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2031 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2033 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2034 Kito Cheng <kito.cheng@gmail.com>
2036 * config/nds32/constraints.md (Ufe): New memory constraint.
2037 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2038 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2039 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2041 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2042 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2044 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2046 * config/nds32/nds32.md: Use optimize_size in the condition for
2047 alu-shift instructions.
2049 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2051 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2053 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2055 * config/nds32/nds32.md (negsi2): Refine pattern.
2057 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2058 Chung-Ju Wu <jasonwucj@gmail.com>
2060 * config/nds32/iterators.md (shift_rotate): New code iterator.
2061 (shift): New code attribute.
2062 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2063 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2064 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2065 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2066 bit-wise operations.
2067 (andsi3, *andsi3): Ditto.
2068 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2069 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2070 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2071 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2072 nds32_ior_operand, nds32_xor_operand): New predicates.
2074 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2076 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2077 (addsi3, subsi3): ... this.
2079 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2081 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2083 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2085 * config/nds32/nds32.md: Adjust indention.
2087 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2089 * config/nds32/nds32.md (feature): New attribute.
2091 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2093 * config/nds32/nds32.md (subtype): New attribute.
2095 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2098 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2099 expansion to perform a bitwise AND of the argument followed by a
2100 boolean negation of the result.
2102 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2104 PR rtl-optimization/84878
2105 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2106 the basic block. Assert the use reference is not artificial and that
2107 it has an associated insn.
2109 2018-04-04 Michael Matz <matz@suse.de>
2111 * builtins.c (compute_objsize): Pass correct operand
2112 to array_at_struct_end_p.
2114 2018-04-04 Richard Biener <rguenther@suse.de>
2117 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2118 from contexts for DINFO_LEVEL_TERSE and below.
2120 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2122 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2123 resiter_operand condition.
2124 * config/nds32/nds32.md (*move<mode>): Ditto.
2126 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2127 Monk Chiang <sh.chiang04@gmail.com>
2129 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2131 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2133 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2135 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2136 Kito Cheng <kito.cheng@gmail.com>
2138 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2139 nds32_cond_code_str, output_cond_branch,
2140 output_cond_branch_compare_zero, nds32_expand_cbranch,
2141 nds32_expand_cstore, nds32_expand_movcc,
2142 nds32_output_cbranchsi4_equality_zero,
2143 nds32_output_cbranchsi4_equality_reg,
2144 nds32_output_cbranchsi4_equality_reg_or_const_int,
2145 nds32_output_cbranchsi4_greater_less_zero: New functions.
2146 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2147 nds32_expand_cstore, nds32_expand_movcc,
2148 nds32_output_cbranchsi4_equality_zero,
2149 nds32_output_cbranchsi4_equality_reg,
2150 nds32_output_cbranchsi4_equality_reg_or_const_int,
2151 nds32_output_cbranchsi4_greater_less_zero): Declare.
2152 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2153 nds32_rimm11s_operand): New predicates.
2154 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2155 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2158 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2160 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2161 * config/nds32/nds32.md: Ditto.
2162 * config/nds32/pipelines.md: Ditto.
2164 2018-04-04 Richard Biener <rguenther@suse.de>
2166 PR tree-optimization/85168
2167 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2168 propagating abnormals.
2170 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2172 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2174 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2175 Kito Cheng <kito.cheng@gmail.com>
2177 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2178 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2179 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2180 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2181 * config/nds32/nds32.md (sibcall_internal): New.
2182 (sibcall_register): Remove.
2183 (sibcall_immediate): Remove.
2184 (sibcall_value_internal): New.
2185 (sibcall_value_register): Remove.
2186 (sibcall_value_immediate): Remove.
2187 * config/nds32/predicates.md (nds32_general_register_operand): New.
2188 (nds32_call_address_operand): New.
2190 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2192 PR rtl-optimization/85167
2193 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2194 bb_defs if *split_p, instead preinitialize it to NULL.
2196 PR tree-optimization/85156
2197 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2198 evaluating the argument multiple times.
2200 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2202 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2204 (_mm_cvtpd_ps): Likewise.
2205 (_mm_cvttpd_epi32): Likewise.
2206 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2207 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2208 vector, pixel, and bool following altivec.h include.
2210 2018-04-03 Martin Sebor <msebor@redhat.com>
2212 * doc/extend.texi (Common Function Attributes): Clarify.
2213 (const attribute): Likewise.
2214 (pure attribute): Likewise.
2216 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2219 * config/i386/i386.c (ix86_expand_vector_set): Use
2220 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2222 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2224 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2225 instructions when changing rounding bits to preserve precision bits
2226 in the x87 control word.
2228 2018-04-03 Martin Liska <mliska@suse.cz>
2230 PR tree-optimization/82491
2231 * rtl.h (strip_offset_and_add): Replace += suboffset with
2232 poly_uint64 () + suboffset.
2234 2018-03-29 Martin Liska <mliska@suse.cz>
2235 Martin Jambor <mjambor@suse.cz>
2238 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2239 param_type is not an integral or pointer type.
2241 2018-04-03 Richard Biener <rguenther@suse.de>
2243 * sese.h (recompute_all_dominators): Remove.
2245 2018-04-02 Martin Sebor <msebor@redhat.com>
2247 * doc/invoke.texi (-Wrestrict): Fix typos.
2249 2018-04-02 Jim Wilson <jimw@sifive.com>
2251 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2252 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2253 (<optab>di3, <optab>si3_extend): Likewise.
2254 (<optab>si3_mask, <optab>si3_mask_1): New.
2255 (<optab>di3_mask, <optab>di3_mask_1): New.
2256 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2257 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2258 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2260 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2262 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2265 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2267 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2268 (nds32_canonicalize_comparison): New function.
2270 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2271 Kito Cheng <kito.cheng@gmail.com>
2272 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2274 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2275 * config/nds32/constants.md (unspec_volatile_element): Add
2276 UNSPEC_VOLATILE_RELAX_GROUP.
2277 * config/nds32/nds32-relax-opt.c: New file.
2278 * config/nds32/nds32-predicates.c
2279 (nds32_symbol_load_store_p): New function.
2280 * config/nds32/nds32-protos.h
2281 (nds32_symbol_load_store_p): Declare function.
2282 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2283 * config/nds32/nds32.c
2284 (nds32_register_pass): New function to register pass.
2285 (nds32_register_passes): New function to register passes.
2286 * config/nds32/nds32.md (relax_group): New pattern.
2287 * config/nds32/nds32.opt (mrelax-hint): New option.
2288 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2290 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2292 * config/nds32/t-nds32: Modify files dependency.
2294 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2296 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2297 (PROFILE_HOOK): Define its implementation.
2299 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2301 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2302 type and 32-bit size.
2304 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2307 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2308 (V_128_256): New mode iterator.
2309 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2310 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2311 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2313 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2314 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2316 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2319 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2320 NaN inputs correctly.
2322 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2325 * config/rs6000/vsx.md (??r): New mode attribute.
2326 (*vsx_mov<mode>_64bit): Use it.
2327 (*vsx_mov<mode>_32bit): Likewise.
2329 2018-03-30 Martin Sebor <msebor@redhat.com>
2331 PR tree-optimization/84818
2332 * builtins.c (check_access): Use warning_n.
2334 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2337 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2339 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2342 2018-03-30 Julia Koval <julia.koval@intel.com>
2345 * x86-tune.def (movx, partial_reg_dependency): Enable for
2348 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2351 * lra-constraints.c (process_alt_operands): Move setting
2352 this_alternative_matches below.
2354 2018-03-29 Martin Liska <mliska@suse.cz>
2357 * doc/invoke.texi: Document how LTO works with debug info.
2358 Describe auto-load support of binutils. Mention 'x86-64'
2359 as valid option value of -march option.
2361 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2363 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2366 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2367 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2368 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2371 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2374 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2375 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2376 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2377 for RS6000_BTM_POWERPC64.
2378 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2379 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2380 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2385 2018-03-28 Carl Love <cel@us.ibm.com>
2388 2017-09-27 Carl Love <cel@us.ibm.com>
2390 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2391 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2392 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2395 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2397 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2398 instead of __vector bool.
2399 (_mm_max_pu8): Likewise.
2400 (_mm_min_pi16): Likewise.
2402 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2405 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2406 (DIVWEUO): Likewise.
2408 (DIVDEUO): Likewise.
2409 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2410 DIVWEUO and DIVDEUO.
2411 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2412 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2413 (div_extend): Likewise.
2414 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2416 (__builtin_divweuo): Likewise.
2417 (__builtin_divdeo): Likewise.
2418 (__builtin_divdeuo): Likewise.
2420 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2423 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2424 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2426 PR tree-optimization/82004
2427 * gimple-match-head.c (optimize_pow_to_exp): New function.
2428 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2429 Don't fold to exp if optimize_pow_to_exp is false.
2431 2018-03-28 Martin Liska <mliska@suse.cz>
2434 * calls.c (initialize_argument_information): Fix trailing space.
2435 * common.opt: Fix typo and provide better explanation for
2436 -fsanitize-coverage option.
2437 * config/i386/i386.opt: Fix typo.
2439 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2440 Martin Liska <mliska@suse.cz>
2443 * gimplify.c (asan_poison_variable): Don't do the check for
2444 gimplify_omp_ctxp here.
2445 (gimplify_decl_expr): Do it here.
2446 (gimplify_target_expr): Likewise.
2448 2018-03-28 Martin Liska <mliska@suse.cz>
2451 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2452 chkp_type_bounds_count if MPX is not enabled.
2454 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2456 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2458 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2461 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2462 function to create the function decl for complex long double
2463 multiply and divide for -mabi=ieeelongdouble.
2464 (init_float128_ieee): Call it.
2466 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2469 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2470 -fcf-protection=branch -mibt.
2471 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2473 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2476 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2478 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2481 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2482 extern array declarations.
2484 2018-03-27 Richard Biener <rguenther@suse.de>
2487 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2488 explicit single_use checks.
2490 2018-03-27 Richard Biener <rguenther@suse.de>
2492 PR tree-optimization/85082
2493 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2496 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2498 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2499 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2500 Turn on fasynchronous-unwind-tables and funwind-tables.
2502 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2505 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2506 (*bmi_blsr_<mode>_ccz): Ditto.
2508 2018-03-26 Tom de Vries <tom@codesourcery.com>
2510 PR tree-optimization/85063
2511 * omp-general.c (offloading_function_p): New function. Factor out
2513 * omp-offload.c (pass_omp_target_link::gate): ... here.
2514 * omp-general.h (offloading_function_p): Declare.
2515 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2516 with attribute omp declare target for offloading functions.
2518 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2520 PR tree-optimization/84005
2521 * tree-data-ref.h (get_base_for_alignment): Declare.
2522 * tree-data-ref.c (get_base_for_alignment_1): New function.
2523 (get_base_for_alignment): Likewise.
2524 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2525 get_base_for_alignment to find a suitable base object, instead
2526 of always using drb->base_address.
2528 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2531 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2532 known size by default.
2534 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2537 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2538 and non BLKmode operands.
2540 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2543 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2544 Clean up attributes.
2546 2018-03-23 Richard Biener <rguenther@suse.de>
2549 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2550 we are going to emit early debug for LTO.
2552 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2555 * function.c (match_asm_constraints_1): Don't optimize if input
2556 doesn't satisfy general_operand predicate for output's mode.
2559 * alias.c (write_dependence_p): Don't require for x_canonicalized
2560 non-VOIDmode if x has VOIDmode.
2563 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2564 just don't try to optimize it rather than assert it never happens.
2566 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2568 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2569 macro expansions for definition of ST_INTERNAL_<mode> and
2570 LD_INTERNAL_<mode> builtins.
2571 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2573 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2575 (altivec_expand_st_builtin): Likewise.
2576 (altivec_expand_builtin): Remove calls to deleted functions.
2577 (rs6000_address_for_altivec): Delete this function.
2578 * config/rs6000/vector.md: Remove expands for
2579 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2581 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2584 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2585 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2586 re-computing once computed.
2587 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2588 (arm_init_machine_status): Initialize
2589 machine->static_chain_stack_bytes.
2591 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2594 * doc/extend.texi: Add four new prototypes for vec_ld.
2595 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2596 definitions for more logical presentation.
2597 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2598 entries for V1TI variants of __builtin_altivec_ld builtin.
2599 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2600 handling of V1TI variant of LVX icode pattern.
2601 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2602 (rs6000_gimple_fold_builtin): Likewise.
2603 (altivec_init_builtins): Add code to define
2604 __builtin_altivec_lvx_v1ti function.
2606 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2609 * function.c (match_asm_constraints_1): Don't do the optimization
2610 if input isn't a REG, SUBREG, MEM or constant.
2612 2018-03-22 Tom de Vries <tom@codesourcery.com>
2614 PR tree-optimization/84956
2615 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2616 bb_has_abnormal_pred.
2618 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2621 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2622 DECL_INITIAL (decl) to decl at the end.
2623 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2626 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2628 * doc/extend.texi (__builtin_tgmath): Document when complex
2629 integer types are treated as _Complex _Float64.
2631 2018-03-21 Tom de Vries <tom@codesourcery.com>
2633 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2635 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2637 PR tree-optimization/84960
2638 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2639 if it is ENTRY block, move them into single succ of ENTRY in that case.
2641 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2643 PR tree-optimization/84811
2644 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2645 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2646 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2647 (known_subrange_p): Update accordingly. Cast each value involved
2648 in the size comparison, rather than casting the result of the
2651 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2653 PR tree-optimization/84982
2654 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2655 by flipping the least significant bit rather than all bits from
2656 bitpos to bitpos + bitsize - 1.
2658 2018-03-21 Nathan Sidwell <nathan@acm.org>
2660 * doc/extend.texi (Deprecated Features): Remove mention of
2661 long-deleted deprecations.
2663 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2666 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2667 * configure: Regenerate.
2669 2018-03-21 Tom de Vries <tom@codesourcery.com>
2671 PR tree-optimization/83126
2672 * tree-parloops.c (num_phis): New function.
2673 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2675 2018-03-21 Nathan Sidwell <nathan@acm.org>
2677 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2678 mention anon-struct/union members and trailing attributes.
2680 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2682 PR tree-optimization/84969
2683 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2684 builtin memset partitions if they set different rhs values.
2686 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2688 PR rtl-optimization/84989
2689 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2690 VEC_DUPLICATE with scalar result mode.
2692 2018-03-21 Martin Liska <mliska@suse.cz>
2695 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2696 not intended return statement.
2698 2018-03-21 Martin Liska <mliska@suse.cz>
2701 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2702 (chkp_find_bound_slots_1): Limit number of iterations.
2704 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2707 * Minor grammar fixes for x86 options.
2709 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2712 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2713 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2715 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2718 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2719 (altivec_lvx_<mode>_1op): Likewise.
2720 (altivec_stvx_<mode>_2op): Likewise.
2721 (altivec_stvx_<mode>_1op): Likewise.
2722 (altivec_lvx_<VM2:mode>): New define_expand.
2723 (altivec_stvx_<VM2:mode>): Likewise.
2724 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2725 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2726 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2727 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2728 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2729 (rs6000_gen_lvx): Likewise.
2730 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2731 (altivec_expand_stv_builtin): Likewise.
2732 (altivec_expand_builtin): Likewise.
2733 * config/rs6000/vector.md: Likewise.
2735 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2738 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2741 2018-03-20 Richard Biener <rguenther@suse.de>
2744 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2745 sign-conversions as zero, fall back to standard scalar_stmt
2748 2018-03-20 Martin Liska <mliska@suse.cz>
2751 * predict.c (rebuild_frequencies): Handle case when we have
2752 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2754 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2757 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2758 flag_section_anchors.
2759 * varasm.c (use_blocks_for_decl_p): Remove hack for
2760 dw2_force_const_mem.
2763 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2765 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2766 be created, use lowpart_subreg of operands[0] rather than operands[0]
2768 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2769 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2770 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2771 and n constraint instead of aarch64_shift_imm_di and Usd.
2772 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2773 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2775 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2778 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2779 to favor GPR over NEON registers.
2780 (<shift>di3_neon): Likewise.
2782 2018-03-20 Tom de Vries <tom@codesourcery.com>
2785 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2786 (nvptx_process_pars): Emit bar.sync asap and alap.
2788 2018-03-20 Tom de Vries <tom@codesourcery.com>
2791 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2792 seen_label if seen_label is already set.
2794 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2797 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2798 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2799 Use 1U instead of 1. Formatting fixes.
2802 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2803 instead of TREE_TYPE (s1) for the return value.
2805 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2807 PR tree-optimization/84946
2808 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2809 bitsize + bitsize in poly_uint64 rather than poly_int64.
2812 * dwarf2asm.c: Include fold-const.c.
2813 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2814 of decl rather than decl itself.
2816 PR rtl-optimization/84643
2817 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2819 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2822 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2823 calling assemble_variable.
2825 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2828 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2829 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2831 2018-03-19 Jim Wilson <jimw@sifive.com>
2834 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2835 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2836 (riscv_first_stack_step): Likewise.
2837 (riscv_option_override): Use STACK_BOUNDARY instead of
2839 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2841 (BIGGEST_ALIGNMENT): Set to 128.
2842 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2843 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2846 2018-03-19 Richard Biener <rguenther@suse.de>
2848 PR tree-optimization/84933
2849 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2850 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2852 2018-03-19 Richard Biener <rguenther@suse.de>
2854 PR tree-optimization/84859
2855 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2856 (cond_if_else_store_replacement): Perform sinking operation on
2857 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2858 Generalize what a BB with a single eligible store is.
2860 2018-03-19 Richard Biener <rguenther@suse.de>
2862 PR tree-optimization/84929
2863 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2864 chrec_is_positive against non-chrec arg.
2866 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2869 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2871 2018-03-18 Martin Liska <mliska@suse.cz>
2873 PR rtl-optimization/84635
2874 * regrename.c (build_def_use): Use matches_mode only when
2877 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2879 PR tree-optimization/84913
2880 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2881 vectorize chains of COND_EXPRs.
2883 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2885 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2887 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2889 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2891 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2893 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2895 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2896 Kito Cheng <kito.cheng@gmail.com>
2898 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2899 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2900 (nds32_adjust_reg_alloc_order): New function.
2901 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2903 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2905 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2906 nds32_print_operand, nds32_print_operand_address): Use
2907 HOST_WIDE_INT_PRINT_DEC instead.
2909 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2911 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2913 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2916 * config/i386/i386.c (initial_ix86_tune_features,
2917 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2919 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2920 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2921 rather than 1u << ix86_tune. Formatting fix.
2922 (ix86_option_override_internal): Change ix86_arch_mask from
2923 unsigned int to unsigned HOST_WIDE_INT, initialize to
2924 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2925 (ix86_function_specific_restore): Likewise.
2927 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2930 * postreload.c (reload_combine_recognize_pattern): Perform
2931 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2932 truncate_int_for_mode the result for the destination's mode.
2935 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2936 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2938 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2941 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2944 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2945 regno_allocno_class_array and sorted_pseudos.
2946 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2947 insns where regno is used.
2949 2018-03-16 Martin Liska <mliska@suse.cz>
2952 * multiple_target.c (create_dispatcher_calls): Redirect
2953 reference in the symbol table.
2955 2018-03-16 Martin Liska <mliska@suse.cz>
2958 * multiple_target.c (create_dispatcher_calls): Redirect also
2961 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2965 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2966 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2967 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2969 2018-03-16 Julia Koval <julia.koval@intel.com>
2971 * doc/invoke.texi (Skylake Server): Add CLWB.
2972 Cannonlake): Remove CLWB.
2974 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2976 PR tree-optimization/84841
2977 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2979 (FLOAT_ONE_CONST_TYPE): Define.
2980 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2981 (sort_by_operand_rank): Put entries with higher constant_type last
2982 rather than first to match comments.
2984 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2986 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2989 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2992 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2993 check and use address of target always.
2995 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2998 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2999 (indirect_thunk_bnd_needed): Likewise.
3000 (indirect_thunks_used): Likewise.
3001 (indirect_thunks_bnd_used): Likewise.
3002 (indirect_return_needed): New.
3003 (indirect_return_bnd_needed): Likewise.
3004 (output_indirect_thunk_function): Add a bool argument for
3006 (output_indirect_thunk_function): Don't generate alias for
3007 function return thunk.
3008 (ix86_code_end): Call output_indirect_thunk_function to generate
3009 function return thunks.
3010 (ix86_output_function_return): Set indirect_return_bnd_needed
3011 and indirect_return_needed instead of indirect_thunk_bnd_needed
3012 and indirect_thunk_needed.
3014 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3016 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3017 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3018 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3020 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3021 Paul Hua <paul.hua.gm@gmail.com>
3024 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3026 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3028 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3029 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3032 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3035 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3036 instead of GET_MODE_SIZE when comparing Units.
3038 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3041 * varasm.c (hash_section): Return an unchangeble hash value
3042 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3043 Return !aarch64_can_use_per_function_literal_pools_p ().
3045 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3048 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3049 rather than address of cmode as last argument to prepare_cmp_insn.
3051 2018-03-15 Julia Koval <julia.koval@intel.com>
3053 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3054 F_AVX512VNNI, F_AVX512BITALG): New.
3056 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3059 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3060 insn for floating-point loads and stores.
3062 2018-03-14 Carl Love <cel@us.ibm.com>
3064 * config/rs6000/rs6000-c.c: Add macro definitions for
3065 ALTIVEC_BUILTIN_VEC_PERMXOR.
3066 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3067 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3068 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3069 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3071 * config/doc/extend.texi: Add prototypes for vec_permxor.
3073 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3076 * diagnostic-show-locus.c (class layout_point): Convert m_line
3077 from int to linenum_type.
3078 (line_span::comparator): Use linenum "compare" function when
3079 comparing line numbers.
3080 (test_line_span): New function.
3081 (layout_range::contains_point): Convert param "row" from int to
3083 (layout_range::intersects_line_p): Likewise.
3084 (layout::will_show_line_p): Likewise.
3085 (layout::print_source_line): Likewise.
3086 (layout::should_print_annotation_line_p): Likewise.
3087 (layout::print_annotation_line): Likewise.
3088 (layout::print_leading_fixits): Likewise.
3089 (layout::annotation_line_showed_range_p): Likewise.
3090 (struct line_corrections): Likewise for field m_row.
3091 (line_corrections::line_corrections): Likewise for param "row".
3092 (layout::print_trailing_fixits): Likewise.
3093 (layout::get_state_at_point): Likewise.
3094 (layout::get_x_bound_for_row): Likewise.
3095 (layout::print_line): Likewise.
3096 (diagnostic_show_locus): Likewise for locals "last_line" and
3098 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3099 * input.c (selftest::test_linenum_comparisons): New function.
3100 (selftest::input_c_tests): Call it.
3101 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3102 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3103 * selftest.h (ASSERT_GT): New macro.
3104 (ASSERT_GT_AT): New macro.
3105 (ASSERT_LT): New macro.
3106 (ASSERT_LT_AT): New macro.
3108 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3110 PR rtl-optimization/84780
3111 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3113 2018-03-14 Martin Liska <mliska@suse.cz>
3115 * tree.c (record_node_allocation_statistics): Use
3116 get_stats_node_kind.
3117 (get_stats_node_kind): New function extracted from
3118 record_node_allocation_statistics.
3119 (free_node): Use get_stats_node_kind.
3121 2018-03-14 Richard Biener <rguenther@suse.de>
3123 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3124 that the value-set of ANTIC_IN doesn't grow.
3127 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3129 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3130 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3132 2018-03-14 Julia Koval <julia.koval@intel.com>
3134 * config.gcc (icelake-client, icelake-server): New.
3136 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3137 (initial_ix86_arch_features): Ditto.
3138 (PTA_SKYLAKE): Add SGX.
3139 (PTA_ICELAKE): Remove.
3140 (PTA_ICELAKE_CLIENT): New.
3141 (PTA_ICELAKE_SERVER): New.
3142 (ix86_option_override_internal): Split up icelake on icelake client and
3144 (get_builtin_code_for_version): Ditto.
3145 (fold_builtin_cpu): Ditto.
3146 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3147 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3148 * config/i386/i386.h (processor_type): Ditto.
3149 * doc/invoke.texi: Ditto.
3151 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3154 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3155 INTEGER_CST offset, add it together with bitpos / 8 and
3156 sign extend based on POINTER_SIZE.
3160 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3163 * config/i386/constraints.md (Yc): New register constraint.
3164 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3165 Use Yc constraint for alternative 2 of operand 0. Remove
3166 preferred_for_speed attribute.
3168 2018-03-14 Richard Biener <rguenther@suse.de>
3170 PR tree-optimization/84830
3171 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3172 with the old one to avoid oscillations.
3174 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3177 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3179 (assign_by_spills): Return a flag of reload assignment failure.
3180 Do not process the reload assignment failures. Do not spill other
3181 reload pseudos if they has the same reg class. Update n if
3183 (lra_assign): Add a return arg. Set up from the result of
3184 assign_by_spills call.
3185 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3186 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3187 usage_insns if it is not NULL.
3188 (spill_hard_reg_in_range): New function.
3189 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3190 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3191 function prototypes.
3192 (lra_assign): Change prototype.
3193 * lra.c (lra): Add code to deal with fails by splitting hard reg
3196 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3198 * config/riscv/riscv.opt (mrelax): New option.
3199 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3200 "norelax" when riscv_mrelax is disabled.
3201 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3203 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3206 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3207 reassociation for int modes.
3209 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3211 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3212 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3214 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3215 * config/aarch64/aarch64-sve.md
3216 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3217 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3218 (*extend<mode><Vwide>2): Rename to...
3219 (aarch64_sve_extend<mode><Vwide>2): ...this.
3220 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3221 renaming the old pattern to...
3222 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3224 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3225 define_expand, renaming the old pattern to...
3226 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3227 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3228 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3229 account when deciding which SVE instruction the optab should use.
3230 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3232 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3234 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3235 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3236 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3237 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3238 (tlsdesc_small_<mode>): Turn a define_expand and use
3239 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3240 (tlsdesc_small_advsimd_<mode>): ...this.
3241 (tlsdesc_small_sve_<mode>): New pattern.
3243 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3245 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3246 (UNSPEC_UMUL_HIGHPART): New constants.
3247 (MUL_HIGHPART): New int iteraor.
3248 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3249 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3251 (*<su>mul<mode>3_highpart): New define_insn.
3253 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3256 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3259 2018-03-13 Martin Liska <mliska@suse.cz>
3262 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3264 (sem_item_optimizer::~sem_item_optimizer): Release it.
3265 (sem_item_optimizer::merge_classes): Register variable aliases.
3266 (sem_item_optimizer::fixup_pt_set): New function.
3267 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3268 * ipa-icf.h: Declare new variables and functions.
3270 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3273 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3274 integer_pow2p@2 and test integer_pow2p in condition.
3275 (A < 0 ? C : 0): Similarly for @1.
3278 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3279 characters starting at p contain '\0' character, don't look beyond
3283 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3284 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3287 * reg-stack.c (change_stack): Change update_end var from int to
3288 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3289 also call set_block_for_insn on the newly added insns and rescan.
3292 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3293 on the last operand.
3296 * tree.c (stabilize_reference_1): Return save_expr (e) for
3297 STATEMENT_LIST even if it doesn't have side-effects.
3299 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3301 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3303 2018-03-12 Renlin Li <renlin.li@arm.com>
3305 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3306 aarch64_output_scalar_simd_mov_immediate.
3308 2018-03-12 Martin Sebor <msebor@redhat.com>
3310 PR tree-optimization/83456
3311 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3312 for perfectly overlapping calls to memcpy.
3313 (gimple_fold_builtin_memory_chk): Same.
3314 (gimple_fold_builtin_strcpy): Handle no-warning.
3315 (gimple_fold_builtin_stxcpy_chk): Same.
3316 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3318 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3320 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3321 parameter. Use it for SFmode.
3322 (rs6000_function_arg_advance_1): Adjust.
3323 (rs6000_function_arg): Adjust.
3324 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3326 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3328 PR rtl-optimization/84169
3329 PR rtl-optimization/84780
3330 * combine.c (can_combine_p): Check for a 2-insn combination whether
3331 the destination register is used between the two insns, too.
3333 2018-03-12 Richard Biener <rguenther@suse.de>
3335 PR tree-optimization/84803
3336 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3337 for refs DR analysis didn't process.
3339 2018-03-12 Richard Biener <rguenther@suse.de>
3341 PR tree-optimization/84777
3342 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3343 force-vectorize loops ignore whether we are optimizing for size.
3345 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3347 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3348 (TARGET_MD_ASM_ADJUST): Define.
3350 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3351 Kito Cheng <kito.cheng@gmail.com>
3352 Chung-Ju Wu <jasonwucj@gmail.com>
3354 * config/nds32/nds32.c (nds32_compute_stack_frame,
3355 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3356 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3357 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3358 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3359 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3360 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3361 * config/nds32/nds32.md (prologue, epilogue): Use macro
3362 NDS32_V3PUSH_AVAILABLE_P to do checking.
3364 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3367 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3368 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3369 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3370 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3371 addition of most attributes on !orig_type_die or the attribute not
3372 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3374 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3375 Chung-Ju Wu <jasonwucj@gmail.com>
3377 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3379 * config/nds32/nds32.opt (mvh): New option.
3381 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3382 Chung-Ju Wu <jasonwucj@gmail.com>
3384 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3386 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3387 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3390 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3391 Chung-Ju Wu <jasonwucj@gmail.com>
3393 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3395 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3396 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3398 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3399 Kito Cheng <kito.cheng@gmail.com>
3400 Chung-Ju Wu <jasonwucj@gmail.com>
3402 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3403 UNSPEC_FFMISM and UNSPEC_FLMISM.
3404 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3405 for ffb, ffmism and flmism.
3406 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3407 (unspec_ffmism): Ditto.
3408 (unspec_flmism): Ditto.
3409 (nds32_expand_builtin_impl): Check if string extension is available.
3410 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3411 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3413 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3416 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3419 * lra-assigns.c (assign_by_spills): Return a flag of reload
3420 assignment failure. Do not process the reload assignment
3421 failures. Do not spill other reload pseudos if they has the same
3423 (lra_assign): Add a return arg. Set up from the result of
3424 assign_by_spills call.
3425 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3426 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3427 usage_insns if it is not NULL.
3428 (spill_hard_reg_in_range): New function.
3429 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3430 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3431 function prototypes.
3432 (lra_assign): Change prototype.
3433 * lra.c (lra): Add code to deal with fails by splitting hard reg
3436 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3439 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3441 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3444 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3445 (dw_val_node): Add val_symbolic_view.
3446 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3447 (symview_upper_bound): New.
3448 (new_line_info_table): Initialize symviews_since_reset.
3449 (dwarf2out_source_line): Count symviews_since_reset and set
3450 symview_upper_bound.
3451 (dw_val_equal_p): Handle symview.
3452 (add_AT_symview): New.
3453 (print_dw_val): Handle symview.
3454 (attr_checksum, attr_checksum_ordered): Likewise.
3455 (same_dw_val_p, size_of_die): Likewise.
3456 (value_format, output_die): Likewise.
3457 (add_high_low_attributes): Use add_AT_symview for entry_view.
3458 (dwarf2out_finish): Reset symview_upper_bound, clear
3461 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3464 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3465 Add strict argument and use it.
3466 (rs6000_split_multireg_move): Update for new strict argument.
3467 (mem_operand_gpr): Disallow all non-offsettable addresses.
3468 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3470 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3473 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3474 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3475 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3478 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3479 decl, use remap_type if we want to use the type.
3481 2018-03-09 Martin Sebor <msebor@redhat.com>
3483 PR tree-optimization/84526
3484 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3486 (builtin_access::generic_overlap): Be prepared to handle non-array
3489 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3491 PR rtl-optimization/84682
3492 * lra-constraints.c (process_address_1): Check is_address flag
3493 for address constraints.
3494 (process_alt_operands): Likewise.
3495 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3496 preprocess_constraints.
3497 * recog.h (preprocess_constraints): Add oploc parameter.
3499 * recog.c (preprocess_constraints): Test address_operand for
3500 CT_ADDRESS constraints.
3502 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3505 * lra-assigns.c (assign_by_spills): Return a flag of reload
3506 assignment failure. Do not process the reload assignment
3507 failures. Do not spill other reload pseudos if they has the same
3509 (lra_assign): Add a return arg. Set up from the result of
3510 assign_by_spills call.
3511 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3512 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3513 usage_insns if it is not NULL.
3514 (spill_hard_reg_in_range): New function.
3515 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3516 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3517 function prototypes.
3518 (lra_assign): Change prototype.
3519 * lra.c (lra): Add code to deal with fails by splitting hard reg
3522 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3525 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3526 Accept complain bool parameter. Only emit errors if it is true.
3527 (arm_parse_cpu_option_name): Likewise.
3528 (arm_target_thumb_only): Adjust callers of the above.
3529 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3530 prototype to take a default true bool parameter.
3531 (arm_parse_arch_option_name): Likewise.
3533 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3534 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3538 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3539 * configure: Regenerate.
3540 * configure.ac ("linker --version-script option"): New.
3541 ("linker soname option"): New.
3543 2018-03-09 Richard Biener <rguenther@suse.de>
3545 PR tree-optimization/84775
3546 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3547 immediate uses of predicate stmts and mark them modified.
3550 PR tree-optimization/84178
3551 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3553 (version_loop_for_if_conversion): Delay update_ssa call.
3554 (tree_if_conversion): Delay update_ssa until after predicate
3557 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3560 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3561 when the function accesses prior frames.
3563 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3566 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3567 gen_llsym, otherwise call maybe_gen_llsym.
3570 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3571 has ',' character inside of it.
3573 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3576 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3577 as clobbering CC_REGNUM.
3579 2018-03-08 Richard Biener <rguenther@suse.de>
3582 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3583 (follow_copies_to_constant): Do not follow SSA names registered
3586 2018-03-08 Richard Biener <rguenther@suse.de>
3588 PR tree-optimization/84178
3589 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3591 (version_loop_for_if_conversion): Delay update_ssa call.
3592 (tree_if_conversion): Delay update_ssa until after predicate
3595 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3597 PR tree-optimization/84178
3598 * tree-if-conv.c (release_bb_predicate): Remove the
3599 the assertion that the stmts have NULL use_ops.
3600 Discard the statements, asserting that they haven't
3601 yet been added to a BB.
3603 2018-03-08 Richard Biener <rguenther@suse.de>
3605 PR tree-optimization/84746
3606 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3607 (phi_translate): Pass in destination ANTIC_OUT set.
3608 (phi_translate_1): Likewise. For a simplified result lookup
3609 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3610 (phi_translate_set): Adjust.
3611 (do_pre_regular_insertion): Likewise.
3612 (do_pre_partial_partial_insertion): Likewise.
3614 2018-03-08 Martin Liska <mliska@suse.cz>
3616 PR gcov-profile/84735
3617 * doc/gcov.texi: Document usage of profile files.
3618 * gcov-io.h: Document changes in the format.
3620 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3624 * dwarf2out.c (struct dw_line_info_table): Update comments for
3626 (FORCE_RESET_NEXT_VIEW): New.
3627 (FORCE_RESETTING_VIEW_P): New.
3628 (RESETTING_VIEW_P): Check for -1 too.
3629 (ZERO_VIEW_P): Likewise.
3630 (new_line_info_table): Force-reset next view.
3631 (dwarf2out_begin_function): Likewise.
3632 (dwarf2out_source_line): Simplify zero_view_p initialization.
3633 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3634 view directly. Omit view when omitting .loc at line 0.
3636 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3638 PR tree-optimization/84740
3639 * tree-switch-conversion.c (process_switch): Call build_constructors
3640 only if info.phi_count is non-zero.
3642 PR tree-optimization/84739
3643 * tree-tailcall.c (find_tail_calls): Check call arguments against
3644 DECL_ARGUMENTS (current_function_decl) rather than
3645 DECL_ARGUMENTS (func) when checking for tail recursion.
3647 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3649 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3650 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3651 Volker Reichelt's entry and add entries for people that perform
3652 GCC fuzzy testing and report numerous bugs.
3654 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3657 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3658 readonly data in sdata, if that is disabled.
3659 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3660 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3661 -mreadonly-in-sdata option.
3663 2018-03-07 Martin Sebor <msebor@redhat.com>
3665 PR tree-optimization/84468
3666 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3667 basic block when looking for nul assignment.
3669 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3672 * except.h (output_function_exception_table): Adjust prototype.
3673 * except.c (output_function_exception_table): Remove FNNAME parameter
3674 and add SECTION parameter. Ouput one part of the table at a time.
3675 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3676 the first part of the exception table and emit unwind directives.
3677 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3678 (i386_pe_seh_cold_init): Likewise.
3679 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3680 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3681 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3682 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3683 * config/i386/winnt.c: Include except.h.
3684 (struct seh_frame_state): Add reg_offset, after_prologue and
3685 in_cold_section fields.
3686 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3687 (i386_pe_seh_cold_init): New function.
3688 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3689 to seh->in_cold_section.
3690 (seh_emit_push): Record the offset of the push.
3691 (seh_emit_save): Record the offet of the save.
3692 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3693 Test seh->after_prologue to disregard the epilogue.
3694 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3695 (i386_pe_end_cold_function): New function.
3697 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3700 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3701 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3704 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3706 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3707 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3710 * multiple_target.c: Include tree-inline.h and intl.h.
3711 (expand_target_clones): Diagnose and fail if node->definition and
3712 !tree_versionable_function_p (node->decl).
3714 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3716 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3718 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3719 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3720 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3722 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3725 * combine.c (try_combine): Use reg_or_subregno instead of handling
3726 just paradoxical SUBREGs and REGs.
3728 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3730 * config/arc/arc.c (arc_finalize_pic): Remove function.
3731 (arc_must_save_register): We use single base PIC register, remove
3732 checks to save/restore the PIC register.
3733 (arc_expand_prologue): Likewise.
3734 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3736 (arc_verify_short): Likewise.
3737 (arc_attr_type): Likewise.
3738 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3739 (walk_stores): Likewise.
3740 (arc_address_cost): Make it static.
3741 (arc_verify_short): Likewise.
3742 (branch_dest): Likewise.
3743 (arc_attr_type): Likewise.
3744 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3745 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3746 (arc_final_prescan_insn): Remove inserting the nops due to
3747 hardware hazards. It is done in reorg step.
3748 (insn_length_variant_t): Remove.
3749 (insn_length_parameters_t): Likewise.
3750 (arc_insn_length_parameters): Likewise.
3751 (arc_get_insn_variants): Likewise.
3752 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3754 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3757 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3760 PR tree-optimization/84687
3761 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3763 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3765 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3767 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3768 Rename to ppc_speculation_barrier.
3769 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3770 __builtin_ppc_speculation_barrier.
3772 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3775 * combine.c (combine_simplify_rtx): Don't try to simplify if
3776 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3779 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3781 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3782 to 32 bytes when compiling for POWER9.
3784 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3787 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3788 regparm >= 3 with no arg reg available also for calls with
3789 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3792 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3794 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3796 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3799 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3801 2018-03-05 Richard Biener <rguenther@suse.de>
3803 PR tree-optimization/84486
3804 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3805 When inserting a __builtin_assume_aligned call set the LHS
3806 SSA name alignment info accordingly.
3808 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3810 PR tree-optimization/84114
3811 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3812 Avoid reassociation of FLOAT_MODE addition.
3814 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3816 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3817 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3818 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3819 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3820 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3821 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3822 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3824 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3825 __builtin_ia32_wbinvd): New builtins.
3826 (SPECIAL_ARGS2): New.
3827 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3828 (SPECIAL_ARGS2): New.
3829 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3830 (ix86_valid_target_attribute_inner_p): Ditto.
3831 (ix86_init_mmx_sse_builtins): Add special_args2.
3832 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3833 TARGET_WBNOINVD_P): New.
3834 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3835 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3836 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3837 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3838 * config/i386/pconfigintrin.h: New file.
3839 * config/i386/wbnoinvdintrin.h: Ditto.
3840 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3841 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3843 2018-03-05 Richard Biener <rguenther@suse.de>
3845 PR tree-optimization/84670
3846 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3848 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3849 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3850 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3851 doesn't grow if all successors (recursively) were visited at least
3854 2018-03-05 Richard Biener <rguenther@suse.de>
3856 PR tree-optimization/84650
3857 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3858 if executed in the loop pipeline.
3860 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3862 * doc/configfiles.texi (Configuration Files): Move info about
3863 conditionalizing $target-protos.h to...
3864 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3865 differs from $target-protos.h.
3867 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3868 Chung-Ju Wu <jasonwucj@gmail.com>
3870 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3871 * config/nds32/nds32-multiple.md (setmemsi): Define.
3872 * config/nds32/nds32-memory-manipulation.c
3873 (nds32_gen_dup_4_byte_to_word_value): New.
3874 (emit_setmem_word_loop): New.
3875 (emit_setmem_byte_loop): New.
3876 (nds32_expand_setmem_loop): New.
3877 (nds32_expand_setmem_loop_v3m): New.
3878 (nds32_expand_setmem_unroll): New.
3879 (nds32_expand_setmem): New.
3881 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3882 Chung-Ju Wu <jasonwucj@gmail.com>
3884 * config/nds32/nds32-memory-manipulation.c
3885 (nds32_emit_load_store): New.
3886 (nds32_emit_post_inc_load_store): New.
3887 (nds32_emit_mem_move): New.
3888 (nds32_emit_mem_move_block): New.
3889 (nds32_expand_movmemsi_loop_unknown_size): New.
3890 (nds32_expand_movmemsi_loop_known_size): New.
3891 (nds32_expand_movmemsi_loop): New.
3892 (nds32_expand_movmemsi_unroll): New.
3893 (nds32_expand_movmemqi): Rename ...
3894 (nds32_expand_movmemsi): ... to this.
3895 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3896 (movmemsi): ... to this.
3897 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3898 (nds32_expand_movmemsi): ... to this.
3900 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3901 Monk Chiang <sh.chiang04@gmail.com>
3902 Chung-Ju Wu <jasonwucj@gmail.com>
3904 * config/nds32/nds32-protos.h
3905 (nds32_expand_load_multiple): New arguments.
3906 (nds32_expand_store_multiple): Ditto.
3907 (nds32_valid_multiple_load_store): Rename ...
3908 (nds32_valid_multiple_load_store_p): ... to this.
3909 * config/nds32/nds32-memory-manipulation.c
3910 (nds32_expand_load_multiple): Refine implementation.
3911 (nds32_expand_store_multiple): Ditto.
3912 * config/nds32/nds32-multiple.md
3913 (load_multiple): Update nds32_expand_load_multiple interface.
3914 (store_multiple): Update nds32_expand_store_multiple interface.
3915 * config/nds32/nds32-predicates.c
3916 (nds32_valid_multiple_load_store): Rename ...
3917 (nds32_valid_multiple_load_store_p): ... to this and refine
3919 * config/nds32/predicates.md
3920 (nds32_load_multiple_and_update_address_operation): New predicate.
3921 (nds32_store_multiple_and_update_address_operation): New predicate.
3923 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3924 Chung-Ju Wu <jasonwucj@gmail.com>
3926 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3927 (combo): New attribute.
3928 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3930 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3932 * config/nds32/nds32.opt: Change -mcmodel= default value.
3934 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3935 Monk Chiang <sh.chiang04@gmail.com>
3936 Chung-Ju Wu <jasonwucj@gmail.com>
3938 * config/nds32/constants.md (unspec_element): New enum.
3939 * config/nds32/constraints.md (Umw): New constraint.
3940 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3941 * config/nds32/nds32-intrinsic.md: Likewise.
3942 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3943 (nds32_valid_smw_lwm_base_p): New.
3944 (nds32_output_smw_single_word): New.
3945 (nds32_output_lmw_single_word): New.
3946 (nds32_expand_unaligned_load): New.
3947 (nds32_expand_unaligned_store): New.
3948 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3949 (nds32_output_smw_single_word): Declare.
3950 (nds32_output_lmw_single_word): Declare.
3951 (nds32_expand_unaligned_load): Declare.
3952 (nds32_expand_unaligned_store): Declare.
3953 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3954 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3955 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3956 NDS32_BUILTIN_UASTORE_DW.
3957 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3960 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3961 Kito Cheng <kito.cheng@gmail.com>
3962 Chung-Ju Wu <jasonwucj@gmail.com>
3964 * config/nds32/nds32-intrinsic.c
3965 (nds32_expand_builtin_null_ftype_reg): Delete.
3966 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3967 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3968 (nds32_read_argument): New.
3969 (nds32_legitimize_target): Ditto.
3970 (nds32_legitimize_argument): Ditto.
3971 (nds32_check_constant_argument): Ditto.
3972 (nds32_expand_unop_builtin): Ditto.
3973 (nds32_expand_unopimm_builtin): Ditto.
3974 (nds32_expand_binop_builtin): Ditto.
3975 (nds32_builtin_decl_impl): Ditto.
3976 (builtin_description): Ditto.
3977 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3978 (nds32_init_builtins_impl): Ditto.
3979 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3980 (nds32_builtin_decl): New.
3981 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3982 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3984 2018-03-02 Jeff Law <law@redhat.com>
3986 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3987 (redundant_insn, fill_simple_delay_slots): Likewise.
3988 (fill_slots_from_thread): Likewise.
3989 * resource.c (mark_referenced_resources): Likewise.
3990 (mark_set_resources, find_dead_or_set_registers): Likewise.
3992 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3994 * substring-locations.h (format_warning_va): Formatting fix for
3996 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3998 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3999 * substring-locations.c: Include intl.h.
4000 (format_warning_va): Turned into small wrapper around
4001 format_warning_n_va, renamed to ...
4002 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4003 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4005 (format_warning_at_substring_n): New function.
4006 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4007 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4008 format_warning_at_substring with just a shorter name instead of
4009 const function pointer.
4010 (fmtwarn_n): New function.
4011 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4012 appropriate, get rid of all the fmtstr temporaries, move conditionals
4013 with G_() wrapped string literals directly into fmtwarn arguments,
4014 cast dir.len to (int), formatting fixes.
4016 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4018 * doc/invoke.texi: Remove "Cilk Plus" references.
4020 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4021 Richard Biener <rguenther@suse.de>
4024 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4025 for error or warning attributes if CALL_FROM_THUNK_P is set.
4028 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4031 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4032 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4035 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4036 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4038 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4039 instead of -1U in last predictors element's probability member.
4041 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4044 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4045 arguments if they are comparable.
4047 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4049 PR tree-optimization/84634
4050 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4051 masks and masked_loop_p with a single loop_masks, making sure it's
4052 null for bb vectorization.
4054 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4056 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4057 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4058 loop->force_vectorize to check whether there is no alias.
4060 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4063 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4065 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4067 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4068 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4069 instead of a loop around prev_real_insn.
4070 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4074 * config/i386/i386.c (ix86_print_operand): Use conditional
4075 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4078 2018-03-02 Richard Biener <rguenther@suse.de>
4080 PR tree-optimization/84427
4081 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4082 (bitmap_set_subtract_values): Rewrite to handle multiple
4085 (prune_clobbered_mems): Likewise.
4086 (phi_translate): Take edge instead of pred/phiblock.
4087 (phi_translate_1): Likewise.
4088 (phi_translate_set): Likewise. Insert all translated
4089 exprs for a value into the set, keeping possibly multiple
4090 expressions per value.
4091 (compute_antic_aux): Adjust for phi_translate changes.
4092 When intersecting union the expressions and prune those
4093 not in the final value set, keeping possibly multiple
4094 expressions per value. Do not use value-insertion
4095 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4096 all expressions. Add verification that the value-sets
4097 only shrink during iteration.
4098 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4099 (do_pre_regular_insertion): Likewise.
4100 (do_pre_partial_partial_insertion): Likewise.
4102 2018-03-02 Richard Biener <rguenther@suse.de>
4105 * config/darwin.c (saved_debug_info_level): New static global.
4106 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4107 (darwin_asm_lto_end): Restore debug info generation settings.
4109 2018-03-01 Martin Liska <mliska@suse.cz>
4112 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4115 2018-03-01 Richard Biener <rguenther@suse.de>
4118 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4119 type annotation with LTO when debug was disabled at compile-time.
4121 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4123 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4125 (mips_final_postscan_insn): Likewise.
4127 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4129 PR rtl-optimization/84528
4130 * alias.c (init_alias_target): Add commentary.
4131 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4132 a unique base value if the frame pointer is not eliminated
4133 to the stack pointer.
4135 2018-03-01 Tom de Vries <tom@codesourcery.com>
4137 PR rtl-optimization/83327
4138 * lra-int.h (hard_regs_spilled_into): Declare.
4139 * lra.c (hard_regs_spilled_into): Define.
4140 (init_reg_info): Init hard_regs_spilled_into.
4141 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4142 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4143 (process_bb_lives): Handle hard_regs_spilled_into.
4144 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4145 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4147 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4149 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4150 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4151 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4152 * config/rs6000/aix72.h: New file.
4154 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4156 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4157 instead of warning_at with conditional singular and plural messages
4161 * stor-layout.c (update_alignment_for_field): For
4162 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4163 && !DECL_PACKED (field), do the alignment update, just use
4164 only desired_align instead of MAX (type_align, desired_align)
4166 (place_field): Don't do known_align < desired_align handling
4167 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4168 is non-NULL, instead do it after rli->prev_field handling and
4169 only if not within a bitfield word. For DECL_PACKED (field)
4170 use type_align of BITS_PER_UNIT.
4172 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4174 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4175 superfluous parentheses and trailing spaces.
4177 2018-02-28 Richard Biener <rguenther@suse.de>
4179 PR tree-optimization/84584
4180 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4181 SCoPs with fake exit edge.
4183 2018-02-28 Martin Liska <mliska@suse.cz>
4186 * timevar.c (timer::print): Fix format to properly print 100%
4189 2018-02-28 Richard Biener <rguenther@suse.de>
4192 * genmatch.c (capture_info::walk_match): Do not mark
4193 captured expressions without operands as expr_p given
4194 they act more like predicates and should be subject to
4195 "lost tail" side-effect preserving.
4197 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4199 PR rtl-optimization/81611
4200 * auto-inc-dec.c (attempt_change): Move dead note from
4201 mem_insn if it's the next use of regno
4202 (find_address): Take address use of reg holding
4203 non-incremented value. Add parm to limit search to the named
4205 (merge_in_block): Attempt to use a mem insn that is the next
4206 use of the original regno.
4208 2018-02-27 Martin Sebor <msebor@redhat.com>
4211 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4212 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4214 2018-02-27 Martin Sebor <msebor@redhat.com>
4216 PR translation/84207
4217 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4218 n argument to unsigned HOST_WIDE_INT.
4219 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4220 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4221 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4222 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4224 2018-02-27 Richard Biener <rguenther@suse.de>
4226 PR tree-optimization/84512
4227 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4228 Do not use the estimate returned from record_stmt_cost for
4229 the scalar iteration cost but sum properly using add_stmt_cost.
4231 2018-02-27 Richard Biener <rguenther@suse.de>
4233 PR tree-optimization/84466
4234 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4235 Adjust last change to less strictly validate use operands.
4237 2018-02-27 Martin Liska <mliska@suse.cz>
4239 PR gcov-profile/84548
4240 * gcov.c (process_file): Allow partial overlap and consider it
4241 also as group functions.
4242 (output_lines): Properly calculate range of lines for a group.
4244 2018-02-27 Martin Liska <mliska@suse.cz>
4246 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4247 'ggc' suffixes. Change first column width.
4248 (timer::print): Fix formatting of the column.
4250 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4252 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4253 preserve inline entry blocks for the sake of debug inline
4254 entry point markers alone.
4255 (remove_unused_locals): Suggest in comments a better place to
4256 force the preservation of inline entry blocks that are
4257 otherwise unused, but do not preserve them.
4259 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4261 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4263 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4266 * config/i386/constraints.md (Bs): Replace
4267 ix86_indirect_branch_register with
4268 TARGET_INDIRECT_BRANCH_REGISTER.
4270 * config/i386/i386.md (indirect_jump): Likewise.
4271 (tablejump): Likewise.
4272 (*sibcall_memory): Likewise.
4273 (*sibcall_value_memory): Likewise.
4274 Peepholes of indirect call and jump via memory: Likewise.
4275 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4276 (*sibcall_value_GOT_32): Likewise.
4277 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4278 (GOT_memory_operand): Likewise.
4279 (call_insn_operand): Likewise.
4280 (sibcall_insn_operand): Likewise.
4281 (GOT32_symbol_operand): Likewise.
4282 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4284 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4286 PR rtl-optimization/83496
4287 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4288 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4289 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4290 redundant insn, if any.
4291 (relax_delay_slots): Likewise.
4292 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4294 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4296 PR tree-optimization/83965
4297 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4298 that grouped statements are part of a reduction chain. Return
4299 true if the statement is not marked as a reduction itself but
4301 (vect_recog_dot_prod_pattern): Don't check whether the statement
4302 is part of a group here.
4303 (vect_recog_sad_pattern): Likewise.
4304 (vect_recog_widen_sum_pattern): Likewise.
4306 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4309 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4311 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4314 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4316 (ix86_output_indirect_function_return): New prototype.
4317 (ix86_split_simple_return_pop_internal): Likewise.
4318 * config/i386/i386.c (indirect_return_via_cx): New.
4319 (indirect_return_via_cx_bnd): Likewise.
4320 (indirect_thunk_name): Handle return va CX_REG.
4321 (output_indirect_thunk_function): Create alias for
4322 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4323 (ix86_output_indirect_jmp): Remove the bool argument.
4324 (ix86_output_indirect_function_return): New function.
4325 (ix86_split_simple_return_pop_internal): Likewise.
4326 * config/i386/i386.md (*indirect_jump): Don't pass false
4327 to ix86_output_indirect_jmp.
4328 (*tablejump_1): Likewise.
4329 (simple_return_pop_internal): Change it to define_insn_and_split.
4330 Call ix86_split_simple_return_pop_internal to split it for
4332 (simple_return_indirect_internal): Call
4333 ix86_output_indirect_function_return instead of
4334 ix86_output_indirect_jmp.
4336 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4339 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4340 memset and value initialization afterwards.
4342 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4344 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4346 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4349 * common/config/aarch64/aarch64-common.c
4350 (aarch_option_optimization_table[]): Switch
4351 off fomit-frame-pointer
4353 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4354 Chung-Ju Wu <jasonwucj@gmail.com>
4356 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4358 (store_multiple): Ditto.
4360 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4362 * config.gcc: Add --with-cpu support for nds32 target.
4363 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4364 * config/nds32/nds32.opt: Add -mcpu= option.
4366 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4368 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4369 isel=yes): Warn for these deprecated options.
4371 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4373 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4374 ISA_2_5_MASKS_EMBEDDED.
4376 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4378 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4379 p->max as pointers rather than using iterative_hash_expr.
4381 2018-02-23 Carl Love <cel@us.ibm.com>
4383 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4384 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4386 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4387 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4388 P8V_BUILTIN_VEC_VUNSIGNED2.
4390 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4393 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4394 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4396 * lra-constraints.c (curr_insn_transform): Set up
4397 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4399 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4400 * lra-eliminations.c (spill_pseudos): Ditto.
4401 (process_insn_for_elimination): Ditto.
4402 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4403 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4404 LRA_NON_CLOBBERED_ALT.
4406 2018-02-22 Martin Sebor <msebor@redhat.com>
4408 PR tree-optimization/84480
4409 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4410 to maybe_diag_stxncpy_trunc. Call it.
4411 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4412 from gimple_fold_builtin_strcpy. Print inlining stack.
4413 (handle_builtin_stxncpy): Print inlining stack.
4414 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4416 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4419 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4420 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4421 and -fcheck-pointer-bounds are used together.
4422 (indirect_thunk_prefix): New enum.
4423 (indirect_thunk_need_prefix): New function.
4424 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4425 "_nt" instead of "_bnd" for NOTRACK prefix.
4426 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4427 (output_indirect_thunk_function): Likewise.
4429 (ix86_code_end): Update output_indirect_thunk_function calls.
4430 (ix86_output_indirect_branch_via_reg): Replace
4431 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4432 (ix86_output_indirect_branch_via_push): Likewise.
4433 (ix86_output_function_return): Likewise.
4434 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4435 incompatible with -fcf-protection=branch and
4436 -fcheck-pointer-bounds.
4438 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4441 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4442 Change gcc_assert call to output_operand_lossage.
4444 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4446 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4448 2018-02-22 DJ Delorie <dj@redhat.com>
4449 Sebastian Perta <sebastian.perta@renesas.com>
4450 Oleg Endo <olegendo@gcc.gnu.org>
4452 * config/rx/rx.c (rx_rtx_costs): New function.
4453 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4455 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4457 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4459 2018-02-22 Martin Liska <mliska@suse.cz>
4462 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4463 Add "native" as a possible value.
4465 2018-02-22 Martin Liska <mliska@suse.cz>
4468 * config/i386/i386.c (ix86_option_override_internal):
4469 Add "native" as a possible value for -march and -mtune.
4471 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4474 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4475 to all type variants.
4477 PR tree-optimization/84503
4478 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4479 width as info->bitpos + info->bitsize - start.
4480 (merged_store_group::merge_overlapping): Simplify width computation.
4481 (check_no_overlap): New function.
4482 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4483 start + width and last_order of the group, fail if check_no_overlap
4485 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4486 to group if check_no_overlap fails.
4488 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4490 * config/rs6000/altivec.md: Delete contraint arguments to
4491 define_expand, define_split, and define_peephole2, and in
4492 define_insn_and_split if always unused.
4493 * config/rs6000/darwin.md: Ditto.
4494 * config/rs6000/dfp.md: Ditto.
4495 * config/rs6000/rs6000.md: Ditto.
4496 * config/rs6000/sync.md: Ditto.
4497 * config/rs6000/vector.md: Ditto.
4498 * config/rs6000/vsx.md: Ditto.
4500 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4502 * config/rs6000/altivec.md: Write output control strings as braced
4503 blocks instead of double-quoted strings.
4504 * config/rs6000/darwin.md: Ditto.
4505 * config/rs6000/rs6000.md: Ditto.
4506 * config/rs6000/vector.md: Ditto.
4507 * config/rs6000/vsx.md: Ditto.
4509 2018-02-21 Jason Merrill <jason@redhat.com>
4511 PR c++/84314 - ICE with templates and fastcall attribute.
4512 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4514 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4516 * ipa-cp.c (determine_versionability): Fix comment typos.
4518 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4521 * ipa-cp.c (determine_versionability): Do not version functions caling
4524 2018-02-21 Martin Liska <mliska@suse.cz>
4527 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4528 Add "native" as a possible value.
4529 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4530 the macro when native cpu detection is available.
4532 2018-02-21 Martin Liska <mliska@suse.cz>
4535 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4536 Add "native" as a possible value.
4537 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4538 when native cpu detection is available.
4540 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4541 Martin Sebor <msebor@redhat.com>
4543 PR tree-optimization/84478
4544 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4546 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4547 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4548 support which is conservatively correct, for 2 only stay conservative
4549 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4550 argument to the 2 argument get_range_strlen, adjust 6 arg
4551 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4553 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4554 (gimple_fold_builtin_strlen): Pass true as last argument to
4557 2018-02-20 Martin Sebor <msebor@redhat.com>
4560 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4561 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4562 (builtin_memref::builtin_memref): Factor out parts into
4563 set_base_and_offset and call it.
4565 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4568 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4569 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4570 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4571 search at the associated MODE_INT.
4573 2018-02-20 Jeff Law <law@redhat.com>
4576 PR tree-optimization/81592
4578 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4579 for range data rather than using global data.
4580 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4581 range data rather than using global data.
4582 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4583 pass it to children as needed.
4584 (struct directive::fmtresult): Similarly.
4585 (struct directive::set_width): Similarly.
4586 (struct directive::set_precision): Similarly.
4587 (format_integer, format_directive, parse_directive): Similarly.
4588 (format_none): Accept unnamed vr_values parameter.
4589 (format_percent, format_floating, format_character): Similarly.
4590 (format_string, format_plain): Similarly.
4591 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4592 the EVRP range analyzer for range data rather than using global data.
4593 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4594 gimple-ssa-evrp-analyze.h
4595 (class sprintf_dom_walker): Add after_dom_children member function.
4596 Add evrp_range_analyzer member.
4597 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4598 range analyzer as needed.
4599 (sprintf_dom_walker::after_dom_children): New member function.
4600 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4602 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4603 (evrp_range_analyzer::pop_to_marker): Likewise.
4605 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4607 PR tree-optimization/84419
4608 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4609 with the required type if its current type is compatible but
4612 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4615 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4616 after vectorization.
4618 2018-02-20 Martin Liska <mliska@suse.cz>
4621 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4622 possible values if we don't have a hint.
4624 2018-02-20 Martin Liska <mliska@suse.cz>
4628 * final.c (shorten_branches): Build align_tab array with one
4630 * opts.c (finish_options): Add alignment option limit check.
4631 (MAX_CODE_ALIGN): Likewise.
4632 (MAX_CODE_ALIGN_VALUE): Likewise.
4633 * doc/invoke.texi: Document maximum allowed option value for
4634 all -falign-* options.
4636 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4639 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4640 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4641 * var-tracking.c (emit_note_insn_var_location): Remove all references
4642 to NOTE_INSN_CALL_ARG_LOCATION.
4643 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4644 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4645 Use copy_rtx_if_shared.
4646 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4647 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4648 (dwarf2out_var_location): Remove handling of
4649 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4651 * final.c (final_scan_insn): Remove all references to
4652 NOTE_INSN_CALL_ARG_LOCATION.
4653 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4654 before dumping final insns.
4655 * except.c (emit_note_eh_region_end): Remove all references to
4656 NOTE_INSN_CALL_ARG_LOCATION.
4657 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4658 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4659 * config/arc/arc.c (hwloop_optimize): Likewise.
4660 * config/arm/arm.c (create_fix_barrier): Likewise.
4661 * config/s390/s390.c (s390_chunkify_start): Likewise.
4662 * config/sh/sh.c (find_barrier): Likewise.
4663 * config/i386/i386.c (rest_of_insert_endbranch,
4664 ix86_seh_fixup_eh_fallthru): Likewise.
4665 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4666 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4667 * config/frv/frv.c (frv_function_prologue): Likewise.
4668 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4670 (note_outside_basic_block_p): Remove all references to
4671 NOTE_INSN_CALL_ARG_LOCATION.
4672 * gengtype.c (adjust_field_rtx_def): Likewise.
4673 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4675 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4676 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4679 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4682 PR tree-optimization/84452
4683 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4684 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4687 2018-02-19 Martin Liska <mliska@suse.cz>
4690 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4692 2018-02-19 Martin Liska <mliska@suse.cz>
4693 Richard Sandiford <richard.sandiford@linaro.org>
4695 PR tree-optimization/82491
4696 * gimple-fold.c (get_base_constructor): Make earlier bail out
4699 2018-02-19 Carl Love <cel@us.ibm.com>
4701 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4702 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4704 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4705 P8V_BUILTIN_VEC_NEG.
4707 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4709 * config/rl78/rl78.md (movdf): New define expand.
4711 2018-02-19 Martin Liska <mliska@suse.cz>
4714 * doc/invoke.texi: Fix typo.
4715 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4717 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4719 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4720 handle rs6000_single_float and rs6000_double_float specially for
4723 2018-02-16 Jeff Law <law@redhat.com>
4725 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4728 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4731 * ipa-inline.c (inline_small_functions): Fix a typo.
4733 2018-02-16 Nathan Sidwell <nathan@acm.org>
4735 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4737 2018-02-16 Carl Love <cel@us.ibm.com>
4739 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4740 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4741 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4742 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4743 expansion to P8V_BUILTIN_VEC_FLOAT2.
4745 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4747 PR rtl-optimization/70023
4748 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4749 src_regno into account.
4751 2018-02-16 Carl Love <cel@us.ibm.com>
4753 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4754 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4755 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4756 * config/rs6000/rs6000.c: Remove case statements for
4757 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4758 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4759 and P9V_BUILTIN_VEC_VINSERT4B.
4760 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4761 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4762 * config/rs6000/vsx.md:
4763 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4766 2018-02-16 Carl Love <cel@us.ibm.com>
4768 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4770 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4772 * config/rs6000/rs6000-c.c: Add the definitions for
4773 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4774 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4775 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4776 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4777 definition for insert4b and define insn *insert3b_internal.
4778 * doc/extend.texi: Add documentation for vec_extract4b.
4780 2018-02-16 Nathan Sidwell <nathan@acm.org>
4782 * doc/extend.texi (Backwards Compatibility): Mention friend
4783 injection. Note for-scope is deprecated.
4784 * doc/invoke.texi (-ffriend-injection): Deprecate.
4786 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4788 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4789 that moved to I2, also allow destinations that are a paradoxical
4790 subreg (instead of a normal reg).
4792 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4795 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4798 2018-02-16 Richard Biener <rguenther@suse.de>
4800 PR tree-optimization/84037
4801 PR tree-optimization/84016
4803 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4804 Adjust vec_construct for the fact we need additional higher latency
4805 128bit inserts for AVX256 and AVX512 vector builds.
4806 (ix86_add_stmt_cost): Scale vector construction cost for
4809 2018-02-16 Richard Biener <rguenther@suse.de>
4811 PR tree-optimization/84417
4812 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4813 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4814 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4816 2018-02-16 Martin Liska <mliska@suse.cz>
4819 * internal-fn.def (ASAN_CHECK): Set proper flags.
4820 (ASAN_MARK): Likewise.
4822 2018-02-16 Julia Koval <julia.koval@intel.com>
4824 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4825 from PTA_CANNONLAKE.
4827 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4830 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4831 Use ++iter rather than iter++ for std::list iterators.
4832 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4833 defer deleting them until all nodes in the forest are processed. Do
4834 free even leaf nodes. Change to_process into auto_vec.
4837 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4838 * vec.h (vec_default_construct): Use memset instead of placement new
4839 if BROKEN_VALUE_INITIALIZATION is defined.
4840 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4841 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4844 PR rtl-optimization/83723
4845 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4846 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4847 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4849 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4851 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4853 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4855 PR rtl-optimization/81443
4856 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4857 from inner REGs to paradoxical SUBREGs.
4859 2018-02-16 Richard Biener <rguenther@suse.de>
4861 PR tree-optimization/84399
4862 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4863 For operands we can analyze at their definition make sure we can
4864 analyze them at each use as well.
4866 2018-02-16 Richard Biener <rguenther@suse.de>
4868 PR tree-optimization/84190
4869 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4870 volatile accesses if the decl isn't volatile.
4872 2018-02-15 Jason Merrill <jason@redhat.com>
4874 PR c++/84314 - ICE with templates and fastcall attribute.
4875 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4876 TYPE_CANONICAL on an existing type.
4878 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4880 PR tree-optimization/84383
4881 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4882 dstoff nor call operand_equal_p if dstbase is NULL.
4884 PR tree-optimization/84334
4885 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4886 also a CONSTANT_CLASS_P, punt.
4888 2018-02-14 Jim Wilson <jimw@sifive.com>
4890 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4891 first SMALL_OPERAND check. New local min_second_step. Move assert
4892 to where locals are set. Add TARGET_RVC support.
4893 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4895 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4897 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4899 2018-02-14 Martin Sebor <msebor@redhat.com>
4901 PR tree-optimization/83698
4902 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4903 arrays constrain the offset range to their bounds.
4904 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4905 (builtin_access::overlap): Avoid setting the size of overlap if it's
4907 (maybe_diag_overlap): Also consider arrays when deciding what values
4908 of offsets to include in diagnostics.
4910 2018-02-14 Martin Sebor <msebor@redhat.com>
4913 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4914 that correspond to the kind of a declaration.
4916 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4919 * config/pa/pa.md: Load address of PIC label using the linkage table
4920 if the label is nonlocal.
4922 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4924 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4925 warning message if user requests -maltivec=be.
4926 * doc/invoke.texi: Document deprecation of -maltivec=be.
4928 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4931 * config/rs6000/rs6000-c.c: Update definitions for
4932 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4933 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4935 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4938 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4939 add _get_ssp intrinsics. Remove argument from
4940 __builtin_ia32_rdssp[d|q].
4941 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4942 * config/i386/i386-builtin.def: Remove argument from
4943 __builtin_ia32_rdssp[d|q].
4944 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4945 ix86_expand_special_args_builtin for _rdssp[d|q].
4946 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4947 Clear register before usage.
4948 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4949 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4951 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4953 PR tree-optimization/84357
4954 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4955 operand 1 of an ARRAY_REF too.
4957 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4960 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4961 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4963 (set_of_reg): New struct.
4964 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4965 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4966 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4968 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4969 Split into bitclr, bitset, bitinvert patterns if appropriate.
4970 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4971 use rx_fuse_in_memory_bitop.
4972 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4973 to named insn, correct maximum insn length.
4975 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4978 * machmode.def: Define a complex mode for PARTIAL_INT.
4979 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4981 * doc/rtl.texi: Document CSPImode.
4982 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4984 (msp430_hard_regno_nregs_with_padding): Likewise.
4986 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4989 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4991 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4993 PR rtl-optimization/84169
4994 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4995 we generated a parallel as new i3 and we split that to new i2 and i3
4996 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4997 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4998 those to i2, not i1. Partially rewrite this scan code.
5000 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5003 * stor-layout.c (place_field): For variable length fields, adjust
5004 offset_align afterwards not just based on the field's alignment,
5005 but also on the size.
5008 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5009 of exps and logs in the use_exp2 case.
5011 2018-02-13 Jeff Law <law@redhat.com>
5013 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5016 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5019 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5023 * common.opt (gas-loc-support, gas-locview-support): New.
5024 (ginline-points, ginternal-reset-location-views): New.
5025 * doc/invoke.texi: Document them. Use @itemx where intended.
5026 (gvariable-location-views): Adjust.
5027 * target.def (reset_location_view): New.
5028 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5029 (TARGET_RESET_LOCATION_VIEW): New.
5030 * doc/tm.texi: Rebuilt.
5031 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5032 (dwarf2out_default_as_locview_support): New.
5033 (output_asm_line_debug_info): Use option variables.
5034 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5035 (output_loc_list): Likewise.
5036 (add_high_low_attributes): Check option variables.
5037 Don't output entry view attribute in strict mode.
5038 (gen_inlined_subroutine_die): Check option variables.
5039 (dwarf2out_inline_entry): Likewise.
5040 (init_sections_and_labels): Likewise.
5041 (dwarf2out_early_finish): Likewise.
5042 (maybe_reset_location_view): New, from...
5043 (dwarf2out_var_location): ... here. Call it.
5044 * debug.h (dwarf2out_default_as_loc_support): Declare.
5045 (dwarf2out_default_as_locview_support): Declare.
5046 * hooks.c (hook_int_rtx_insn_0): New.
5047 * hooks.h (hook_int_rtx_insn_0): Declare.
5048 * toplev.c (process_options): Take -gas-loc-support and
5049 -gas-locview-support from dwarf2out. Enable
5050 -gvariable-location-views by default only with locview
5051 assembler support. Enable -ginternal-reset-location-views by
5052 default only if the target defines the corresponding hook.
5053 Enable -ginline-points by default if location views are
5054 enabled; force it disabled if statement frontiers are
5056 * tree-inline.c (expand_call_inline): Check option variables.
5057 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5059 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5061 PR tree-optimization/84321
5062 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5063 handling. Also check whether the anti-range contains any values
5064 that satisfy the mask; switch to a VR_RANGE if not.
5066 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5069 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5071 2018-02-13 Martin Jambor <mjambor@suse.cz>
5074 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5075 of call statements, also set location of a load to a temporary.
5077 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5079 * config/rl78/rl78.c (add_vector_labels): New function.
5080 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5081 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5082 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5083 which checks that no arguments are passed.
5084 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5085 * doc/extend.texi: Documentation for the new attribute.
5087 2018-02-13 Andreas Schwab <schwab@suse.de>
5089 * config/riscv/linux.h (CPP_SPEC): Define.
5091 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5094 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5095 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5096 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5097 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5098 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5099 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5100 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5101 temporarily for AES and PCLMUL builtins.
5103 PR tree-optimization/84339
5104 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5105 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5109 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5110 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5111 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5113 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5115 * omp-simd-clone.h: New file.
5116 * omp-simd-clone.c: Include omp-simd-clone.h.
5117 (expand_simd_clones): No longer static.
5118 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5119 cgraph.h and omp-simd-clone.h.
5120 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5121 (vect_recog_widen_shift_pattern): Formatting fix.
5122 (vect_pattern_recog_1): Don't check optab for calls.
5125 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5126 operands[2] into a REG before using gen_lowpart on it.
5128 2018-02-12 Jeff Law <law@redhat.com>
5131 * config/sh/sh.c (find_barrier): Consider a sibling call
5134 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5135 successfully back substituting a reg.
5137 2018-02-12 Richard Biener <rguenther@suse.de>
5139 PR tree-optimization/84037
5140 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5141 parameter, move visited init to caller.
5142 (vect_slp_analyze_operations): Separate cost from validity
5143 check, initialize visited once for all instances.
5144 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5146 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5148 (vect_model_promotion_demotion_cost): Likewise.
5149 (vectorizable_bswap): Guard cost modeling with !slp_node
5150 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5152 (vectorizable_call): Likewise.
5153 (vectorizable_conversion): Likewise.
5154 (vectorizable_assignment): Likewise.
5155 (vectorizable_shift): Likewise.
5156 (vectorizable_operation): Likewise.
5157 (vectorizable_store): Likewise.
5158 (vectorizable_load): Likewise.
5159 (vectorizable_condition): Likewise.
5160 (vectorizable_comparison): Likewise.
5162 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5165 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5166 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5167 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5169 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5172 * params.def (inline-min-speedup): Increase from 8 to 15.
5173 (max-inline-insns-auto): Decrease from 40 to 30.
5174 * ipa-split.c (consider_split): Add some buffer for function to
5175 be considered inlining candidate.
5176 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5179 2018-02-12 Richard Biener <rguenther@suse.de>
5181 PR tree-optimization/84037
5182 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5183 matched stmts if we cannot swap the non-matched ones.
5185 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5187 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5188 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5189 _mm_maskz_scalef_round_ss): New intrinsics.
5190 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5191 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5192 __builtin_ia32_scalefss_round): Remove.
5193 (__builtin_ia32_scalefsd_mask_round,
5194 __builtin_ia32_scalefss_mask_round): New intrinsics.
5195 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5196 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5197 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5198 "<round_constraint>")): Changed to ...
5199 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5200 "<round_scalar_constraint>")): ... this.
5201 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5202 %0, %1, %2<round_op3>}"): Changed to ...
5203 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5204 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5205 %2<round_scalar_mask_op3>}"): ... this.
5206 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5208 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5210 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5211 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5212 (_mm_maskz_sqrt_round_ss): New intrinsics.
5213 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5214 (__builtin_ia32_sqrtsd_mask_round)
5215 (__builtin_ia32_sqrtss_mask_round): New builtins.
5216 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5217 (__builtin_ia32_sqrtss_round): Remove.
5218 (__builtin_ia32_sqrtsd_mask_round)
5219 (__builtin_ia32_sqrtss_mask_round): New builtins.
5220 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5221 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5222 ((match_operand:VF_128 1 "vector_operand"
5223 "xBm,<round_constraint>")): Changed to ...
5224 ((match_operand:VF_128 1 "vector_operand"
5225 "xBm,<round_scalar_constraint>")): ... this.
5226 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5227 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5228 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5229 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5230 %<iptr>1<round_scalar_mask_op3>}): ... this.
5231 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5232 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5234 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5237 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5238 Cast vec_cmpeq result to correct type.
5239 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5240 Cast vec_cmpgt result to correct type.
5242 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5244 * final.c (final_scan_insn_1): Renamed from...
5245 (final_scan_insn): ... this. New wrapper, to recover
5246 seen from the outermost call in recursive ones.
5247 * config/sparc/sparc.c (output_return): Drop seen from call.
5248 (output_sibcall): Likewise.
5249 * config/visium/visium.c (output_branch): Likewise.
5251 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5253 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5256 2018-02-10 Alan Modra <amodra@gmail.com>
5259 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5260 Specify LR as an input.
5262 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5265 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5266 remove_member_access_dummy_vars): New functions.
5267 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5268 lower_omp_1, execute_lower_omp): Use them.
5270 PR rtl-optimization/84308
5271 * shrink-wrap.c (spread_components): Release todo vector.
5273 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5275 PR rtl-optimization/57193
5276 * ira-color.c (struct allocno_color_data): Add member
5277 conflict_allocno_hard_prefs.
5278 (update_conflict_allocno_hard_prefs): New.
5279 (bucket_allocno_compare_func): Add a preference based on
5280 conflict_allocno_hard_prefs.
5281 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5282 (color_allocnos): Remove a dead code. Initiate
5283 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5285 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5288 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5289 constraint from =wa to wa. Avoid a subreg on the output operand,
5290 instead use a pseudo and subreg it in a move.
5291 (p9_xxbrd_<mode>): Changed to ...
5292 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5293 (p9_xxbrd_v2df): New expander.
5294 (p9_xxbrw_<mode>): Changed to ...
5295 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5296 (p9_xxbrw_v4sf): New expander.
5298 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5300 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5302 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5305 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5306 multiply in 32-bit mode.
5307 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5308 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5311 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5313 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5314 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5315 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5316 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5318 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5321 * dwarf2out.c (is_trivial_indirect_ref): New function.
5322 (dwarf2out_late_global_decl): Do not generate a location
5323 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5324 and that are not defined in the current unit.
5326 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5328 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5329 instead of a libcall for UNORDERED.
5331 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5334 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5335 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5337 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5340 * config/s390/s390.c (s390_set_current_function): Invoke
5341 s390_indirect_branch_settings also if fndecl didn't change.
5343 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5345 * config/rs6000/rs6000.md (blockage): Set length to zero.
5347 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5349 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5351 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5354 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5355 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5359 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5360 PARALLEL incoming that failed vt_get_decl_and_offset check.
5363 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5364 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5366 (get_variable_section): For decls in named .bss* sections pass true as
5367 second argument to bss_initializer_p.
5369 2018-02-09 Marek Polacek <polacek@redhat.com>
5370 Jakub Jelinek <jakub@redhat.com>
5373 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5374 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5375 Sync some changes from cxx_fold_indirect_ref.
5377 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5379 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5381 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5382 (BLOCK_INLINE_ENTRY_LABEL): New.
5383 (dwarf2out_var_location): Disregard inline entry markers.
5384 (inline_entry_data): New struct.
5385 (inline_entry_data_hasher): New hashtable type.
5386 (inline_entry_data_hasher::hash): New.
5387 (inline_entry_data_hasher::equal): New.
5388 (inline_entry_data_table): New variable.
5389 (add_high_low_attributes): Add DW_AT_entry_pc and
5390 DW_AT_GNU_entry_view attributes if a pending entry is found
5391 in inline_entry_data_table. Add old entry_pc attribute only
5392 if debug nonbinding markers are disabled.
5393 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5394 markers are enabled.
5395 (block_within_block_p, dwarf2out_inline_entry): New.
5396 (dwarf2out_finish): Check that no entries remained in
5397 inline_entry_data_table.
5398 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5399 (final_scan_insn, notice_source_line): Likewise.
5400 (rest_of_clean_state): Skip inline entry markers.
5401 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5403 * gimple.c (gimple_build_debug_inline_entry): New.
5404 * gimple.h (enum gimple_debug_subcode): Add
5405 GIMPLE_DEBUG_INLINE_ENTRY.
5406 (gimple_build_debug_inline_entry): Declare.
5407 (gimple_debug_inline_entry_p): New.
5408 (gimple_debug_nonbind_marker_p): Adjust.
5409 * insn-notes.def (INLINE_ENTRY): New.
5410 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5411 inline entry marker notes.
5412 (print_insn): Likewise.
5413 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5414 (INSN_DEBUG_MARKER_KIND): Likewise.
5415 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5416 * tree-inline.c (expand_call_inline): Build and insert
5417 debug_inline_entry stmt.
5418 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5419 inline entry blocks early, if nonbind markers are enabled.
5420 (dump_scope_block): Dump fragment info.
5421 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5422 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5423 (gimple_build_debug_inline_entry): New.
5424 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5425 Enable/disable inline entry points too.
5426 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5427 (DEBUG_INSN): Describe inline entry markers.
5429 * common.opt (gvariable-location-views): New.
5430 (gvariable-location-views=incompat5): New.
5431 * config.in: Rebuilt.
5432 * configure: Rebuilt.
5433 * configure.ac: Test assembler for view support.
5434 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5435 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5436 * dwarf2out.c (var_loc_view): New typedef.
5437 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5438 (dwarf2out_locviews_in_attribute): New.
5439 (dwarf2out_locviews_in_loclist): New.
5440 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5441 (enum dw_line_info_opcode): Add LI_adv_address.
5442 (struct dw_line_info_table): Add view.
5443 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5444 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5445 (zero_view_p): New variable.
5446 (ZERO_VIEW_P): New macro.
5447 (output_asm_line_debug_info): New.
5448 (struct var_loc_node): Add view.
5449 (add_AT_view_list, AT_loc_list): New.
5450 (add_var_loc_to_decl): Add view param. Test it against last.
5451 (new_loc_list): Add view params. Record them.
5452 (AT_loc_list_ptr): Handle loc and view lists.
5453 (view_list_to_loc_list_val_node): New.
5454 (print_dw_val): Handle dw_val_class_view_list.
5455 (size_of_die): Likewise.
5456 (value_format): Likewise.
5457 (loc_list_has_views): New.
5458 (gen_llsym): Set vl_symbol too.
5459 (maybe_gen_llsym, skip_loc_list_entry): New.
5460 (dwarf2out_maybe_output_loclist_view_pair): New.
5461 (output_loc_list): Output view list or entries too.
5462 (output_view_list_offset): New.
5463 (output_die): Handle dw_val_class_view_list.
5464 (output_dwarf_version): New.
5465 (output_compilation_unit_header): Use it.
5466 (output_skeleton_debug_sections): Likewise.
5467 (output_rnglists, output_line_info): Likewise.
5468 (output_pubnames, output_aranges): Update version comments.
5469 (output_one_line_info_table): Output view numbers in asm comments.
5470 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5471 Call maybe_gen_llsym.
5472 (loc_list_from_tree_1): Adjust.
5473 (add_AT_location_description): Create view list attribute if
5474 needed, check it's absent otherwise.
5475 (convert_cfa_to_fb_loc_list): Adjust.
5476 (maybe_emit_file): Call output_asm_line_debug_info for test.
5477 (dwarf2out_var_location): Reset views as needed. Precompute
5478 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5479 attribute. Set view.
5480 (new_line_info_table): Reset next view.
5481 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5482 (dwarf2out_source_line): Likewise. Output view resets and labels to
5483 the assembler, or select appropriate line info opcodes.
5484 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5485 (optimize_string_length): Catch it. Adjust.
5486 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5487 dw_val_class_view_list, and remove it if no longer needed.
5488 (hash_loc_list): Hash view numbers.
5489 (loc_list_hasher::equal): Compare them.
5490 (optimize_location_lists): Check whether a view list symbol is
5491 needed, and whether the locview attribute is present, and
5492 whether they match. Remove the locview attribute if no longer
5494 (index_location_lists): Call skip_loc_list_entry for test.
5495 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5496 Use output_dwarf_version.
5497 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5498 (struct dw_val_node): Add val_view_list.
5499 * final.c (SEEN_NEXT_VIEW): New.
5500 (set_next_view_needed): New.
5501 (clear_next_view_needed): New.
5502 (maybe_output_next_view): New.
5503 (final_start_function): Rename to...
5504 (final_start_function_1): ... this. Take pointer to FIRST,
5505 add SEEN parameter. Emit param bindings in the initial view.
5506 (final_start_function): Reintroduce SEEN-less interface.
5507 (final): Rename to...
5508 (final_1): ... this. Take SEEN parameter. Output final pending
5509 next view at the end.
5510 (final): Reintroduce seen-less interface.
5511 (final_scan_insn): Output pending next view before switching
5512 sections or ending a block. Mark the next view as needed when
5513 outputting variable locations. Notify debug backend of section
5514 changes, and of location view changes.
5515 (rest_of_handle_final): Adjust.
5516 * toplev.c (process_options): Autodetect value for debug variable
5517 location views option. Warn on incompat5 without -gdwarf-5.
5518 * doc/invoke.texi (gvariable-location-views): New.
5519 (gvariable-location-views=incompat5): New.
5520 (gno-variable-location-views): New.
5522 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5524 PR tree-optimization/84136
5525 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5526 that the result of find_edge is non-NULL.
5528 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5531 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5532 storing integer register in SImode. Fix cost of 256 and 512
5533 byte aligned SSE register store.
5535 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5537 * config/i386/i386.c (ix86_multiplication_cost): Fix
5538 multiplication cost for TARGET_AVX512DQ.
5540 2018-02-08 Marek Polacek <polacek@redhat.com>
5542 PR tree-optimization/84238
5543 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5546 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5548 PR tree-optimization/84265
5549 * tree-vect-stmts.c (vectorizable_store): Don't treat
5550 VMAT_CONTIGUOUS accesses as grouped.
5551 (vectorizable_load): Likewise.
5553 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5555 PR tree-optimization/81635
5556 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5557 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5558 (test_round_for_mask): New functions.
5559 (wide_int_cc_tests): Call test_round_for_mask.
5560 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5561 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5562 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5563 range returned by get_range_info.
5565 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5568 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5569 * symtab.c: Include builtins.h
5570 (symtab_node::output_to_lto_symbol_table_p): Move here
5571 from lto-streamer-out.c:output_symbol_p.
5572 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5573 (output_symbol_p): Move all logic to symtab.c
5574 (produce_symtab): Update.
5576 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5578 * config/s390/s390-opts.h (enum indirect_branch): Define.
5579 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5580 (s390_indirect_branch_via_thunk)
5581 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5582 (enum s390_indirect_branch_type): Define.
5583 * config/s390/s390.c (struct s390_frame_layout, struct
5584 machine_function): Remove.
5585 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5586 (indirect_branch_table_label_no, indirect_branch_table_name):
5588 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5589 (enum s390_indirect_branch_option): Define.
5590 (s390_return_addr_from_memory): New function.
5591 (s390_handle_string_attribute): New function.
5592 (s390_attribute_table): Add new attribute handler.
5593 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5594 (s390_indirect_branch_via_thunk): New function.
5595 (s390_indirect_branch_via_inline_thunk): New function.
5596 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5597 sibling call optimization for non z10 compiles.
5598 (s390_emit_call): Force indirect branch target to be a single
5599 register. Add r1 clobber for non-z10 compiles.
5600 (s390_emit_epilogue): Emit return jump via return_use expander.
5601 (s390_reorg): Handle JUMP_INSNs as execute targets.
5602 (s390_option_override_internal): Perform validity checks for the
5603 new command line options.
5604 (s390_indirect_branch_attrvalue): New function.
5605 (s390_indirect_branch_settings): New function.
5606 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5607 (s390_output_indirect_thunk_function): New function.
5608 (s390_code_end): Implement target hook.
5609 (s390_case_values_threshold): Implement target hook.
5610 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5612 * config/s390/s390.h (struct s390_frame_layout)
5613 (struct machine_function): Move here from s390.c.
5614 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5615 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5616 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5617 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5618 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5619 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5620 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5621 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5622 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5623 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5624 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5625 (mnemonic attribute): Add values which aren't recognized
5627 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5628 pattern for branch conversion. Fix mnemonic attribute.
5629 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5630 indirect branch via thunk if requested.
5631 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5632 ("*indirect_jump"): Disable for branch conversion using out of
5634 ("indirect_jump_via_thunk<mode>_z10")
5635 ("indirect_jump_via_thunk<mode>")
5636 ("indirect_jump_via_inlinethunk<mode>_z10")
5637 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5638 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5639 ("casesi_jump_via_inlinethunk<mode>_z10")
5640 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5641 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5642 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5643 ("*indirect2_jump"): Disable for branch conversion.
5644 ("casesi_jump"): Turn into expander and expand patterns for branch
5646 ("return_use"): New expander.
5647 ("*return"): Emit return via thunk and rename it to ...
5648 ("*return<mode>"): ... this one.
5649 * config/s390/s390.opt: Add new options and and enum for the
5652 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5654 * lra-constraints.c (match_reload): Unconditionally use
5655 gen_lowpart_SUBREG, rather than selecting between that
5656 and equivalent gen_rtx_SUBREG code.
5658 2018-02-08 Richard Biener <rguenther@suse.de>
5660 PR tree-optimization/84233
5661 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5662 changed flag instead of boguously re-using phi_inserted.
5664 2018-02-08 Martin Jambor <mjambor@suse.cz>
5666 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5667 static local variables.
5669 2018-02-08 Richard Biener <rguenther@suse.de>
5671 PR tree-optimization/84278
5672 * tree-vect-stmts.c (vectorizable_store): When looking for
5673 smaller vector types to perform grouped strided loads/stores
5674 make sure the mode is supported by the target.
5675 (vectorizable_load): Likewise.
5677 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5679 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5680 Increase LDP/STP opportunities by adding adjacent callee-saves.
5682 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5684 PR rtl-optimization/84068
5685 PR rtl-optimization/83459
5686 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5688 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5690 PR tree-optimization/84224
5691 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5692 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5695 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5698 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5699 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5700 count, remove one USE.
5702 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5704 * doc/install.texi (Configuration): Document the
5705 --with-long-double-format={ibm,ieee} PowerPC configuration
5709 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5710 Convert from define_expand to be define_insn_and_split. Rework
5711 float/double/_Float128 conversions to QI/HI/SImode to work with
5712 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5713 conversions to QI/HImode types did a store and then a load to
5714 truncate the value. For conversions to VSX registers, don't split
5715 the insn, instead emit the code directly. Use the code iterator
5716 any_fix to combine signed and unsigned conversions.
5717 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5718 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5719 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5720 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5721 (fix_<mode>di2_hw): Likewise.
5722 (fixuns_<mode>di2_hw): Likewise.
5723 (fix_<mode>si2_hw): Likewise.
5724 (fixuns_<mode>si2_hw): Likewise.
5725 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5726 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5727 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5728 fix<uns>_trunc<SFDF:mode>si2_p8.
5729 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5731 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5732 (fix<uns>_<mode>_mem): Likewise.
5733 (fctiw<u>z_<mode>_mem): Likewise.
5734 (fix<uns>_<mode>_mem): Likewise.
5735 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5736 the register allocator from doing a direct move to the GPRs to do
5737 a store, and instead use the ISA 3.0 store byte/half-word from
5738 vector register instruction. For IEEE 128-bit floating point,
5739 also optimize stores of 32-bit ints.
5740 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5742 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5744 * genextract.c (push_pathstr_operand): New function to support
5746 (walk_rtx): Call push_pathstr_operand.
5747 (print_path): Support [a-zA-Z].
5749 2018-02-07 Richard Biener <rguenther@suse.de>
5751 PR tree-optimization/84037
5752 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5753 (cse_and_gimplify_to_preheader): Declare.
5754 (vect_get_place_in_interleaving_chain): Likewise.
5755 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5757 (_loop_vec_info::~_loop_vec_info): Delete it.
5758 (cse_and_gimplify_to_preheader): New function.
5759 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5760 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5761 (vectorizable_load): Likewise. For grouped stores always base
5762 the IV on the first element.
5763 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5764 condition before gimplifying.
5766 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5768 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5769 *DIV_EXPR and *MOD_EXPR.
5771 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5774 * config/i386/i386.c (ix86_option_override_internal): Mask out
5775 the CF_SET bit when checking -fcf-protection.
5777 2018-02-07 Tom de Vries <tom@codesourcery.com>
5780 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5783 2018-02-07 Richard Biener <rguenther@suse.de>
5785 PR tree-optimization/84204
5786 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5789 PR tree-optimization/84205
5790 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5791 special-case isl_ast_op_zdiv_r.
5793 PR tree-optimization/84223
5794 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5795 Only add conditions from within the region.
5796 (gather_bbs::after_dom_children): Adjust.
5798 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5801 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5802 * config/avr/avr.md: Only post-reload split REG-REG moves if
5803 either register is GENERAL_REG_P.
5805 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5807 PR tree-optimization/84235
5808 * tree-ssa-scopedtables.c
5809 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5810 if the subtraction is performed in floating point type where NaNs are
5811 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5812 build 1. Formatting fix.
5814 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5817 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5818 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5819 and skip it regardless of bb boundaries. Use CALL_P macro,
5820 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5821 unnecessarily, formatting fix.
5823 2018-02-06 Michael Collison <michael.collison@arm.com>
5825 * config/arm/thumb2.md:
5826 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5827 (*thumb_mov_notscc): Ditto.
5829 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5832 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5833 unsigned_fix, not "s".
5835 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5837 * configure.ac (gcc_fn_eh_frame_ro): New function.
5838 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5839 correct .eh_frame permissions.
5840 * configure: Regenerate.
5842 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5844 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5847 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5849 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5850 Display warning message for -mno-speculate-indirect-jumps.
5852 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5854 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5856 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5858 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5860 PR tree-optimization/84225
5861 * tree-eh.c (find_trapping_overflow): Only call
5862 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5864 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5867 * config/i386/i386.c: Reimplement the check of possible options
5868 -mibt/-mshstk conbination. Change error messages.
5869 * doc/invoke.texi: Fix a typo: remove extra '='.
5871 2018-02-06 Marek Polacek <polacek@redhat.com>
5873 PR tree-optimization/84228
5874 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5876 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5879 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5880 emitted arch directives.
5881 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5882 __ARM_FEATURE_COPROC before changing architectures.
5884 2018-02-06 Richard Biener <rguenther@suse.de>
5886 * config/i386/i386.c (print_reg): Fix typo.
5887 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5889 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5891 * configure: Regenerate.
5893 2018-02-05 Martin Sebor <msebor@redhat.com>
5895 PR tree-optimization/83369
5896 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5899 2018-02-05 Martin Liska <mliska@suse.cz>
5901 * doc/invoke.texi: Cherry-pick upstream r323995.
5903 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5905 * ira.c (ira_init_register_move_cost): Adjust comment.
5907 2018-02-05 Martin Liska <mliska@suse.cz>
5909 PR gcov-profile/84137
5910 * doc/gcov.texi: Fix typo in documentation.
5912 2018-02-05 Martin Liska <mliska@suse.cz>
5914 PR gcov-profile/83879
5915 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5916 using dlopen functionality.
5918 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5920 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5921 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5922 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5923 _mm_maskz_range_round_ss): New intrinsics.
5924 (__builtin_ia32_rangesd128_round)
5925 (__builtin_ia32_rangess128_round): Remove.
5926 (__builtin_ia32_rangesd128_mask_round,
5927 __builtin_ia32_rangess128_mask_round): New builtins.
5928 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5929 __builtin_ia32_rangess128_round): Remove.
5930 (__builtin_ia32_rangesd128_mask_round,
5931 __builtin_ia32_rangess128_mask_round): New builtins.
5932 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5933 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5934 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5935 "<round_saeonly_constraint>")): Changed to ...
5936 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5937 "<round_saeonly_scalar_constraint>")): ... this.
5938 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5939 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5940 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5941 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5942 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5944 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5946 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5948 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5949 Remove all values except native, 8540 and 8548.
5951 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5953 * config/i386/i386.c (ix86_output_function_return): Pass
5954 INVALID_REGNUM, instead of -1, as invalid register number to
5955 indirect_thunk_name and output_indirect_thunk.
5957 2018-02-02 Julia Koval <julia.koval@intel.com>
5959 * config.gcc: Add -march=icelake.
5960 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5961 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5962 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5963 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5964 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5965 (processor_target_table): Add icelake.
5966 (ix86_option_override_internal): Handle new PTAs.
5967 (get_builtin_code_for_version): Handle icelake.
5968 (M_INTEL_COREI7_ICELAKE): New.
5969 (fold_builtin_cpu): Handle icelake.
5970 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5971 * doc/invoke.texi: Add -march=icelake.
5973 2018-02-02 Julia Koval <julia.koval@intel.com>
5975 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5976 to wide_int_bitmask.
5977 * wide-int-bitmask.h: New.
5979 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5982 * config/i386/i386.md: Replace Pmode with word_mode in
5983 builtin_setjmp_setup and builtin_longjmp to support x32.
5985 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5989 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5991 (rs6000_supported_cpu_names): New static variable.
5992 (linux_cpu_translation_table): Likewise.
5993 (elf_platform) <cpu>: Define new static variable and use it.
5994 Translate kernel AT_PLATFORM name to canonical name if needed.
5995 Error if platform name is unknown.
5997 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6000 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6002 2018-02-01 Jeff Law <law@redhat.com>
6005 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6006 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6007 the scratch if RELEASE_VIA_POP is false.
6008 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6009 If we have to save a temporary register, decrement SIZE appropriately.
6010 Pass new arguments to release_scratch_register_on_entry.
6011 (ix86_adjust_stack_and_probe): Likewise.
6012 (ix86_emit_probe_stack_range): Pass new arguments to
6013 release_scratch_register_on_entry.
6015 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6017 PR rtl-optimization/84157
6018 * combine.c (change_zero_ext): Use REG_P predicate in
6019 front of HARD_REGISTER_P predicate.
6021 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6023 * config/avr/avr.c (avr_option_override): Move disabling of
6024 -fdelete-null-pointer-checks to...
6025 * common/config/avr/avr-common.c (avr_option_optimization_table):
6028 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6030 PR tree-optimization/81635
6031 * tree-data-ref.c (split_constant_offset_1): For types that
6032 wrap on overflow, try to use range info to prove that wrapping
6035 2018-02-01 Renlin Li <renlin.li@arm.com>
6038 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6040 (aarch64_register_move_cost): Likewise.
6041 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6043 (REG_CLASS_NAMES): Likewise.
6044 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6045 TAILCALL_ADDR_REGS. Remove IP registers.
6046 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6048 2018-02-01 Richard Biener <rguenther@suse.de>
6050 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6051 for specifying RPO order and allow NULL for that.
6052 * domwalk.c (dom_walker::dom_walker): Likewise.
6053 (dom_walker::walk): Handle NULL RPO order.
6054 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6056 (rewrite_update_dom_walker): Likewise.
6057 (mark_def_dom_walker): Likewise.
6059 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6061 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6062 (aarch64_maybe_expand_sve_subreg_move): Declare.
6063 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6064 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6066 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6067 that are semantically a reverse operation.
6068 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6069 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6070 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6072 (aarch64_can_change_mode_class): For big-endian, forbid changes
6073 between two SVE modes if they have different element sizes.
6075 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6077 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6078 the TImode handling for big-endian targets.
6080 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6082 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6083 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6085 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6086 Remove BSWAP handing for big-endian targets and use the form of
6087 LD1RQ appropariate for the mode.
6089 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6091 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6092 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6095 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6098 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6099 check for operands that need to go through aarch64_sve_reload_be.
6101 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6103 PR tree-optimization/81661
6104 PR tree-optimization/84117
6105 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6106 * tree-eh.c: Include gimplify.h.
6107 (find_trapping_overflow, replace_trapping_overflow,
6108 rewrite_to_non_trapping_overflow): New functions.
6109 * tree-vect-loop.c: Include tree-eh.h.
6110 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6111 * tree-data-ref.c: Include tree-eh.h.
6112 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6114 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6116 PR rtl-optimization/84123
6117 * combine.c (change_zero_ext): Check if hard register satisfies
6118 can_change_dest_mode before calling gen_lowpart_SUBREG.
6120 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6123 * ira.c (ira_init_register_move_cost): Remove assert.
6125 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6127 PR rtl-optimization/84071
6128 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6129 * doc/tm.texi: Regenerate.
6131 2018-01-31 Richard Biener <rguenther@suse.de>
6133 PR tree-optimization/84132
6134 * tree-data-ref.c (analyze_miv_subscript): Properly
6135 check whether evolution_function_is_affine_multivariate_p
6136 before calling gcd_of_steps_may_divide_p.
6138 2018-01-31 Julia Koval <julia.koval@intel.com>
6141 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6142 * config/i386/i386.md (rdpid_rex64) New.
6143 (rdpid): Make 32bit only.
6145 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6148 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6149 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6151 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6154 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6156 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6158 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6160 PR rtl-optimization/84071
6161 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6162 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6164 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6166 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6167 (arc_attribute_table): Add 'aux' attribute.
6168 (arc_in_small_data_p): Consider aux like variables.
6169 (arc_is_aux_reg_p): New function.
6170 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6171 (arc_get_aux_arg): New function.
6172 (prepare_move_operands): Handle aux-register access.
6173 (arc_handle_aux_attribute): New function.
6174 * doc/extend.texi (ARC Variable attributes): Add subsection.
6176 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6178 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6179 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6180 (arc_attribute_table): Add 'uncached' attribute.
6181 (arc_print_operand): Print '.di' flag for uncached memory
6183 (arc_in_small_data_p): Do not consider for small data the uncached
6185 (arc_is_uncached_mem_p): New function.
6186 * config/arc/predicates.md (compact_store_memory_operand): Check
6187 for uncached memory accesses.
6188 (nonvol_nonimm_operand): Likewise.
6189 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6191 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6194 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6195 falign-loops=): Add Optimization flag.
6197 2018-01-30 Jeff Law <law@redhat.com>
6200 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6201 INT_REGISTERS_SAVED. Check it prior to calling
6202 get_scratch_register_on_entry.
6203 (ix86_adjust_stack_and_probe): Similarly.
6204 (ix86_emit_probe_stack_range): Similarly.
6205 (ix86_expand_prologue): Corresponding changes.
6207 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6210 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6211 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6213 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6216 * lra-constraints.c (curr_insn_transform): Process AND in the
6219 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6221 PR rtl-optimization/83986
6222 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6223 dependence against last_pending_memory_flush in addition to
6226 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6228 PR tree-optimization/81611
6229 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6232 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6235 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6238 2018-01-30 Richard Biener <rguenther@suse.de>
6239 Jakub Jelinek <jakub@redhat.com>
6241 PR tree-optimization/84111
6242 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6243 inner loops added during recursion, as they don't have up-to-date
6246 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6249 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6250 (can_inline_edge_by_limits_p): ... here.
6251 (can_early_inline_edge_p, check_callers,
6252 update_caller_keys, update_callee_keys, recursive_inlining,
6253 add_new_edges_to_heap, speculation_useful_p,
6254 inline_small_functions,
6255 inline_small_functions, flatten_function,
6256 inline_to_all_callers_1): Update.
6258 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6260 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6263 2018-01-30 Richard Biener <rguenther@suse.de>
6265 PR tree-optimization/83008
6266 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6267 invariant and constant vector uses in stmts when they need
6270 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6273 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6274 * configure: Regenerate.
6276 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6278 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6280 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6281 Use gen_rtx_REG rather than gen_lowpart.
6283 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6285 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6286 rather than 0 when creating partial subregs.
6288 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6290 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6293 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6296 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6297 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6298 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6299 flags. This restores the settings used before the 2017-07-24.
6300 Turning off pre increment/decrement/modify allows IVOPTS to
6301 optimize DF/SF loops where the index is an int.
6303 2018-01-29 Richard Biener <rguenther@suse.de>
6304 Kelvin Nilsen <kelvin@gcc.gnu.org>
6307 * tree-vect-stmts.c (vectorizable_call): Don't call
6308 targetm.vectorize_builtin_md_vectorized_function if callee is
6311 2018-01-22 Carl Love <cel@us.ibm.com>
6313 * doc/extend.tex: Fix typo in second arg in
6314 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6316 2018-01-29 Richard Biener <rguenther@suse.de>
6318 PR tree-optimization/84086
6319 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6320 (flush_ssaname_freelist): When SSA names were released reset
6321 the SCEV hash table.
6323 2018-01-29 Richard Biener <rguenther@suse.de>
6325 PR tree-optimization/84057
6326 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6327 removed paths when removing edges.
6329 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6331 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6332 -mfunction-return=@var{choice}.
6334 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6337 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6339 (layout::print_source_line): Likewise.
6340 (test_get_line_width_without_trailing_whitespace): Add test cases.
6342 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6345 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6348 2018-01-26 Jim Wilson <jimw@sifive.com>
6350 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6352 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6355 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6357 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6358 and CMP + SUB-immediate -> SUBS.
6360 2018-01-26 Martin Sebor <msebor@redhat.com>
6362 PR tree-optimization/83896
6363 * tree-ssa-strlen.c (get_string_len): Rename...
6364 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6365 Avoid assuming length is constant.
6366 (handle_char_store): Use HOST_WIDE_INT for string length.
6368 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6371 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6372 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6374 2018-01-26 Richard Biener <rguenther@suse.de>
6376 PR rtl-optimization/84003
6377 * dse.c (record_store): Only record redundant stores when
6378 the earlier store aliases at least all accesses the later one does.
6380 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6382 PR rtl-optimization/83985
6383 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6384 REG_CFA_RESTORE insns.
6385 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6386 REG_CFA_RESTORE insns here.
6389 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6390 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6392 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6394 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6395 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6396 (arc_init): Likewise.
6397 (arc_override_options): Likewise.
6398 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6400 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6402 * config/arc/arc.h (TARGET_DBNZ): Define.
6403 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6404 properly set the tune attribute.
6405 (dbnz): Use TARGET_DBNZ guard.
6406 * config/arc/arc.opt (mtune): Add core3 option.
6408 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6410 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6411 recognize new pic like addresses.
6412 (arc_delegitimize_address): Clean up.
6414 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6416 * config/arc/arc-arches.def: Option mrf16 valid for all
6418 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6419 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6420 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6421 * config/arc/arc-tables.opt: Regenerate.
6422 * config/arc/arc.c (arc_conditional_register_usage): Handle
6423 reduced register file case.
6424 (arc_file_start): Set must have build attributes.
6425 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6427 * config/arc/arc.opt (mrf16): Add new option.
6428 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6429 * config/arc/genmultilib.awk: Handle new mrf16 option.
6430 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6431 * config/arc/t-multilib: Regenerate.
6432 * doc/invoke.texi (ARC Options): Document mrf16 option.
6434 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6436 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6437 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6438 (arc_attribute_table): Add 'secure_call' attribute.
6439 (arc_print_operand): Print secure call operand.
6440 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6442 (arc_is_secure_call_p): New function. * config/arc/arc.md
6443 (call_i): Add support for sjli instruction.
6444 (call_value_i): Likewise.
6445 * config/arc/constraints.md (Csc): New constraint.
6447 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6448 John Eric Martin <John.Martin@emmicro-us.com>
6450 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6451 * config/arc/arc.c (_arc_jli_section): New struct.
6452 (arc_jli_section): New type.
6453 (rc_jli_sections): New static variable.
6454 (arc_handle_jli_attribute): New function.
6455 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6456 (arc_file_end): New function.
6457 (TARGET_ASM_FILE_END): Define.
6458 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6459 (arc_add_jli_section): New function.
6460 (jli_call_scan): Likewise.
6461 (arc_reorg): Call jli_call_scan.
6462 (arc_output_addsi): Remove 'S' from printing asm operand.
6463 (arc_is_jli_call_p): New function.
6464 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6466 (movhi_insn): Likewise.
6467 (movsi_insn): Likewise.
6468 (movsi_set_cc_insn): Likewise.
6469 (loadqi_update): Likewise.
6470 (load_zeroextendqisi_update): Likewise.
6471 (load_signextendqisi_update): Likewise.
6472 (loadhi_update): Likewise.
6473 (load_zeroextendhisi_update): Likewise.
6474 (load_signextendhisi_update): Likewise.
6475 (loadsi_update): Likewise.
6476 (loadsf_update): Likewise.
6477 (movsicc_insn): Likewise.
6478 (bset_insn): Likewise.
6479 (bxor_insn): Likewise.
6480 (bclr_insn): Likewise.
6481 (bmsk_insn): Likewise.
6482 (bicsi3_insn): Likewise.
6483 (cmpsi_cc_c_insn): Likewise.
6484 (movsi_ne): Likewise.
6485 (movsi_cond_exec): Likewise.
6486 (clrsbsi2): Likewise.
6497 (call_i): Remove 'S' asm letter, add jli instruction.
6498 (call_value_i): Likewise.
6499 * config/arc/arc.op (mjli-always): New option.
6500 * config/arc/constraints.md (Cji): New constraint.
6501 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6503 (subsf3_fpx): Likewise.
6504 (mulsf3_fpx): Likewise.
6505 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6507 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6509 * doc/invoke.texi (ARC): Document mjli-always option.
6511 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6513 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6514 avoid addition with 0 and use incw and decw where possible.
6516 2018-01-26 Richard Biener <rguenther@suse.de>
6518 PR tree-optimization/81082
6519 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6520 association if it requires casting to unsigned.
6521 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6522 from fold_plusminus_mult_expr to catch important cases late when
6523 range info is available.
6525 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6527 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6528 * configure.ac (hidden_linkonce): New test.
6529 * configure: Regenerate.
6530 * config.in: Regenerate.
6532 2018-01-26 Julia Koval <julia.koval@intel.com>
6534 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6535 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6536 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6537 _mm_mask_bitshuffle_epi64_mask): Fix type.
6538 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6539 USI_FTYPE_V4DI_V4DI_USI): Remove.
6540 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6541 __builtin_ia32_vpshufbitqmb256_mask,
6542 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6543 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6544 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6546 2018-01-26 Alan Modra <amodra@gmail.com>
6549 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6550 UNSPEC_VBPERMQ. Sort other unspecs.
6552 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6554 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6556 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6559 * predict.c (drop_profile): Do not push/pop cfun; update also
6561 (handle_missing_profiles): Fix logic looking for zero profiles.
6563 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6566 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6567 on functions with #pragma omp declare simd or functions with simd
6569 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6570 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6571 Remove trailing \n from warning_at calls.
6573 2018-01-25 Tom de Vries <tom@codesourcery.com>
6576 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6577 for neutered workers.
6579 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6582 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6583 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6585 2018-01-24 Jeff Law <law@redhat.com>
6588 * i386.c (get_probe_interval): Move to earlier point.
6589 (ix86_compute_frame_layout): If -fstack-clash-protection and
6590 the frame is larger than the probe interval, then use pushes
6591 to save registers rather than reg->mem moves.
6592 (ix86_expand_prologue): Remove conditional for int_registers_saved
6595 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6598 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6599 min/max for never referenced object.
6601 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6604 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6606 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6607 attributes from DECL_ATTRIBUTES (decl) without affecting
6608 DECL_ATTRIBUTES (current_function_decl).
6609 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6610 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6612 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6614 PR tree-optimization/83979
6615 * fold-const.c (fold_comparison): Use constant_boolean_node
6616 instead of boolean_{true,false}_node.
6618 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6620 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6623 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6625 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6626 Simplify the clause that sets the length attribute.
6627 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6628 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6629 clause that sets the length attribute.
6630 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6632 2018-01-24 Tom de Vries <tom@codesourcery.com>
6635 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6636 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6637 Add strict parameter.
6638 (prevent_branch_around_nothing): Insert dummy insn between branch to
6639 label and label with no ptx insn inbetween.
6640 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6642 2018-01-24 Tom de Vries <tom@codesourcery.com>
6645 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6646 for neutered threads in warp.
6647 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6649 2018-01-24 Richard Biener <rguenther@suse.de>
6651 PR tree-optimization/83176
6652 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6655 2018-01-24 Richard Biener <rguenther@suse.de>
6657 PR tree-optimization/82819
6658 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6659 code generating pluses that are no-ops in the target precision.
6661 2018-01-24 Richard Biener <rguenther@suse.de>
6664 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6666 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6668 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6669 to merge probabilities.
6670 * predict.c (probably_never_executed): Also mark as cold functions
6671 with global 0 profile and guessed local profile.
6672 * profile-count.c (profile_probability::combine_with_count): New
6674 * profile-count.h (profile_probability::operator*,
6675 profile_probability::operator*=, profile_probability::operator/,
6676 profile_probability::operator/=): Reduce precision to adjusted
6677 and set value to guessed on contradictory divisions.
6678 (profile_probability::combine_with_freq): Remove.
6679 (profile_probability::combine_wiht_count): Declare.
6680 (profile_count::force_nonzero):: Set to adjusted.
6681 (profile_count::probability_in):: Set quality to adjusted.
6682 * tree-ssa-tail-merge.c (replace_block_by): Use
6685 2018-01-23 Andrew Waterman <andrew@sifive.com>
6686 Jim Wilson <jimw@sifive.com>
6688 * config/riscv/riscv.c (riscv_stack_boundary): New.
6689 (riscv_option_override): Set riscv_stack_boundary. Handle
6690 riscv_preferred_stack_boundary_arg.
6691 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6692 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6693 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6694 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6695 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6696 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6698 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6701 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6702 of struct ix86_frame.
6703 (ix86_expand_epilogue): Likewise. Add a local variable for
6704 the reg_save_offset field in struct ix86_frame.
6706 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6708 PR tree-optimization/82604
6709 * tree-loop-distribution.c (enum partition_kind): New enum item
6710 PKIND_PARTIAL_MEMSET.
6711 (partition_builtin_p): Support above new enum item.
6712 (generate_code_for_partition): Ditto.
6713 (compute_access_range): Differentiate cases that equality can be
6714 proven at all loops, the innermost loops or no loops.
6715 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6716 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6717 (finalize_partitions, distribute_loop): Don't fuse partition of
6718 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6719 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6722 2018-01-23 Martin Liska <mliska@suse.cz>
6724 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6725 order to ignore the predictor.
6726 (PRED_POLYMORPHIC_CALL): Likewise.
6727 (PRED_RECURSIVE_CALL): Likewise.
6729 2018-01-23 Martin Liska <mliska@suse.cz>
6731 * tree-profile.c (tree_profiling): Print function header to
6732 aware reader which function we are working on.
6733 * value-prof.c (gimple_find_values_to_profile): Do not print
6734 not interesting value histograms.
6736 2018-01-23 Martin Liska <mliska@suse.cz>
6738 * profile-count.h (enum profile_quality): Add
6739 profile_uninitialized as the first value. Do not number values
6740 as they are zero based.
6741 (profile_count::verify): Update sanity check.
6742 (profile_probability::verify): Likewise.
6744 2018-01-23 Nathan Sidwell <nathan@acm.org>
6746 * doc/invoke.texi (ffor-scope): Deprecate.
6748 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6750 PR tree-optimization/83510
6751 * domwalk.c (set_all_edges_as_executable): New function.
6752 (dom_walker::dom_walker): Convert bool param
6753 "skip_unreachable_blocks" to enum reachability. Move setup of
6754 edge flags to set_all_edges_as_executable and only do it when
6755 reachability is REACHABLE_BLOCKS.
6756 * domwalk.h (enum dom_walker::reachability): New enum.
6757 (dom_walker::dom_walker): Convert bool param
6758 "skip_unreachable_blocks" to enum reachability.
6759 (set_all_edges_as_executable): New decl.
6760 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6761 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6763 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6764 but converting true to REACHABLE_BLOCKS.
6765 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6767 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6768 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6769 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6771 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6772 if check_all_array_refs will be called.
6774 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6776 * tree.c (selftest::test_location_wrappers): Add more test
6779 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6781 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6782 (selftest::test_bit_in_range): Likewise.
6784 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6787 * doc/sourcebuild.texi (vect_float): Say that the selector
6788 only describes the situation when -funsafe-math-optimizations is on.
6789 (vect_float_strict): Document.
6791 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6793 PR tree-optimization/83965
6794 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6795 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6796 instead of checking only for a reduction.
6797 (vect_recog_widen_sum_pattern): Likewise.
6799 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6801 * predict.c (probably_never_executed): Only use precise profile info.
6802 (compute_function_frequency): Skip after inlining hack since we now
6803 have quality checking.
6805 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6807 * profile-count.h (profile_probability::very_unlikely,
6808 profile_probability::unlikely, profile_probability::even): Set
6809 precision to guessed.
6811 2018-01-23 Richard Biener <rguenther@suse.de>
6813 PR tree-optimization/83963
6814 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6815 Properly terminate dominator walk when crossing the exit edge not
6816 when visiting its source block.
6818 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6821 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6822 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6824 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6826 PR tree-optimization/83957
6827 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6828 semicolon after for body surrounded by braces.
6830 PR tree-optimization/83081
6831 * profile-count.h (profile_probability::split): New method.
6832 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6833 Use profile_probability::split.
6834 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6835 when splitting a single conditional jump into 2.
6837 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6839 PR tree-optimization/69452
6840 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6843 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6845 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6846 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6847 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6849 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6851 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6852 * config/rl78/rl78.md (movdi): New define_expand.
6853 * config/rl78/rl78.c (rl78_split_movdi): New function.
6855 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6858 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6860 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6861 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6862 128-bit to produce an UNSPEC move to get the double word with the
6863 signbit and then a shift directly to do signbit.
6864 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6865 implementation with a new version that just does either a direct
6866 move or a regular move. Move memory interface to separate insns.
6867 Move insns so they are next to the expander.
6868 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6869 with signbit move. Split big and little endian case.
6870 (signbit<mode>2_dm_mem_le): Likewise.
6871 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6872 (signbit<mode>2_dm2): Likewise.
6874 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6876 * config/rl78/rl78.md (anddi3): New define_expand.
6878 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6880 * config/rl78/rl78.md (umindi3): New define_expand.
6882 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6884 * config/rl78/rl78.md (smindi3): New define_expand.
6886 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6888 * config/rl78/rl78.md (smaxdi3): New define_expand.
6890 2018-01-22 Carl Love <cel@us.ibm.com>
6892 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6893 LVX_V1TI): Add macro expansion.
6894 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6895 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6896 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6897 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6898 Change check to determine if the instruction is a byte reversing
6899 entry. Fix typo in comment.
6900 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6901 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6902 Add def_builtin calls for new builtins.
6903 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6904 Add define_insn expansion.
6906 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6908 * config/rl78/rl78.md (umaxdi3): New define_expand.
6910 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6912 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6913 for non-QImode registers.
6915 2018-01-22 Richard Biener <rguenther@suse.de>
6917 PR tree-optimization/83963
6918 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6919 including the loop exit block.
6920 (scop_detection::merge_sese): Likewise.
6921 (scop_detection::add_scop): Do it here instead.
6923 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6925 * doc/sourcebuild.texi (arm_softfloat): Document.
6927 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6930 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6931 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6932 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6934 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6935 David Edelsohn <dje.gcc@gmail.com>
6938 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6939 Change "crset eq" to "crset 2".
6940 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6941 (*call_indirect_aix<mode>_nospec): Likewise.
6942 (*call_value_indirect_aix<mode>_nospec): Likewise.
6943 (*call_indirect_elfv2<mode>_nospec): Likewise.
6944 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6945 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6946 change assembly output from . to $.
6947 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6948 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6949 (*tablejump<mode>_internal1_nospec): Likewise.
6951 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6954 * config/sh/sh_optimize_sett_clrt.cc:
6955 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6957 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6959 PR tree-optimization/83940
6960 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6961 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6962 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6963 use it to pass back the definition type.
6964 (vect_check_store_rhs): Likewise rhs_dt_out.
6965 (vect_build_gather_load_calls): Add a mask_dt argument and use
6966 it instead of a call to vect_is_simple_use.
6967 (vectorizable_store): Update calls to vect_check_load_store_mask
6968 and vect_check_store_rhs. Use the dt returned by the latter instead
6969 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6970 instead of calls to vect_is_simple_use. Pass the scalar rather
6971 than the vector operand to vect_is_simple_use when handling
6972 second and subsequent copies of an rhs value.
6973 (vectorizable_load): Update calls to vect_check_load_store_mask
6974 and vect_build_gather_load_calls. Use the cached mask_dt and
6975 gs_info.offset_dt instead of calls to vect_is_simple_use.
6977 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6980 * tree-emutls.c: Include gimplify.h.
6981 (lower_emutls_2): New function.
6982 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6983 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6984 it before further processing.
6987 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6988 UINTVAL (trueop1) instead of INTVAL (op1).
6990 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6994 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6995 INCOMING_FRAME_SP_OFFSET if not defined.
6996 (scan_trace): Add ENTRY argument. If true and
6997 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6998 emit a note to adjust the CFA offset.
6999 (create_cfi_notes): Adjust scan_trace callers.
7000 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7001 INCOMING_FRAME_SP_OFFSET in the CIE.
7002 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7003 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7005 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7006 * doc/tm.texi: Regenerated.
7008 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7010 PR rtl-optimization/83147
7011 * lra-constraints.c (remove_inheritance_pseudos): Use
7012 lra_substitute_pseudo_within_insn.
7014 2018-01-19 Tom de Vries <tom@codesourcery.com>
7015 Cesar Philippidis <cesar@codesourcery.com>
7018 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7020 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7023 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7024 spaces for function labels.
7026 2018-01-19 Martin Liska <mliska@suse.cz>
7028 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7029 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7030 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7031 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7032 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7033 (PRED_CONST_RETURN): Change from 69 to 65.
7034 (PRED_NULL_RETURN): Change from 91 to 71.
7035 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7036 (PRED_LOOP_GUARD): Change from 66 to 73.
7038 2018-01-19 Martin Liska <mliska@suse.cz>
7040 * predict.c (predict_insn_def): Add new assert.
7041 (struct branch_predictor): Change type to signed integer.
7042 (test_prediction_value_range): Amend test to cover
7044 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7045 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7046 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7047 (PRED_LOOP_IV_COMPARE): Likewise.
7048 * predict.h (PROB_UNINITIALIZED): Define new constant.
7050 2018-01-19 Martin Liska <mliska@suse.cz>
7052 * predict.c (dump_prediction): Add new format for
7053 analyze_brprob.py script which is enabled with -details
7055 * profile-count.h (precise_p): New function.
7057 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7059 PR tree-optimization/83922
7060 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7061 there are no statements that need masking.
7062 (vect_active_double_reduction_p): New function.
7063 (vect_analyze_loop_operations): Use it when handling phis that
7064 are not in the loop header.
7066 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7068 PR tree-optimization/83914
7069 * tree-vect-loop.c (vectorizable_induction): Don't convert
7070 init_expr or apply the peeling adjustment for inductions
7071 that are nested within the vectorized loop.
7073 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7075 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7078 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7082 * function.h (gimplify_parameters): Add gimple_seq * argument.
7083 * function.c: Include gimple.h and options.h.
7084 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7085 for the added local temporaries if needed.
7086 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7087 if there are any parameter cleanups, wrap whole body into a
7088 try/finally with the cleanups.
7090 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7093 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7094 Use GET_MODE_CLASS for scalar floating point.
7096 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7100 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7101 Fix call of call_cgraph_insertion_hooks.
7103 2018-01-18 Martin Sebor <msebor@redhat.com>
7105 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7107 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7110 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7113 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7116 * common.opt: (-ffile-prefix-map): New option.
7117 * opts.c (common_handle_option): Defer it.
7118 * opts-global.c (handle_common_deferred_options): Handle it.
7119 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7120 * file-prefix-map.h: New file.
7121 (remap_debug_filename, add_debug_prefix_map): ...here.
7122 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7123 * final.c (debug_prefix_map, add_debug_prefix_map
7124 remap_debug_filename): Move to...
7125 * file-prefix-map.c: New file.
7126 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7127 generalize, get rid of alloca(), use strrchr() instead of strchr().
7128 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7129 Implement in terms of add_prefix_map().
7130 (remap_macro_filename, remap_debug_filename): Implement in term of
7132 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7133 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7134 * dbxout.c: Include file-prefix-map.h.
7135 * varasm.c: Likewise.
7136 * vmsdbgout.c: Likewise.
7137 * xcoffout.c: Likewise.
7138 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7139 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7140 * doc/invoke.texi (-ffile-prefix-map): Document.
7141 (-fdebug-prefix-map): Update description.
7143 2018-01-18 Martin Liska <mliska@suse.cz>
7145 * config/i386/i386.c (indirect_thunk_name): Document that also
7147 (output_indirect_thunk): Document why both instructions
7148 (pause and lfence) are generated.
7150 2018-01-18 Richard Biener <rguenther@suse.de>
7152 PR tree-optimization/83887
7153 * graphite-scop-detection.c
7154 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7155 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7156 (scop_detection::merge_sese): Re-implement with a flood-fill
7157 algorithm that properly finds a SESE region if it exists.
7159 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7162 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7163 pointer_diff optimizations use view_convert instead of convert.
7165 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7167 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7168 Generate different code for -mno-speculate-indirect-jumps.
7169 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7170 (*call_indirect_aix<mode>): Disable for
7171 -mno-speculate-indirect-jumps.
7172 (*call_indirect_aix<mode>_nospec): New define_insn.
7173 (*call_value_indirect_aix<mode>): Disable for
7174 -mno-speculate-indirect-jumps.
7175 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7176 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7177 -mno-speculate-indirect-jumps.
7178 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7180 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7182 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7183 long double type, set the flags for noting the default long double
7184 type, even if we don't pass or return a long double type.
7186 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7189 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7192 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7194 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7195 support for merge[hl].
7196 (fold_mergehl_helper): New helper function.
7197 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7198 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7199 (altivec_vmrglw_direct): Add xxmrglw insn.
7201 2018-01-17 Andrew Waterman <andrew@sifive.com>
7203 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7204 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7206 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7209 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7210 call the lto_location_cache before reading the
7211 DECL_SOURCE_LOCATION of the types.
7213 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7214 Richard Sandiford <richard.sandiford@linaro.org>
7216 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7217 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7218 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7219 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7220 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7222 * config/aarch64/constraints.md (aarch64_movti_operand):
7224 * config/aarch64/predicates.md (Uti): Add new constraint.
7226 2018-01-17 Carl Love <cel@us.ibm.com>
7228 * config/rs6000/vsx.md (define_expand xl_len_r,
7229 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7230 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7232 (define_expand, define_insn): Move the shift left from the
7233 define_insn to the define_expand for lxvl and stxvl instructions.
7234 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7235 and XL_LEN_R definitions to PURE.
7237 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7239 * config/i386/i386.c (indirect_thunk_name): Declare regno
7240 as unsigned int. Compare regno with INVALID_REGNUM.
7241 (output_indirect_thunk): Ditto.
7242 (output_indirect_thunk_function): Ditto.
7243 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7244 in the call to output_indirect_thunk_function.
7246 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7249 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7250 rather than the size of inner_type to determine the stack slot size
7251 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7253 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7256 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7259 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7261 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7262 endian Linux systems to optionally enable multilibs for selecting
7263 the long double type if the user configured an explicit type.
7264 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7265 have no long double multilibs if not defined.
7266 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7267 warn if the user used -mabi={ieee,ibm}longdouble and we built
7268 multilibs for long double.
7269 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7270 appropriate multilib option.
7271 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7273 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7274 for building long double multilibs.
7275 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7277 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7279 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7282 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7284 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7287 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7290 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7293 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7295 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7296 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7298 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7300 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7301 different rtl trees depending on TARGET_64BIT.
7302 (rs6000_gen_lvx): Likewise.
7304 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7306 * config/visium/visium.md (nop): Tweak comment.
7307 (hazard_nop): Likewise.
7309 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7311 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7312 -mspeculate-indirect-jumps.
7313 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7314 for -mno-speculate-indirect-jumps.
7315 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7316 (*call_value_indirect_elfv2<mode>): Disable for
7317 -mno-speculate-indirect-jumps.
7318 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7319 (indirect_jump): Emit different RTL for
7320 -mno-speculate-indirect-jumps.
7321 (*indirect_jump<mode>): Disable for
7322 -mno-speculate-indirect-jumps.
7323 (*indirect_jump<mode>_nospec): New define_insn.
7324 (tablejump): Emit different RTL for
7325 -mno-speculate-indirect-jumps.
7326 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7327 (tablejumpsi_nospec): New define_expand.
7328 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7329 (tablejumpdi_nospec): New define_expand.
7330 (*tablejump<mode>_internal1): Disable for
7331 -mno-speculate-indirect-jumps.
7332 (*tablejump<mode>_internal1_nospec): New define_insn.
7333 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7336 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7338 * caller-save.c (insert_save): Drop unnecessary parameter. All
7341 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7342 Richard Biener <rguenth@suse.de>
7345 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7346 return early, inline manually is_gimple_sizepos. Make sure if we
7347 call gimplify_expr we don't end up with a gimple constant.
7348 * tree.c (variably_modified_type_p): Don't return true for
7349 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7350 * gimplify.h (is_gimple_sizepos): Remove.
7352 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7354 PR tree-optimization/83857
7355 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7356 vectorizable_live_operation for pure SLP statements.
7357 (vectorizable_live_operation): Handle PHIs.
7359 2018-01-16 Richard Biener <rguenther@suse.de>
7361 PR tree-optimization/83867
7362 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7363 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7365 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7368 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7369 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7370 If off is not INTEGER_CST, issue a may not be aligned warning
7371 rather than isn't aligned. Use isn%'t rather than isn't.
7372 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7374 <case MULT_EXPR>: Improve the case when bottom and one of the
7375 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7376 operand, in that case check if the other operand is multiple of
7377 bottom divided by the INTEGER_CST operand.
7379 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7382 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7383 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7384 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7385 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7386 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7387 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7388 (pa_function_arg_size): New function.
7390 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7392 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7393 in a separate statement.
7395 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7397 PR tree-optimization/83847
7398 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7399 group gathers and scatters.
7401 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7403 PR rtl-optimization/86620
7404 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7406 PR rtl-optimization/83213
7407 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7408 to last if both are JUMP_INSNs.
7410 PR tree-optimization/83843
7411 * gimple-ssa-store-merging.c
7412 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7413 store_immediate_info for bswap/nop orig_stores.
7415 2018-01-15 Andrew Waterman <andrew@sifive.com>
7417 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7419 <UDIV>: Increase cost if !TARGET_DIV.
7421 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7423 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7424 (define_attr "cr_logical_3op"): New.
7425 (cceq_ior_compare): Adjust.
7426 (cceq_ior_compare_complement): Adjust.
7427 (*cceq_rev_compare): Adjust.
7428 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7429 (is_cracked_insn): Adjust.
7430 (insn_must_be_first_in_group): Adjust.
7431 * config/rs6000/40x.md: Adjust.
7432 * config/rs6000/440.md: Adjust.
7433 * config/rs6000/476.md: Adjust.
7434 * config/rs6000/601.md: Adjust.
7435 * config/rs6000/603.md: Adjust.
7436 * config/rs6000/6xx.md: Adjust.
7437 * config/rs6000/7450.md: Adjust.
7438 * config/rs6000/7xx.md: Adjust.
7439 * config/rs6000/8540.md: Adjust.
7440 * config/rs6000/cell.md: Adjust.
7441 * config/rs6000/e300c2c3.md: Adjust.
7442 * config/rs6000/e500mc.md: Adjust.
7443 * config/rs6000/e500mc64.md: Adjust.
7444 * config/rs6000/e5500.md: Adjust.
7445 * config/rs6000/e6500.md: Adjust.
7446 * config/rs6000/mpc.md: Adjust.
7447 * config/rs6000/power4.md: Adjust.
7448 * config/rs6000/power5.md: Adjust.
7449 * config/rs6000/power6.md: Adjust.
7450 * config/rs6000/power7.md: Adjust.
7451 * config/rs6000/power8.md: Adjust.
7452 * config/rs6000/power9.md: Adjust.
7453 * config/rs6000/rs64.md: Adjust.
7454 * config/rs6000/titan.md: Adjust.
7456 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7458 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7459 ix86_indirect_branch_register logic.
7461 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7463 * config/i386/constraints.md (Bs): Update
7464 ix86_indirect_branch_register check. Don't check
7465 ix86_indirect_branch_register with GOT_memory_operand.
7467 * config/i386/predicates.md (GOT_memory_operand): Don't check
7468 ix86_indirect_branch_register here.
7469 (GOT32_symbol_operand): Likewise.
7471 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7473 * config/i386/predicates.md (constant_call_address_operand):
7474 Rewrite ix86_indirect_branch_register logic.
7475 (sibcall_insn_operand): Likewise.
7477 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7479 * config/i386/constraints.md (Bs): Replace
7480 ix86_indirect_branch_thunk_register with
7481 ix86_indirect_branch_register.
7483 * config/i386/i386.md (indirect_jump): Likewise.
7484 (tablejump): Likewise.
7485 (*sibcall_memory): Likewise.
7486 (*sibcall_value_memory): Likewise.
7487 Peepholes of indirect call and jump via memory: Likewise.
7488 * config/i386/i386.opt: Likewise.
7489 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7490 (GOT_memory_operand): Likewise.
7491 (call_insn_operand): Likewise.
7492 (sibcall_insn_operand): Likewise.
7493 (GOT32_symbol_operand): Likewise.
7495 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7498 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7499 type rather than type addr's type points to.
7500 (expand_omp_atomic_mutex): Likewise.
7501 (expand_omp_atomic): Likewise.
7503 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7506 * config/i386/i386.c (output_indirect_thunk_function): Use
7507 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7508 for __x86_return_thunk.
7510 2018-01-15 Richard Biener <rguenther@suse.de>
7513 * expmed.c (extract_bit_field_1): Fix typo.
7515 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7518 * config/arm/iterators.md (VF): New mode iterator.
7519 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7520 Remove integer-related logic from pattern.
7521 (neon_vabd<mode>_3): Likewise.
7523 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7526 * common.opt (fstrict-overflow): No longer an alias.
7527 (fwrapv-pointer): New option.
7528 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7529 also for pointer types based on flag_wrapv_pointer.
7530 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7531 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7532 opts->x_flag_wrapv got set.
7533 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7534 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7535 POINTER_TYPE_OVERFLOW_UNDEFINED.
7536 * match.pd: Likewise in address comparison pattern.
7537 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7539 2018-01-15 Richard Biener <rguenther@suse.de>
7542 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7543 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7544 Reset type names to their identifier if their TYPE_DECL doesn't
7545 have linkage (and thus is used for ODR and devirt).
7546 (save_debug_info_for_decl): Remove.
7547 (save_debug_info_for_type): Likewise.
7548 (add_tree_to_fld_list): Adjust.
7549 * tree-pretty-print.c (dump_generic_node): Make dumping of
7550 type names more robust.
7552 2018-01-15 Richard Biener <rguenther@suse.de>
7554 * BASE-VER: Bump to 8.0.1.
7556 2018-01-14 Martin Sebor <msebor@redhat.com>
7559 * builtins.c (check_access): Avoid warning when the no-warning bit
7562 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7564 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7565 * ira-color (allocno_hard_regs_compare): Likewise.
7567 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7570 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7571 Use .pushsection/.popsection.
7573 2018-01-14 Martin Sebor <msebor@redhat.com>
7576 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7578 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7580 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7581 entry from extra_headers.
7582 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7583 extra_headers, make the list bitwise identical to the i?86-*-* one.
7585 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7587 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7588 -mcmodel=large with -mindirect-branch=thunk,
7589 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7590 -mfunction-return=thunk-extern.
7591 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7592 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7593 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7595 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7597 * config/i386/i386.c (print_reg): Print the name of the full
7598 integer register without '%'.
7599 (ix86_print_operand): Handle 'V'.
7600 * doc/extend.texi: Document 'V' modifier.
7602 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7604 * config/i386/constraints.md (Bs): Disallow memory operand for
7605 -mindirect-branch-register.
7607 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7608 (GOT_memory_operand): Likewise.
7609 (call_insn_operand): Likewise.
7610 (sibcall_insn_operand): Likewise.
7611 (GOT32_symbol_operand): Likewise.
7612 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7613 for -mindirect-branch-register.
7614 (tablejump): Likewise.
7615 (*sibcall_memory): Likewise.
7616 (*sibcall_value_memory): Likewise.
7617 Disallow peepholes of indirect call and jump via memory for
7618 -mindirect-branch-register.
7619 (*call_pop): Replace m with Bw.
7620 (*call_value_pop): Likewise.
7621 (*sibcall_pop_memory): Replace m with Bs.
7622 * config/i386/i386.opt (mindirect-branch-register): New option.
7623 * doc/invoke.texi: Document -mindirect-branch-register option.
7625 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7627 * config/i386/i386-protos.h (ix86_output_function_return): New.
7628 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7629 set function_return_type.
7630 (indirect_thunk_name): Add ret_p to indicate thunk for function
7632 (output_indirect_thunk_function): Pass false to
7633 indirect_thunk_name.
7634 (ix86_output_indirect_branch_via_reg): Likewise.
7635 (ix86_output_indirect_branch_via_push): Likewise.
7636 (output_indirect_thunk_function): Create alias for function
7637 return thunk if regno < 0.
7638 (ix86_output_function_return): New function.
7639 (ix86_handle_fndecl_attribute): Handle function_return.
7640 (ix86_attribute_table): Add function_return.
7641 * config/i386/i386.h (machine_function): Add
7642 function_return_type.
7643 * config/i386/i386.md (simple_return_internal): Use
7644 ix86_output_function_return.
7645 (simple_return_internal_long): Likewise.
7646 * config/i386/i386.opt (mfunction-return=): New option.
7647 (indirect_branch): Mention -mfunction-return=.
7648 * doc/extend.texi: Document function_return function attribute.
7649 * doc/invoke.texi: Document -mfunction-return= option.
7651 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7653 * config/i386/i386-opts.h (indirect_branch): New.
7654 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7655 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7656 with local indirect jump when converting indirect call and jump.
7657 (ix86_set_indirect_branch_type): New.
7658 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7659 (indirectlabelno): New.
7660 (indirect_thunk_needed): Likewise.
7661 (indirect_thunk_bnd_needed): Likewise.
7662 (indirect_thunks_used): Likewise.
7663 (indirect_thunks_bnd_used): Likewise.
7664 (INDIRECT_LABEL): Likewise.
7665 (indirect_thunk_name): Likewise.
7666 (output_indirect_thunk): Likewise.
7667 (output_indirect_thunk_function): Likewise.
7668 (ix86_output_indirect_branch_via_reg): Likewise.
7669 (ix86_output_indirect_branch_via_push): Likewise.
7670 (ix86_output_indirect_branch): Likewise.
7671 (ix86_output_indirect_jmp): Likewise.
7672 (ix86_code_end): Call output_indirect_thunk_function if needed.
7673 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7675 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7676 (ix86_attribute_table): Add indirect_branch.
7677 * config/i386/i386.h (machine_function): Add indirect_branch_type
7678 and has_local_indirect_jump.
7679 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7681 (tablejump): Likewise.
7682 (*indirect_jump): Use ix86_output_indirect_jmp.
7683 (*tablejump_1): Likewise.
7684 (simple_return_indirect_internal): Likewise.
7685 * config/i386/i386.opt (mindirect-branch=): New option.
7686 (indirect_branch): New.
7689 (thunk-inline): Likewise.
7690 (thunk-extern): Likewise.
7691 * doc/extend.texi: Document indirect_branch function attribute.
7692 * doc/invoke.texi: Document -mindirect-branch= option.
7694 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7697 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7699 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7701 * ipa-inline.c (want_inline_small_function_p): Return false if
7702 inlining has already failed with CIF_FINAL_ERROR.
7703 (update_caller_keys): Call want_inline_small_function_p before
7705 (update_callee_keys): Likewise.
7707 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7709 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7711 (rs6000_quadword_masked_address_p): Likewise.
7712 (quad_aligned_load_p): Likewise.
7713 (quad_aligned_store_p): Likewise.
7714 (const_load_sequence_p): Add comment to describe the outer-most loop.
7715 (mimic_memory_attributes_and_flags): New function.
7716 (rs6000_gen_stvx): Likewise.
7717 (replace_swapped_aligned_store): Likewise.
7718 (rs6000_gen_lvx): Likewise.
7719 (replace_swapped_aligned_load): Likewise.
7720 (replace_swapped_load_constant): Capitalize argument name in
7721 comment describing this function.
7722 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7723 and stores that access quad-word aligned addresses and replace
7724 with stvx or lvx instructions when appropriate.
7725 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7726 New function prototype.
7727 (rs6000_quadword_masked_address_p): Likewise.
7728 (rs6000_gen_lvx): Likewise.
7729 (rs6000_gen_stvx): Likewise.
7730 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7731 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7732 when memory address is aligned.
7733 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7734 this split to select lvx instruction when memory address is aligned.
7735 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7736 instruction when memory address is aligned.
7737 (*vsx_le_perm_load_v16qi): Likewise.
7738 (four unnamed splitters): Modify to select the stvx instruction
7739 when memory is aligned.
7741 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7743 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7744 which appears in the queue multiple times.
7746 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7747 Alan Hayward <alan.hayward@arm.com>
7748 David Sherwood <david.sherwood@arm.com>
7750 * tree-vectorizer.h (vec_lower_bound): New structure.
7751 (_loop_vec_info): Add check_nonzero and lower_bounds.
7752 (LOOP_VINFO_CHECK_NONZERO): New macro.
7753 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7754 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7755 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7756 fields. Make seg_len the distance travelled, not including the
7758 (dr_direction_indicator): Declare.
7759 (dr_zero_step_indicator): Likewise.
7760 (dr_known_forward_stride_p): Likewise.
7761 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7763 (runtime_alias_check_p): Allow runtime alias checks with
7765 (operator ==): Compare access_size and align.
7766 (prune_runtime_alias_test_list): Rework for new distinction between
7767 the access_size and seg_len.
7768 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7770 (get_segment_min_max): New function.
7771 (create_intersect_range_checks): Use it.
7772 (dr_step_indicator): New function.
7773 (dr_direction_indicator): Likewise.
7774 (dr_zero_step_indicator): Likewise.
7775 (dr_known_forward_stride_p): Likewise.
7776 * tree-loop-distribution.c (data_ref_segment_size): Return
7777 DR_STEP * (niters - 1).
7778 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7780 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7781 (vect_preserves_scalar_order_p): New function, split out from...
7782 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7783 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7784 (vect_vfa_access_size): New function.
7785 (vect_vfa_align): Likewise.
7786 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7787 (dump_lower_bound): New function.
7788 (vect_check_lower_bound): Likewise.
7789 (vect_small_gap_p): Likewise.
7790 (vectorizable_with_step_bound_p): Likewise.
7791 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7792 depencies if the vectorization factor is 1. Convert the checks
7793 for nonzero steps into checks on the bounds of DR_STEP. Try using
7794 a bunds check for variable steps if the minimum required step is
7795 relatively small. Update calls to the dr_with_seg_len
7796 constructor and to vect_compile_time_alias.
7797 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7799 (vect_loop_versioning): Call it.
7800 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7802 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7804 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7805 Alan Hayward <alan.hayward@arm.com>
7806 David Sherwood <david.sherwood@arm.com>
7808 * doc/sourcebuild.texi (vect_scatter_store): Document.
7809 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7811 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7813 * genopinit.c (main): Add supports_vec_scatter_store and
7814 supports_vec_scatter_store_cached to target_optabs.
7815 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7816 IFN_MASK_SCATTER_STORE.
7817 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7819 * internal-fn.h (internal_store_fn_p): Declare.
7820 (internal_fn_stored_value_index): Likewise.
7821 * internal-fn.c (scatter_store_direct): New macro.
7822 (expand_scatter_store_optab_fn): New function.
7823 (direct_scatter_store_optab_supported_p): New macro.
7824 (internal_store_fn_p): New function.
7825 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7826 IFN_MASK_SCATTER_STORE.
7827 (internal_fn_mask_index): Likewise.
7828 (internal_fn_stored_value_index): New function.
7829 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7831 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7832 * optabs-query.c (supports_vec_scatter_store_p): New function.
7833 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7834 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7835 true for scatter stores.
7836 (vect_gather_scatter_fn_p): Handle scatter stores too.
7837 (vect_check_gather_scatter): Consider using scatter stores if
7838 supports_vec_scatter_store_p.
7839 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7841 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7842 internal_fn_stored_value_index.
7843 (check_load_store_masking): Handle scatter stores too.
7844 (vect_get_store_rhs): Make public.
7845 (vectorizable_call): Use internal_store_fn_p.
7846 (vectorizable_store): Handle scatter store internal functions.
7847 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7848 when deciding whether the end of the group has been reached.
7849 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7850 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7851 (mask_scatter_store<mode>): New insns.
7853 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7854 Alan Hayward <alan.hayward@arm.com>
7855 David Sherwood <david.sherwood@arm.com>
7857 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7858 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7859 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7861 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7862 Use vect_truncate_gather_scatter_offset if we can't treat the
7863 operation as a normal gather load or scatter store.
7864 (get_group_load_store_type): Take the gather_scatter_info
7865 as argument. Try using a gather load or scatter store for
7866 single-element groups.
7867 (get_load_store_type): Update calls to get_group_load_store_type
7868 and vect_use_strided_gather_scatters_p.
7870 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7871 Alan Hayward <alan.hayward@arm.com>
7872 David Sherwood <david.sherwood@arm.com>
7874 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7875 optional tree argument.
7876 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7878 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7879 but continue to use the current value as a fallback.
7880 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7881 to compare the updates.
7882 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7883 (get_load_store_type): Use it when handling a strided access.
7884 (vect_get_strided_load_store_ops): New function.
7885 (vect_get_data_ptr_increment): Likewise.
7886 (vectorizable_load): Handle strided gather loads. Always pass
7887 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7889 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7890 Alan Hayward <alan.hayward@arm.com>
7891 David Sherwood <david.sherwood@arm.com>
7893 * doc/md.texi (gather_load@var{m}): Document.
7894 (mask_gather_load@var{m}): Likewise.
7895 * genopinit.c (main): Add supports_vec_gather_load and
7896 supports_vec_gather_load_cached to target_optabs.
7897 * optabs-tree.c (init_tree_optimization_optabs): Use
7898 ggc_cleared_alloc to allocate target_optabs.
7899 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7900 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7902 * internal-fn.h (internal_load_fn_p): Declare.
7903 (internal_gather_scatter_fn_p): Likewise.
7904 (internal_fn_mask_index): Likewise.
7905 (internal_gather_scatter_fn_supported_p): Likewise.
7906 * internal-fn.c (gather_load_direct): New macro.
7907 (expand_gather_load_optab_fn): New function.
7908 (direct_gather_load_optab_supported_p): New macro.
7909 (direct_internal_fn_optab): New function.
7910 (internal_load_fn_p): Likewise.
7911 (internal_gather_scatter_fn_p): Likewise.
7912 (internal_fn_mask_index): Likewise.
7913 (internal_gather_scatter_fn_supported_p): Likewise.
7914 * optabs-query.c (supports_at_least_one_mode_p): New function.
7915 (supports_vec_gather_load_p): Likewise.
7916 * optabs-query.h (supports_vec_gather_load_p): Declare.
7917 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7918 and memory_type field.
7919 (NUM_PATTERNS): Bump to 15.
7920 * tree-vect-data-refs.c: Include internal-fn.h.
7921 (vect_gather_scatter_fn_p): New function.
7922 (vect_describe_gather_scatter_call): Likewise.
7923 (vect_check_gather_scatter): Try using internal functions for
7924 gather loads. Recognize existing calls to a gather load function.
7925 (vect_analyze_data_refs): Consider using gather loads if
7926 supports_vec_gather_load_p.
7927 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7928 (vect_get_gather_scatter_offset_type): Likewise.
7929 (vect_convert_mask_for_vectype): Likewise.
7930 (vect_add_conversion_to_patterm): Likewise.
7931 (vect_try_gather_scatter_pattern): Likewise.
7932 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7933 (vect_vect_recog_func_ptrs): Add it.
7934 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7935 internal_fn_mask_index and internal_gather_scatter_fn_p.
7936 (check_load_store_masking): Take the gather_scatter_info as an
7937 argument and handle gather loads.
7938 (vect_get_gather_scatter_ops): New function.
7939 (vectorizable_call): Check internal_load_fn_p.
7940 (vectorizable_load): Likewise. Handle gather load internal
7942 (vectorizable_store): Update call to check_load_store_masking.
7943 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7944 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7945 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7946 (aarch64_gather_scale_operand_d): New predicates.
7947 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7948 (mask_gather_load<mode>): New insns.
7950 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7951 Alan Hayward <alan.hayward@arm.com>
7952 David Sherwood <david.sherwood@arm.com>
7954 * optabs.def (fold_left_plus_optab): New optab.
7955 * doc/md.texi (fold_left_plus_@var{m}): Document.
7956 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7957 * internal-fn.c (fold_left_direct): Define.
7958 (expand_fold_left_optab_fn): Likewise.
7959 (direct_fold_left_optab_supported_p): Likewise.
7960 * fold-const-call.c (fold_const_fold_left): New function.
7961 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7962 * tree-parloops.c (valid_reduction_p): New function.
7963 (gather_scalar_reductions): Use it.
7964 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7965 (vect_finish_replace_stmt): Declare.
7966 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7967 (needs_fold_left_reduction_p): New function, split out from...
7968 (vect_is_simple_reduction): ...here. Accept reductions that
7969 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7970 (vect_force_simple_reduction): Also store the reduction type in
7971 the assignment's STMT_VINFO_REDUC_TYPE.
7972 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7973 (merge_with_identity): New function.
7974 (vect_expand_fold_left): Likewise.
7975 (vectorize_fold_left_reduction): Likewise.
7976 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7977 scalar phi in place for it. Check for target support and reject
7978 cases that would reassociate the operation. Defer the transform
7979 phase to vectorize_fold_left_reduction.
7980 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7981 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7982 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7984 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7986 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7987 call to ifc_temp_var.
7989 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7990 Alan Hayward <alan.hayward@arm.com>
7991 David Sherwood <david.sherwood@arm.com>
7993 * target.def (legitimize_address_displacement): Take the original
7994 offset as a poly_int.
7995 * targhooks.h (default_legitimize_address_displacement): Update
7997 * targhooks.c (default_legitimize_address_displacement): Likewise.
7998 * doc/tm.texi: Regenerate.
7999 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8000 as an argument, moving assert of ad->disp == ad->disp_term to...
8001 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8002 Try calling targetm.legitimize_address_displacement before expanding
8003 the address rather than afterwards, and adjust for the new interface.
8004 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8005 Match the new hook interface. Handle SVE addresses.
8006 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8011 * Makefile.in (OBJS): Add early-remat.o.
8012 * target.def (select_early_remat_modes): New hook.
8013 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8014 * doc/tm.texi: Regenerate.
8015 * targhooks.h (default_select_early_remat_modes): Declare.
8016 * targhooks.c (default_select_early_remat_modes): New function.
8017 * timevar.def (TV_EARLY_REMAT): New timevar.
8018 * passes.def (pass_early_remat): New pass.
8019 * tree-pass.h (make_pass_early_remat): Declare.
8020 * early-remat.c: New file.
8021 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8023 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8025 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8026 Alan Hayward <alan.hayward@arm.com>
8027 David Sherwood <david.sherwood@arm.com>
8029 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8030 vfm1 with a bound_epilog parameter.
8031 (vect_do_peeling): Update calls accordingly, and move the prologue
8032 call earlier in the function. Treat the base bound_epilog as 0 for
8033 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8034 this base when peeling for gaps.
8035 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8036 with fully-masked loops.
8037 (vect_estimate_min_profitable_iters): Handle the single peeled
8038 iteration in that case.
8040 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8041 Alan Hayward <alan.hayward@arm.com>
8042 David Sherwood <david.sherwood@arm.com>
8044 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8045 single-element interleaving even if the size is not a power of 2.
8046 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8047 accesses for single-element interleaving if the group size is
8050 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8051 Alan Hayward <alan.hayward@arm.com>
8052 David Sherwood <david.sherwood@arm.com>
8054 * doc/md.texi (fold_extract_last_@var{m}): Document.
8055 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8056 * optabs.def (fold_extract_last_optab): New optab.
8057 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8058 * internal-fn.c (fold_extract_direct): New macro.
8059 (expand_fold_extract_optab_fn): Likewise.
8060 (direct_fold_extract_optab_supported_p): Likewise.
8061 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8062 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8063 EXTRACT_LAST_REDUCTION.
8064 (get_initial_def_for_reduction): Do not create an initial vector
8065 for EXTRACT_LAST_REDUCTION reductions.
8066 (vectorizable_reduction): Leave the scalar phi in place for
8067 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8068 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8069 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8070 transform phase to vectorizable_condition.
8071 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8073 (vect_finish_stmt_generation): ...here.
8074 (vect_finish_replace_stmt): New function.
8075 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8076 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8078 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8080 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8081 Alan Hayward <alan.hayward@arm.com>
8082 David Sherwood <david.sherwood@arm.com>
8084 * doc/md.texi (extract_last_@var{m}): Document.
8085 * optabs.def (extract_last_optab): New optab.
8086 * internal-fn.def (EXTRACT_LAST): New internal function.
8087 * internal-fn.c (cond_unary_direct): New macro.
8088 (expand_cond_unary_optab_fn): Likewise.
8089 (direct_cond_unary_optab_supported_p): Likewise.
8090 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8091 loops using EXTRACT_LAST.
8092 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8093 (extract_last_<mode>): ...this optab.
8094 (vec_extract<mode><Vel>): Update accordingly.
8096 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8097 Alan Hayward <alan.hayward@arm.com>
8098 David Sherwood <david.sherwood@arm.com>
8100 * target.def (empty_mask_is_expensive): New hook.
8101 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8102 * doc/tm.texi: Regenerate.
8103 * targhooks.h (default_empty_mask_is_expensive): Declare.
8104 * targhooks.c (default_empty_mask_is_expensive): New function.
8105 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8106 if the target says that empty masks are expensive.
8107 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8109 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8111 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8112 Alan Hayward <alan.hayward@arm.com>
8113 David Sherwood <david.sherwood@arm.com>
8115 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8116 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8117 (vect_use_loop_mask_for_alignment_p): New function.
8118 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8119 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8120 niters_skip argument. Make sure that the first niters_skip elements
8121 of the first iteration are inactive.
8122 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8123 Update call to vect_set_loop_masks_directly.
8124 (get_misalign_in_elems): New function, split out from...
8125 (vect_gen_prolog_loop_niters): ...here.
8126 (vect_update_init_of_dr): Take a code argument that specifies whether
8127 the adjustment should be added or subtracted.
8128 (vect_update_init_of_drs): Likewise.
8129 (vect_prepare_for_masked_peels): New function.
8130 (vect_do_peeling): Skip prologue peeling if we're using a mask
8131 instead. Update call to vect_update_inits_of_drs.
8132 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8134 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8135 alignment. Do not include the number of peeled iterations in
8136 the minimum threshold in that case.
8137 (vectorizable_induction): Adjust the start value down by
8138 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8139 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8140 Take the number of skipped iterations into account when calculating
8142 * tree-vect-stmts.c (vect_gen_while_not): New function.
8144 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8145 Alan Hayward <alan.hayward@arm.com>
8146 David Sherwood <david.sherwood@arm.com>
8148 * doc/sourcebuild.texi (vect_fully_masked): Document.
8149 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8151 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8153 (vect_analyze_loop_2): ...here. Don't check the vectorization
8154 factor against the number of loop iterations if the loop is
8157 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8158 Alan Hayward <alan.hayward@arm.com>
8159 David Sherwood <david.sherwood@arm.com>
8161 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8162 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8163 (dump_groups): Update accordingly.
8164 (iv_use::mem_type): New member variable.
8165 (address_p): New function.
8166 (record_use): Add a mem_type argument and initialize the new
8168 (record_group_use): Add a mem_type argument. Use address_p.
8169 Remove obsolete null checks of base_object. Update call to record_use.
8170 (find_interesting_uses_op): Update call to record_group_use.
8171 (find_interesting_uses_cond): Likewise.
8172 (find_interesting_uses_address): Likewise.
8173 (get_mem_type_for_internal_fn): New function.
8174 (find_address_like_use): Likewise.
8175 (find_interesting_uses_stmt): Try find_address_like_use before
8176 calling find_interesting_uses_op.
8177 (addr_offset_valid_p): Use the iv mem_type field as the type
8178 of the addressed memory.
8179 (add_autoinc_candidates): Likewise.
8180 (get_address_cost): Likewise.
8181 (split_small_address_groups_p): Use address_p.
8182 (split_address_groups): Likewise.
8183 (add_iv_candidate_for_use): Likewise.
8184 (autoinc_possible_for_pair): Likewise.
8185 (rewrite_groups): Likewise.
8186 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8187 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8188 (get_alias_ptr_type_for_ptr_address): New function.
8189 (rewrite_use_address): Rewrite address uses in calls that were
8190 identified by find_address_like_use.
8192 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8193 Alan Hayward <alan.hayward@arm.com>
8194 David Sherwood <david.sherwood@arm.com>
8196 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8198 * gimple-expr.h (is_gimple_addressable: Likewise.
8199 * gimple-expr.c (is_gimple_address): Likewise.
8200 * internal-fn.c (expand_call_mem_ref): New function.
8201 (expand_mask_load_optab_fn): Use it.
8202 (expand_mask_store_optab_fn): Likewise.
8204 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8205 Alan Hayward <alan.hayward@arm.com>
8206 David Sherwood <david.sherwood@arm.com>
8208 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8209 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8210 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8211 (cond_umax@var{mode}): Document.
8212 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8213 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8214 (cond_umin_optab, cond_umax_optab): New optabs.
8215 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8216 (COND_IOR, COND_XOR): New internal functions.
8217 * internal-fn.h (get_conditional_internal_fn): Declare.
8218 * internal-fn.c (cond_binary_direct): New macro.
8219 (expand_cond_binary_optab_fn): Likewise.
8220 (direct_cond_binary_optab_supported_p): Likewise.
8221 (get_conditional_internal_fn): New function.
8222 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8223 Cope with reduction statements that are vectorized as calls rather
8225 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8226 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8227 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8228 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8229 (UNSPEC_COND_EOR): New unspecs.
8230 (optab): Add mappings for them.
8231 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8232 (sve_int_op, sve_fp_op): New int attributes.
8234 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8235 Alan Hayward <alan.hayward@arm.com>
8236 David Sherwood <david.sherwood@arm.com>
8238 * optabs.def (while_ult_optab): New optab.
8239 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8240 * internal-fn.def (WHILE_ULT): New internal function.
8241 * internal-fn.h (direct_internal_fn_supported_p): New override
8242 that takes two types as argument.
8243 * internal-fn.c (while_direct): New macro.
8244 (expand_while_optab_fn): New function.
8245 (convert_optab_supported_p): Likewise.
8246 (direct_while_optab_supported_p): New macro.
8247 * wide-int.h (wi::udiv_ceil): New function.
8248 * tree-vectorizer.h (rgroup_masks): New structure.
8249 (vec_loop_masks): New typedef.
8250 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8252 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8253 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8254 (vect_max_vf): New function.
8255 (slpeel_make_loop_iterate_ntimes): Delete.
8256 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8257 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8258 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8259 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8260 internal-fn.h, stor-layout.h and optabs-query.h.
8261 (vect_set_loop_mask): New function.
8262 (add_preheader_seq): Likewise.
8263 (add_header_seq): Likewise.
8264 (interleave_supported_p): Likewise.
8265 (vect_maybe_permute_loop_masks): Likewise.
8266 (vect_set_loop_masks_directly): Likewise.
8267 (vect_set_loop_condition_masked): Likewise.
8268 (vect_set_loop_condition_unmasked): New function, split out from
8269 slpeel_make_loop_iterate_ntimes.
8270 (slpeel_make_loop_iterate_ntimes): Rename to..
8271 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8272 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8273 (vect_do_peeling): Update call accordingly.
8274 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8276 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8277 mask_compare_type, can_fully_mask_p and fully_masked_p.
8278 (release_vec_loop_masks): New function.
8279 (_loop_vec_info): Use it to free the loop masks.
8280 (can_produce_all_loop_masks_p): New function.
8281 (vect_get_max_nscalars_per_iter): Likewise.
8282 (vect_verify_full_masking): Likewise.
8283 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8284 retries, and free the mask rgroups before retrying. Check loop-wide
8285 reasons for disallowing fully-masked loops. Make the final decision
8286 about whether use a fully-masked loop or not.
8287 (vect_estimate_min_profitable_iters): Do not assume that peeling
8288 for the number of iterations will be needed for fully-masked loops.
8289 (vectorizable_reduction): Disable fully-masked loops.
8290 (vectorizable_live_operation): Likewise.
8291 (vect_halve_mask_nunits): New function.
8292 (vect_double_mask_nunits): Likewise.
8293 (vect_record_loop_mask): Likewise.
8294 (vect_get_loop_mask): Likewise.
8295 (vect_transform_loop): Handle the case in which the final loop
8296 iteration might handle a partial vector. Call vect_set_loop_condition
8297 instead of slpeel_make_loop_iterate_ntimes.
8298 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8299 (check_load_store_masking): New function.
8300 (prepare_load_store_mask): Likewise.
8301 (vectorizable_store): Handle fully-masked loops.
8302 (vectorizable_load): Likewise.
8303 (supportable_widening_operation): Use vect_halve_mask_nunits for
8305 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8306 (vect_gen_while): New function.
8307 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8308 (aarch64_uqdec<mode>): New insn.
8310 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8311 Alan Hayward <alan.hayward@arm.com>
8312 David Sherwood <david.sherwood@arm.com>
8314 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8315 (reduc_xor_scal_optab): New optabs.
8316 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8317 (reduc_xor_scal_@var{m}): Document.
8318 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8319 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8321 * fold-const-call.c (fold_const_call): Handle them.
8322 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8323 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8324 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8325 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8326 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8327 (UNSPEC_XORV): New unspecs.
8328 (optab): Add entries for them.
8329 (BITWISEV): New int iterator.
8330 (bit_reduc_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 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8337 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8338 * optabs.def (vec_shl_insert_optab): New optab.
8339 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8340 (duplicate_and_interleave): Likewise.
8341 * tree-vect-loop.c: Include internal-fn.h.
8342 (neutral_op_for_slp_reduction): New function, split out from
8343 get_initial_defs_for_reduction.
8344 (get_initial_def_for_reduction): Handle option 2 for variable-length
8345 vectors by loading the neutral value into a vector and then shifting
8346 the initial value into element 0.
8347 (get_initial_defs_for_reduction): Replace the code argument with
8348 the neutral value calculated by neutral_op_for_slp_reduction.
8349 Use gimple_build_vector for constant-length vectors.
8350 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8351 but the first group_size elements have a neutral value.
8352 Use duplicate_and_interleave otherwise.
8353 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8354 Update call to get_initial_defs_for_reduction. Handle SLP
8355 reductions for variable-length vectors by creating one vector
8356 result for each scalar result, with the elements associated
8357 with other scalar results stubbed out with the neutral value.
8358 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8359 Require IFN_VEC_SHL_INSERT for double reductions on
8360 variable-length vectors, or SLP reductions that have
8361 a neutral value. Require can_duplicate_and_interleave_p
8362 support for variable-length unchained SLP reductions if there
8363 is no neutral value, such as for MIN/MAX reductions. Also require
8364 the number of vector elements to be a multiple of the number of
8365 SLP statements when doing variable-length unchained SLP reductions.
8366 Update call to vect_create_epilog_for_reduction.
8367 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8368 and remove initial values.
8369 (duplicate_and_interleave): Make public.
8370 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8371 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8373 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8374 Alan Hayward <alan.hayward@arm.com>
8375 David Sherwood <david.sherwood@arm.com>
8377 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8378 (can_duplicate_and_interleave_p): New function.
8379 (vect_get_and_check_slp_defs): Take the vector of statements
8380 rather than just the current one. Remove excess parentheses.
8381 Restriction rejectinon of vect_constant_def and vect_external_def
8382 for variable-length vectors to boolean types, or types for which
8383 can_duplicate_and_interleave_p is false.
8384 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8385 (duplicate_and_interleave): New function.
8386 (vect_get_constant_vectors): Use gimple_build_vector for
8387 constant-length vectors and suitable variable-length constant
8388 vectors. Use duplicate_and_interleave for other variable-length
8389 vectors. Don't defer the update when inserting new statements.
8391 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8392 Alan Hayward <alan.hayward@arm.com>
8393 David Sherwood <david.sherwood@arm.com>
8395 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8396 min_profitable_iters doesn't go negative.
8398 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8399 Alan Hayward <alan.hayward@arm.com>
8400 David Sherwood <david.sherwood@arm.com>
8402 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8403 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8404 * optabs.def (vec_mask_load_lanes_optab): New optab.
8405 (vec_mask_store_lanes_optab): Likewise.
8406 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8407 (MASK_STORE_LANES): Likewise.
8408 * internal-fn.c (mask_load_lanes_direct): New macro.
8409 (mask_store_lanes_direct): Likewise.
8410 (expand_mask_load_optab_fn): Handle masked operations.
8411 (expand_mask_load_lanes_optab_fn): New macro.
8412 (expand_mask_store_optab_fn): Handle masked operations.
8413 (expand_mask_store_lanes_optab_fn): New macro.
8414 (direct_mask_load_lanes_optab_supported_p): Likewise.
8415 (direct_mask_store_lanes_optab_supported_p): Likewise.
8416 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8418 (vect_load_lanes_supported): Likewise.
8419 * tree-vect-data-refs.c (strip_conversion): New function.
8420 (can_group_stmts_p): Likewise.
8421 (vect_analyze_data_ref_accesses): Use it instead of checking
8422 for a pair of assignments.
8423 (vect_store_lanes_supported): Take a masked_p parameter.
8424 (vect_load_lanes_supported): Likewise.
8425 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8426 vect_store_lanes_supported and vect_load_lanes_supported.
8427 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8428 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8429 parameter. Don't allow gaps for masked accesses.
8430 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8431 and vect_load_lanes_supported.
8432 (get_load_store_type): Take a masked_p parameter and update
8433 call to get_group_load_store_type.
8434 (vectorizable_store): Update call to get_load_store_type.
8435 Handle IFN_MASK_STORE_LANES.
8436 (vectorizable_load): Update call to get_load_store_type.
8437 Handle IFN_MASK_LOAD_LANES.
8439 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8440 Alan Hayward <alan.hayward@arm.com>
8441 David Sherwood <david.sherwood@arm.com>
8443 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8445 * config/aarch64/aarch64-protos.h
8446 (aarch64_sve_struct_memory_operand_p): Declare.
8447 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8448 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8449 (VPRED, vpred): Handle SVE structure modes.
8450 * config/aarch64/constraints.md (Utx): New constraint.
8451 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8452 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8453 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8454 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8455 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8456 structure modes. Split into pieces after RA.
8457 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8458 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8460 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8461 SVE structure modes.
8462 (aarch64_classify_address): Likewise.
8463 (sizetochar): Move earlier in file.
8464 (aarch64_print_operand): Handle SVE register lists.
8465 (aarch64_array_mode): New function.
8466 (aarch64_sve_struct_memory_operand_p): Likewise.
8467 (TARGET_ARRAY_MODE): Redefine.
8469 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8470 Alan Hayward <alan.hayward@arm.com>
8471 David Sherwood <david.sherwood@arm.com>
8473 * target.def (array_mode): New target hook.
8474 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8475 * doc/tm.texi: Regenerate.
8476 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8477 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8478 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8480 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8483 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8484 Alan Hayward <alan.hayward@arm.com>
8485 David Sherwood <david.sherwood@arm.com>
8487 * fold-const.c (fold_binary_loc): Check the argument types
8488 rather than the result type when testing for a vector operation.
8490 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8492 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8493 * doc/tm.texi: Regenerate.
8495 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8496 Alan Hayward <alan.hayward@arm.com>
8497 David Sherwood <david.sherwood@arm.com>
8499 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8500 (sve): Document new AArch64 extension.
8501 * doc/md.texi (w): Extend the description of the AArch64
8502 constraint to include SVE vectors.
8503 (Upl, Upa): Document new AArch64 predicate constraints.
8504 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8506 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8507 (msve-vector-bits=): New option.
8508 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8509 SVE when these are disabled.
8510 (sve): New extension.
8511 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8512 modes. Adjust their number of units based on aarch64_sve_vg.
8513 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8514 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8515 aarch64_addr_query_type.
8516 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8517 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8518 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8519 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8520 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8521 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8522 (aarch64_simd_imm_zero_p): Delete.
8523 (aarch64_check_zero_based_sve_index_immediate): Declare.
8524 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8525 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8526 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8527 (aarch64_sve_float_mul_immediate_p): Likewise.
8528 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8530 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8531 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8532 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8533 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8534 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8535 (aarch64_regmode_natural_size): Likewise.
8536 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8537 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8539 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8540 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8541 for VG and the SVE predicate registers.
8542 (V_ALIASES): Add a "z"-prefixed alias.
8543 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8544 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8545 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8546 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8547 (REG_CLASS_NAMES): Add entries for them.
8548 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8549 and the predicate registers.
8550 (aarch64_sve_vg): Declare.
8551 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8552 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8553 (REGMODE_NATURAL_SIZE): Define.
8554 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8556 * config/aarch64/aarch64.c: Include cfgrtl.h.
8557 (simd_immediate_info): Add a constructor for series vectors,
8558 and an associated step field.
8559 (aarch64_sve_vg): New variable.
8560 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8561 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8562 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8563 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8564 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8565 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8566 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8567 (aarch64_get_mask_mode): New functions.
8568 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8569 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8570 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8571 predicate modes and predicate registers. Explicitly restrict
8572 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8573 to store a vector mode if it is recognized by
8574 aarch64_classify_vector_mode.
8575 (aarch64_regmode_natural_size): New function.
8576 (aarch64_hard_regno_caller_save_mode): Return the original mode
8578 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8579 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8580 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8581 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8583 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8584 does not overlap dest if the function is frame-related. Handle
8586 (aarch64_split_add_offset): New function.
8587 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8588 them aarch64_add_offset.
8589 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8590 and update call to aarch64_sub_sp.
8591 (aarch64_add_cfa_expression): New function.
8592 (aarch64_expand_prologue): Pass extra temporary registers to the
8593 functions above. Handle the case in which we need to emit new
8594 DW_CFA_expressions for registers that were originally saved
8595 relative to the stack pointer, but now have to be expressed
8596 relative to the frame pointer.
8597 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8599 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8600 IP0 and IP1 values for SVE frames.
8601 (aarch64_expand_vec_series): New function.
8602 (aarch64_expand_sve_widened_duplicate): Likewise.
8603 (aarch64_expand_sve_const_vector): Likewise.
8604 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8605 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8606 into the register, rather than emitting a SET directly.
8607 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8608 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8609 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8610 (offset_9bit_signed_scaled_p): New functions.
8611 (aarch64_replicate_bitmask_imm): New function.
8612 (aarch64_bitmask_imm): Use it.
8613 (aarch64_cannot_force_const_mem): Reject expressions involving
8614 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8615 (aarch64_classify_index): Handle SVE indices, by requiring
8616 a plain register index with a scale that matches the element size.
8617 (aarch64_classify_address): Handle SVE addresses. Assert that
8618 the mode of the address is VOIDmode or an integer mode.
8619 Update call to aarch64_classify_symbol.
8620 (aarch64_classify_symbolic_expression): Update call to
8621 aarch64_classify_symbol.
8622 (aarch64_const_vec_all_in_range_p): New function.
8623 (aarch64_print_vector_float_operand): Likewise.
8624 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8625 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8626 and the FP immediates 1.0 and 0.5.
8627 (aarch64_print_address_internal): Handle SVE addresses.
8628 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8629 (aarch64_regno_regclass): Handle predicate registers.
8630 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8632 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8633 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8634 (aarch64_convert_sve_vector_bits): New function.
8635 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8636 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8638 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8639 Handle SVE vector and predicate modes. Accept VL-based constants
8640 that need only one temporary register, and VL offsets that require
8641 no temporary registers.
8642 (aarch64_conditional_register_usage): Mark the predicate registers
8643 as fixed if SVE isn't available.
8644 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8645 Return true for SVE vector and predicate modes.
8646 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8647 rather than an unsigned int. Handle SVE modes.
8648 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8650 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8652 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8653 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8654 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8655 (aarch64_sve_float_mul_immediate_p): New functions.
8656 (aarch64_sve_valid_immediate): New function.
8657 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8658 Explicitly reject structure modes. Check for INDEX constants.
8659 Handle PTRUE and PFALSE constants.
8660 (aarch64_check_zero_based_sve_index_immediate): New function.
8661 (aarch64_simd_imm_zero_p): Delete.
8662 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8663 vector modes. Accept constants in the range of CNT[BHWD].
8664 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8665 ask for an Advanced SIMD mode.
8666 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8667 (aarch64_simd_vector_alignment): Handle SVE predicates.
8668 (aarch64_vectorize_preferred_vector_alignment): New function.
8669 (aarch64_simd_vector_alignment_reachable): Use it instead of
8671 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8672 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8674 (MAX_VECT_LEN): Delete.
8675 (expand_vec_perm_d): Add a vec_flags field.
8676 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8677 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8678 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8680 (aarch64_evpc_rev): Rename to...
8681 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8682 (aarch64_evpc_rev_global): New function.
8683 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8684 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8686 (aarch64_evpc_sve_tbl): New function.
8687 (aarch64_expand_vec_perm_const_1): Update after rename of
8688 aarch64_evpc_rev. Handle SVE permutes too, trying
8689 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8690 than aarch64_evpc_tbl.
8691 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8692 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8693 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8694 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8695 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8696 (aarch64_expand_sve_vcond): New functions.
8697 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8698 of aarch64_vector_mode_p.
8699 (aarch64_dwarf_poly_indeterminate_value): New function.
8700 (aarch64_compute_pressure_classes): Likewise.
8701 (aarch64_can_change_mode_class): Likewise.
8702 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8703 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8704 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8705 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8706 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8707 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8708 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8709 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8711 (Dn, Dl, Dr): Accept const as well as const_vector.
8712 (Dz): Likewise. Compare against CONST0_RTX.
8713 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8714 of "vector" where appropriate.
8715 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8716 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8717 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8718 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8719 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8720 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8721 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8722 (v_int_equiv): Extend to SVE modes.
8723 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8725 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8726 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8727 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8728 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8729 (SVE_COND_FP_CMP): New int iterators.
8730 (perm_hilo): Handle the new unpack unspecs.
8731 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8733 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8734 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8735 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8736 (aarch64_equality_operator, aarch64_constant_vector_operand)
8737 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8738 (aarch64_sve_nonimmediate_operand): Likewise.
8739 (aarch64_sve_general_operand): Likewise.
8740 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8741 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8742 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8743 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8744 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8745 (aarch64_sve_float_arith_immediate): Likewise.
8746 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8747 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8748 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8749 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8750 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8751 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8752 (aarch64_sve_float_arith_operand): Likewise.
8753 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8754 (aarch64_sve_float_mul_operand): Likewise.
8755 (aarch64_sve_vec_perm_operand): Likewise.
8756 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8757 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8758 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8759 as well as const_vector.
8760 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8761 in file. Use CONST0_RTX and CONSTM1_RTX.
8762 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8763 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8764 Use aarch64_simd_imm_zero.
8765 * config/aarch64/aarch64-sve.md: New file.
8766 * config/aarch64/aarch64.md: Include it.
8767 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8768 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8769 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8770 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8771 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8772 (sve): New attribute.
8773 (enabled): Disable instructions with the sve attribute unless
8775 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8776 aarch64_expand_mov_immediate.
8777 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8778 CNT[BHSD] immediates.
8779 (movti): Split CONST_POLY_INT moves into two halves.
8780 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8781 Split additions that need a temporary here if the destination
8782 is the stack pointer.
8783 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8784 (*add<mode>3_poly_1): New instruction.
8785 (set_clobber_cc): New expander.
8787 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8789 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8790 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8791 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8792 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8793 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8794 Change innermode from fixed_mode_size to machine_mode.
8795 (simplify_subreg): Update call accordingly. Handle a constant-sized
8796 subreg of a variable-length CONST_VECTOR.
8798 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8799 Alan Hayward <alan.hayward@arm.com>
8800 David Sherwood <david.sherwood@arm.com>
8802 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8803 (add_offset_to_base): New function, split out from...
8804 (create_mem_ref): ...here. When handling a scale other than 1,
8805 check first whether the address is valid without the offset.
8806 Add it into the base if so, leaving the index and scale as-is.
8808 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8811 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8812 fold_for_warn before checking if arg2 is INTEGER_CST.
8814 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8816 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8817 (store_multiple_operation): Delete.
8818 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8819 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8820 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8821 guarded by TARGET_STRING.
8822 (rs6000_output_load_multiple): Delete.
8823 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8824 OPTION_MASK_STRING / TARGET_STRING handling.
8825 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8826 (const rs6000_opt_masks) <"string">: Change mask to 0.
8827 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8828 (MASK_STRING): Delete.
8829 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8831 (load_multiple): Delete.
8838 (store_multiple): Delete.
8845 (movmemsi_8reg): Delete.
8846 (corresponding unnamed define_insn): Delete.
8847 (movmemsi_6reg): Delete.
8848 (corresponding unnamed define_insn): Delete.
8849 (movmemsi_4reg): Delete.
8850 (corresponding unnamed define_insn): Delete.
8851 (movmemsi_2reg): Delete.
8852 (corresponding unnamed define_insn): Delete.
8853 (movmemsi_1reg): Delete.
8854 (corresponding unnamed define_insn): Delete.
8855 * config/rs6000/rs6000.opt (mno-string): New.
8856 (mstring): Replace by deprecation warning stub.
8857 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8859 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8861 * regrename.c (regrename_do_replace): If replacing the same
8862 reg multiple times, try to reuse last created gen_raw_REG.
8865 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8866 main to workaround a bug in GDB.
8868 2018-01-12 Tom de Vries <tom@codesourcery.com>
8871 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8873 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8875 PR rtl-optimization/80481
8876 * ira-color.c (get_cap_member): New function.
8877 (allocnos_conflict_by_live_ranges_p): Use it.
8878 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8879 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8881 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8884 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8885 (*saddl_se_1): Ditto.
8887 (*ssubl_se_1): Ditto.
8889 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8891 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8892 rather than wi::to_widest for DR_INITs.
8893 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8894 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8895 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8897 (vect_analyze_group_access_1): Note that here.
8899 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8901 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8902 polynomial type sizes.
8904 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8906 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8907 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8908 (gimple_add_tmp_var): Likewise.
8910 2018-01-12 Martin Liska <mliska@suse.cz>
8912 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8913 (gimple_alloc_sizes): Likewise.
8914 (dump_gimple_statistics): Use PRIu64 in printf format.
8915 * gimple.h: Change uint64_t to int.
8917 2018-01-12 Martin Liska <mliska@suse.cz>
8919 * tree-core.h: Use uint64_t instead of int.
8920 * tree.c (tree_node_counts): Likewise.
8921 (tree_node_sizes): Likewise.
8922 (dump_tree_statistics): Use PRIu64 in printf format.
8924 2018-01-12 Martin Liska <mliska@suse.cz>
8926 * Makefile.in: As qsort_chk is implemented in vec.c, add
8927 vec.o to linkage of gencfn-macros.
8928 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8929 passing the info to record_node_allocation_statistics.
8930 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8932 * ggc-common.c (struct ggc_usage): Add operator== and use
8933 it in operator< and compare function.
8934 * mem-stats.h (struct mem_usage): Likewise.
8935 * vec.c (struct vec_usage): Remove operator< and compare
8936 function. Can be simply inherited.
8938 2018-01-12 Martin Jambor <mjambor@suse.cz>
8941 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8942 * tree-ssa-math-opts.c: Include domwalk.h.
8943 (convert_mult_to_fma_1): New function.
8944 (fma_transformation_info): New type.
8945 (fma_deferring_state): Likewise.
8946 (cancel_fma_deferring): New function.
8947 (result_of_phi): Likewise.
8948 (last_fma_candidate_feeds_initial_phi): Likewise.
8949 (convert_mult_to_fma): Added deferring logic, split actual
8950 transformation to convert_mult_to_fma_1.
8951 (math_opts_dom_walker): New type.
8952 (math_opts_dom_walker::after_dom_children): New method, body moved
8953 here from pass_optimize_widening_mul::execute, added deferring logic
8955 (pass_optimize_widening_mul::execute): Moved most of code to
8956 math_opts_dom_walker::after_dom_children.
8957 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8958 * config/i386/i386.c (ix86_option_override_internal): Added
8959 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8961 2018-01-12 Richard Biener <rguenther@suse.de>
8964 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8965 inline instance vars.
8967 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8970 * config/rx/rx.c (rx_is_restricted_memory_address):
8973 2018-01-12 Richard Biener <rguenther@suse.de>
8975 PR tree-optimization/80846
8976 * target.def (split_reduction): New target hook.
8977 * targhooks.c (default_split_reduction): New function.
8978 * targhooks.h (default_split_reduction): Declare.
8979 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8980 target requests first reduce vectors by combining low and high
8982 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8983 (get_vectype_for_scalar_type_and_size): Export.
8984 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8985 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8986 * doc/tm.texi: Regenerate.
8987 * config/i386/i386.c (ix86_split_reduction): Implement
8988 TARGET_VECTORIZE_SPLIT_REDUCTION.
8990 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8993 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8994 in PIC mode except for TARGET_VXWORKS_RTP.
8995 * config/sparc/sparc.c: Include cfgrtl.h.
8996 (TARGET_INIT_PIC_REG): Define.
8997 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8998 (sparc_pic_register_p): New predicate.
8999 (sparc_legitimate_address_p): Use it.
9000 (sparc_legitimize_pic_address): Likewise.
9001 (sparc_delegitimize_address): Likewise.
9002 (sparc_mode_dependent_address_p): Likewise.
9003 (gen_load_pcrel_sym): Remove 4th parameter.
9004 (load_got_register): Adjust call to above. Remove obsolete stuff.
9005 (sparc_expand_prologue): Do not call load_got_register here.
9006 (sparc_flat_expand_prologue): Likewise.
9007 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9008 (sparc_use_pseudo_pic_reg): New function.
9009 (sparc_init_pic_reg): Likewise.
9010 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9011 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9013 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9015 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9016 Add item for branch_cost.
9018 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9020 PR rtl-optimization/83565
9021 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9022 not extend the result to a larger mode for rotate operations.
9023 (num_sign_bit_copies1): Likewise.
9025 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9028 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9030 Use values-Xc.o for -pedantic.
9031 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9033 2018-01-12 Martin Liska <mliska@suse.cz>
9036 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9038 (possible_polymorphic_call_targets): Use it.
9039 (ipa_devirt): Likewise.
9041 2018-01-12 Martin Liska <mliska@suse.cz>
9043 * profile-count.h (enum profile_quality): Use 0 as invalid
9044 enum value of profile_quality.
9046 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9048 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9049 -mext-string options.
9051 2018-01-12 Richard Biener <rguenther@suse.de>
9053 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9054 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9055 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9057 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9059 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9061 * configure.ac (--with-long-double-format): Add support for the
9062 configuration option to change the default long double format on
9064 * config.gcc (powerpc*-linux*-*): Likewise.
9065 * configure: Regenerate.
9066 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9067 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9068 used without modification.
9070 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9072 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9073 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9074 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9075 MISC_BUILTIN_SPEC_BARRIER.
9076 (rs6000_init_builtins): Likewise.
9077 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9079 (speculation_barrier): New define_insn.
9080 * doc/extend.texi: Document __builtin_speculation_barrier.
9082 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9085 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9086 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9087 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9089 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9090 integral modes instead of "ss" and "sd".
9091 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9092 vectors with 32-bit and 64-bit elements.
9093 (vecdupssescalarmodesuffix): New mode attribute.
9094 (vec_dup<mode>): Use it.
9096 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9099 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9100 frame if argument is passed on stack.
9102 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9105 * ree.c (combine_reaching_defs): Optimize also
9106 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9107 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9109 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9112 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9114 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9117 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9118 after they are computed.
9120 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9122 PR tree-optimization/83695
9123 * gimple-loop-linterchange.cc
9124 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9125 reset cached scev information after interchange.
9126 (pass_linterchange::execute): Remove call to scev_reset_htab.
9128 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9130 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9131 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9132 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9133 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9134 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9135 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9136 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9137 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9138 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9139 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9140 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9141 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9142 (V_lane_reg): Likewise.
9143 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9145 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9146 (vfmal_lane_low<mode>_intrinsic,
9147 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9148 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9149 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9150 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9151 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9152 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9154 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9156 * config/arm/arm-cpus.in (fp16fml): New feature.
9157 (ALL_SIMD): Add fp16fml.
9158 (armv8.2-a): Add fp16fml as an option.
9159 (armv8.3-a): Likewise.
9160 (armv8.4-a): Add fp16fml as part of fp16.
9161 * config/arm/arm.h (TARGET_FP16FML): Define.
9162 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9164 * config/arm/arm-modes.def (V2HF): Define.
9165 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9166 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9167 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9168 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9169 vfmsl_low, vfmsl_high): New set of builtins.
9170 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9171 (vfml_op): New code attribute.
9172 (VFMLHALVES): New int iterator.
9173 (VFML, VFMLSEL): New mode attributes.
9174 (V_reg): Define mapping for V2HF.
9175 (V_hi, V_lo): New mode attributes.
9176 (VF_constraint): Likewise.
9177 (vfml_half, vfml_half_selector): New int attributes.
9178 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9180 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9181 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9183 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9184 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9185 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9186 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9188 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9189 Document new effective target and option set.
9191 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9193 * config/arm/arm-cpus.in (armv8_4): New feature.
9194 (ARMv8_4a): New fgroup.
9195 (armv8.4-a): New arch.
9196 * config/arm/arm-tables.opt: Regenerate.
9197 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9198 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9199 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9200 Add matching rules for -march=armv8.4-a and extensions.
9201 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9203 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9206 * config/rx/rx.md (BW): New mode attribute.
9207 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9209 2018-01-11 Richard Biener <rguenther@suse.de>
9211 PR tree-optimization/83435
9212 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9213 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9214 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9216 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9217 Alan Hayward <alan.hayward@arm.com>
9218 David Sherwood <david.sherwood@arm.com>
9220 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9222 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9223 (aarch64_print_address_internal): Use it to check for a zero offset.
9225 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9226 Alan Hayward <alan.hayward@arm.com>
9227 David Sherwood <david.sherwood@arm.com>
9229 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9230 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9231 Return a poly_int64 rather than a HOST_WIDE_INT.
9232 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9233 rather than a HOST_WIDE_INT.
9234 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9235 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9236 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9237 final_offset from HOST_WIDE_INT to poly_int64.
9238 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9239 to_constant when getting the number of units in an Advanced SIMD
9241 (aarch64_builtin_vectorized_function): Check for a constant number
9243 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9245 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9246 attribute instead of GET_MODE_NUNITS.
9247 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9248 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9249 GET_MODE_SIZE for fixed-size registers.
9250 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9251 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9252 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9253 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9254 (aarch64_print_operand, aarch64_print_address_internal)
9255 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9256 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9257 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9258 Handle polynomial GET_MODE_SIZE.
9259 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9260 wider than SImode without modification.
9261 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9262 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9263 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9264 passing and returning SVE modes.
9265 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9266 rather than GEN_INT.
9267 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9268 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9269 (aarch64_allocate_and_probe_stack_space): Likewise.
9270 (aarch64_layout_frame): Cope with polynomial offsets.
9271 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9272 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9274 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9275 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9276 poly_int64 rather than a HOST_WIDE_INT.
9277 (aarch64_get_separate_components, aarch64_process_components)
9278 (aarch64_expand_prologue, aarch64_expand_epilogue)
9279 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9280 (aarch64_anchor_offset): New function, split out from...
9281 (aarch64_legitimize_address): ...here.
9282 (aarch64_builtin_vectorization_cost): Handle polynomial
9283 TYPE_VECTOR_SUBPARTS.
9284 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9286 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9287 number of elements from the PARALLEL rather than the mode.
9288 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9289 rather than GET_MODE_BITSIZE.
9290 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9291 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9292 (aarch64_expand_vec_perm_const_1): Handle polynomial
9293 d->perm.length () and d->perm elements.
9294 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9295 Apply to_constant to d->perm elements.
9296 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9297 polynomial CONST_VECTOR_NUNITS.
9298 (aarch64_move_pointer): Take amount as a poly_int64 rather
9300 (aarch64_progress_pointer): Avoid temporary variable.
9301 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9302 the mode attribute instead of GET_MODE.
9304 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9305 Alan Hayward <alan.hayward@arm.com>
9306 David Sherwood <david.sherwood@arm.com>
9308 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9309 x exists before using it.
9310 (aarch64_add_constant_internal): Rename to...
9311 (aarch64_add_offset_1): ...this. Replace regnum with separate
9312 src and dest rtxes. Handle the case in which they're different,
9313 including when the offset is zero. Replace scratchreg with an rtx.
9314 Use 2 additions if there is no spare register into which we can
9315 move a 16-bit constant.
9316 (aarch64_add_constant): Delete.
9317 (aarch64_add_offset): Replace reg with separate src and dest
9318 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9319 Use aarch64_add_offset_1.
9320 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9321 an rtx rather than an int. Take the delta as a poly_int64
9322 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9323 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9324 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9325 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9326 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9328 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9329 aarch64_add_constant.
9331 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9333 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9334 Use scalar_float_mode.
9336 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9338 * config/aarch64/aarch64-simd.md
9339 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9340 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9341 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9342 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9343 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9344 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9345 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9346 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9347 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9348 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9350 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9353 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9354 targ_options->x_arm_arch_string is non NULL.
9356 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9358 * config/aarch64/aarch64.h
9359 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9361 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9364 * expmed.c (emit_store_flag_force): Swap if const op0
9365 and change VOIDmode to mode of op0.
9367 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9369 PR rtl-optimization/83761
9370 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9371 than bytes to mode_for_size.
9373 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9376 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9377 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9380 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9383 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9384 when in layout mode.
9385 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9386 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9389 2018-01-10 Michael Collison <michael.collison@arm.com>
9391 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9392 * config/aarch64/aarch64-option-extension.def: Add
9393 AARCH64_OPT_EXTENSION of 'fp16fml'.
9394 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9395 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9396 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9397 * config/aarch64/constraints.md (Ui7): New constraint.
9398 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9399 (VFMLA_SEL_W): Ditto.
9402 (VFMLA16_LOW): New int iterator.
9403 (VFMLA16_HIGH): Ditto.
9404 (UNSPEC_FMLAL): New unspec.
9405 (UNSPEC_FMLSL): Ditto.
9406 (UNSPEC_FMLAL2): Ditto.
9407 (UNSPEC_FMLSL2): Ditto.
9408 (f16mac): New code attribute.
9409 * config/aarch64/aarch64-simd-builtins.def
9410 (aarch64_fmlal_lowv2sf): Ditto.
9411 (aarch64_fmlsl_lowv2sf): Ditto.
9412 (aarch64_fmlalq_lowv4sf): Ditto.
9413 (aarch64_fmlslq_lowv4sf): Ditto.
9414 (aarch64_fmlal_highv2sf): Ditto.
9415 (aarch64_fmlsl_highv2sf): Ditto.
9416 (aarch64_fmlalq_highv4sf): Ditto.
9417 (aarch64_fmlslq_highv4sf): Ditto.
9418 (aarch64_fmlal_lane_lowv2sf): Ditto.
9419 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9420 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9421 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9422 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9423 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9424 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9425 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9426 (aarch64_fmlal_lane_highv2sf): Ditto.
9427 (aarch64_fmlsl_lane_highv2sf): Ditto.
9428 (aarch64_fmlal_laneq_highv2sf): Ditto.
9429 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9430 (aarch64_fmlalq_lane_highv4sf): Ditto.
9431 (aarch64_fmlsl_lane_highv4sf): Ditto.
9432 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9433 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9434 * config/aarch64/aarch64-simd.md:
9435 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9436 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9437 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9438 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9439 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9440 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9441 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9442 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9443 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9444 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9445 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9446 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9447 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9448 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9449 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9450 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9451 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9452 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9453 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9454 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9455 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9456 (vfmlsl_low_u32): Ditto.
9457 (vfmlalq_low_u32): Ditto.
9458 (vfmlslq_low_u32): Ditto.
9459 (vfmlal_high_u32): Ditto.
9460 (vfmlsl_high_u32): Ditto.
9461 (vfmlalq_high_u32): Ditto.
9462 (vfmlslq_high_u32): Ditto.
9463 (vfmlal_lane_low_u32): Ditto.
9464 (vfmlsl_lane_low_u32): Ditto.
9465 (vfmlal_laneq_low_u32): Ditto.
9466 (vfmlsl_laneq_low_u32): Ditto.
9467 (vfmlalq_lane_low_u32): Ditto.
9468 (vfmlslq_lane_low_u32): Ditto.
9469 (vfmlalq_laneq_low_u32): Ditto.
9470 (vfmlslq_laneq_low_u32): Ditto.
9471 (vfmlal_lane_high_u32): Ditto.
9472 (vfmlsl_lane_high_u32): Ditto.
9473 (vfmlal_laneq_high_u32): Ditto.
9474 (vfmlsl_laneq_high_u32): Ditto.
9475 (vfmlalq_lane_high_u32): Ditto.
9476 (vfmlslq_lane_high_u32): Ditto.
9477 (vfmlalq_laneq_high_u32): Ditto.
9478 (vfmlslq_laneq_high_u32): Ditto.
9479 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9480 (AARCH64_FL_FOR_ARCH8_4): New.
9481 (AARCH64_ISA_F16FML): New ISA flag.
9482 (TARGET_F16FML): New feature flag for fp16fml.
9483 (doc/invoke.texi): Document new fp16fml option.
9485 2018-01-10 Michael Collison <michael.collison@arm.com>
9487 * config/aarch64/aarch64-builtins.c:
9488 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9489 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9490 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9491 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9492 (AARCH64_ISA_SHA3): New ISA flag.
9493 (TARGET_SHA3): New feature flag for sha3.
9494 * config/aarch64/iterators.md (sha512_op): New int attribute.
9495 (CRYPTO_SHA512): New int iterator.
9496 (UNSPEC_SHA512H): New unspec.
9497 (UNSPEC_SHA512H2): Ditto.
9498 (UNSPEC_SHA512SU0): Ditto.
9499 (UNSPEC_SHA512SU1): Ditto.
9500 * config/aarch64/aarch64-simd-builtins.def
9501 (aarch64_crypto_sha512hqv2di): New builtin.
9502 (aarch64_crypto_sha512h2qv2di): Ditto.
9503 (aarch64_crypto_sha512su0qv2di): Ditto.
9504 (aarch64_crypto_sha512su1qv2di): Ditto.
9505 (aarch64_eor3qv8hi): Ditto.
9506 (aarch64_rax1qv2di): Ditto.
9507 (aarch64_xarqv2di): Ditto.
9508 (aarch64_bcaxqv8hi): Ditto.
9509 * config/aarch64/aarch64-simd.md:
9510 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9511 (aarch64_crypto_sha512su0qv2di): Ditto.
9512 (aarch64_crypto_sha512su1qv2di): Ditto.
9513 (aarch64_eor3qv8hi): Ditto.
9514 (aarch64_rax1qv2di): Ditto.
9515 (aarch64_xarqv2di): Ditto.
9516 (aarch64_bcaxqv8hi): Ditto.
9517 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9518 (vsha512h2q_u64): Ditto.
9519 (vsha512su0q_u64): Ditto.
9520 (vsha512su1q_u64): Ditto.
9521 (veor3q_u16): Ditto.
9522 (vrax1q_u64): Ditto.
9524 (vbcaxq_u16): Ditto.
9525 * config/arm/types.md (crypto_sha512): New type attribute.
9526 (crypto_sha3): Ditto.
9527 (doc/invoke.texi): Document new sha3 option.
9529 2018-01-10 Michael Collison <michael.collison@arm.com>
9531 * config/aarch64/aarch64-builtins.c:
9532 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9533 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9534 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9535 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9536 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9537 (AARCH64_ISA_SM4): New ISA flag.
9538 (TARGET_SM4): New feature flag for sm4.
9539 * config/aarch64/aarch64-simd-builtins.def
9540 (aarch64_sm3ss1qv4si): Ditto.
9541 (aarch64_sm3tt1aq4si): Ditto.
9542 (aarch64_sm3tt1bq4si): Ditto.
9543 (aarch64_sm3tt2aq4si): Ditto.
9544 (aarch64_sm3tt2bq4si): Ditto.
9545 (aarch64_sm3partw1qv4si): Ditto.
9546 (aarch64_sm3partw2qv4si): Ditto.
9547 (aarch64_sm4eqv4si): Ditto.
9548 (aarch64_sm4ekeyqv4si): Ditto.
9549 * config/aarch64/aarch64-simd.md:
9550 (aarch64_sm3ss1qv4si): Ditto.
9551 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9552 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9553 (aarch64_sm4eqv4si): Ditto.
9554 (aarch64_sm4ekeyqv4si): Ditto.
9555 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9556 (sm3part_op): Ditto.
9557 (CRYPTO_SM3TT): Ditto.
9558 (CRYPTO_SM3PART): Ditto.
9559 (UNSPEC_SM3SS1): New unspec.
9560 (UNSPEC_SM3TT1A): Ditto.
9561 (UNSPEC_SM3TT1B): Ditto.
9562 (UNSPEC_SM3TT2A): Ditto.
9563 (UNSPEC_SM3TT2B): Ditto.
9564 (UNSPEC_SM3PARTW1): Ditto.
9565 (UNSPEC_SM3PARTW2): Ditto.
9566 (UNSPEC_SM4E): Ditto.
9567 (UNSPEC_SM4EKEY): Ditto.
9568 * config/aarch64/constraints.md (Ui2): New constraint.
9569 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9570 * config/arm/types.md (crypto_sm3): New type attribute.
9571 (crypto_sm4): Ditto.
9572 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9573 (vsm3tt1aq_u32): Ditto.
9574 (vsm3tt1bq_u32): Ditto.
9575 (vsm3tt2aq_u32): Ditto.
9576 (vsm3tt2bq_u32): Ditto.
9577 (vsm3partw1q_u32): Ditto.
9578 (vsm3partw2q_u32): Ditto.
9579 (vsm4eq_u32): Ditto.
9580 (vsm4ekeyq_u32): Ditto.
9581 (doc/invoke.texi): Document new sm4 option.
9583 2018-01-10 Michael Collison <michael.collison@arm.com>
9585 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9586 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9587 (AARCH64_FL_FOR_ARCH8_4): New.
9588 (AARCH64_FL_V8_4): New flag.
9589 (doc/invoke.texi): Document new armv8.4-a option.
9591 2018-01-10 Michael Collison <michael.collison@arm.com>
9593 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9594 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9595 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9596 * config/aarch64/aarch64-option-extension.def: Add
9597 AARCH64_OPT_EXTENSION of 'sha2'.
9598 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9599 (crypto): Disable sha2 and aes if crypto disabled.
9600 (crypto): Enable aes and sha2 if enabled.
9601 (simd): Disable sha2 and aes if simd disabled.
9602 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9604 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9605 (TARGET_SHA2): New feature flag for sha2.
9606 (TARGET_AES): New feature flag for aes.
9607 * config/aarch64/aarch64-simd.md:
9608 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9609 conditional on TARGET_AES.
9610 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9611 (aarch64_crypto_sha1hsi): Make pattern conditional
9613 (aarch64_crypto_sha1hv4si): Ditto.
9614 (aarch64_be_crypto_sha1hv4si): Ditto.
9615 (aarch64_crypto_sha1su1v4si): Ditto.
9616 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9617 (aarch64_crypto_sha1su0v4si): Ditto.
9618 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9619 (aarch64_crypto_sha256su0v4si): Ditto.
9620 (aarch64_crypto_sha256su1v4si): Ditto.
9621 (doc/invoke.texi): Document new aes and sha2 options.
9623 2018-01-10 Martin Sebor <msebor@redhat.com>
9625 PR tree-optimization/83781
9626 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9629 2018-01-11 Martin Sebor <msebor@gmail.com>
9630 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9632 PR tree-optimization/83501
9633 PR tree-optimization/81703
9635 * tree-ssa-strlen.c (get_string_cst): Rename...
9636 (get_string_len): ...to this. Handle global constants.
9637 (handle_char_store): Adjust.
9639 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9640 Jim Wilson <jimw@sifive.com>
9642 * config/riscv/riscv-protos.h (riscv_output_return): New.
9643 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9644 (riscv_attribute_table, riscv_output_return),
9645 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9646 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9647 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9649 (riscv_expand_prologue): Add early return for naked function.
9650 (riscv_expand_epilogue): Likewise.
9651 (riscv_function_ok_for_sibcall): Return false for naked function.
9652 (riscv_set_current_function): New.
9653 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9654 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9655 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9656 * doc/extend.texi (RISC-V Function Attributes): New.
9658 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9660 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9661 check for 128-bit long double before checking TCmode.
9662 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9663 128-bit long doubles before checking TFmode or TCmode.
9664 (FLOAT128_IBM_P): Likewise.
9666 2018-01-10 Martin Sebor <msebor@redhat.com>
9668 PR tree-optimization/83671
9669 * builtins.c (c_strlen): Unconditionally return zero for the empty
9671 Use -Warray-bounds for warnings.
9672 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9673 for non-constant array indices with COMPONENT_REF, arrays of
9674 arrays, and pointers to arrays.
9675 (gimple_fold_builtin_strlen): Determine and set length range for
9676 non-constant character arrays.
9678 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9681 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9684 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9686 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9688 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9691 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9692 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9693 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9694 indexed_or_indirect_operand predicate.
9695 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9696 (*vsx_le_perm_load_v8hi): Likewise.
9697 (*vsx_le_perm_load_v16qi): Likewise.
9698 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9699 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9700 (*vsx_le_perm_store_v8hi): Likewise.
9701 (*vsx_le_perm_store_v16qi): Likewise.
9702 (eight unnamed splitters): Likewise.
9704 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9706 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9707 * config/rs6000/emmintrin.h: Likewise.
9708 * config/rs6000/mmintrin.h: Likewise.
9709 * config/rs6000/xmmintrin.h: Likewise.
9711 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9714 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9716 * tree.c (tree_nop_conversion): Return true for location wrapper
9718 (maybe_wrap_with_location): New function.
9719 (selftest::check_strip_nops): New function.
9720 (selftest::test_location_wrappers): New function.
9721 (selftest::tree_c_tests): Call it.
9722 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9723 (maybe_wrap_with_location): New decl.
9724 (EXPR_LOCATION_WRAPPER_P): New macro.
9725 (location_wrapper_p): New inline function.
9726 (tree_strip_any_location_wrapper): New inline function.
9728 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9731 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9732 stack_realign_offset for the largest alignment of stack slot
9734 (ix86_find_max_used_stack_alignment): New function.
9735 (ix86_finalize_stack_frame_flags): Use it. Set
9736 max_used_stack_alignment if we don't realign stack.
9737 * config/i386/i386.h (machine_function): Add
9738 max_used_stack_alignment.
9740 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9742 * config/arm/arm.opt (-mbranch-cost): New option.
9743 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9746 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9749 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9750 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9752 2018-01-10 Richard Biener <rguenther@suse.de>
9755 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9756 early out so it also covers the case where we have a non-NULL
9759 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9761 PR tree-optimization/83753
9762 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9763 for non-strided grouped accesses if the number of elements is 1.
9765 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9768 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9769 * i386.h (TARGET_USE_GATHER): Define.
9770 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9772 2018-01-10 Martin Liska <mliska@suse.cz>
9775 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9776 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9778 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9779 CLEANUP_NO_PARTITIONING is not set.
9781 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9783 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9784 for vectors, as a partial revert of r254296.
9785 * rtl.h (const_vec_p): Delete.
9786 (const_vec_duplicate_p): Don't test for vector CONSTs.
9787 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9788 * expmed.c (make_tree): Likewise.
9791 * common.md (E, F): Use CONSTANT_P instead of checking for
9793 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9794 checking for CONST_VECTOR.
9796 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9799 * predict.c (force_edge_cold): Handle in more sane way edges
9802 2018-01-09 Carl Love <cel@us.ibm.com>
9804 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9806 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9807 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9808 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9809 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9810 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9811 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9812 * config/rs6000/rs6000-protos.h: Add extern defition for
9813 rs6000_generate_float2_double_code.
9814 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9816 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9817 (float2_v2df): Add define_expand.
9819 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9822 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9823 op_mode in the force_to_mode call.
9825 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9827 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9828 instead of checking each element individually.
9829 (aarch64_evpc_uzp): Likewise.
9830 (aarch64_evpc_zip): Likewise.
9831 (aarch64_evpc_ext): Likewise.
9832 (aarch64_evpc_rev): Likewise.
9833 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9834 instead of checking each element individually. Return true without
9836 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9837 whether all selected elements come from the same input, instead of
9838 checking each element individually. Remove calls to gen_rtx_REG,
9839 start_sequence and end_sequence and instead assert that no rtl is
9842 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9844 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9845 order of HIGH and CONST checks.
9847 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9849 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9850 if the destination isn't an SSA_NAME.
9852 2018-01-09 Richard Biener <rguenther@suse.de>
9854 PR tree-optimization/83668
9855 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9857 (canonicalize_loop_form): ... here, renamed from ...
9858 (canonicalize_loop_closed_ssa_form): ... this and amended to
9859 swap successor edges for loop exit blocks to make us use
9860 the RPO order we need for initial schedule generation.
9862 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9864 PR tree-optimization/64811
9865 * match.pd: When optimizing comparisons with Inf, avoid
9866 introducing or losing exceptions from comparisons with NaN.
9868 2018-01-09 Martin Liska <mliska@suse.cz>
9871 * asan.c (shadow_mem_size): Add gcc_assert.
9873 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9875 Don't save registers in main().
9878 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9879 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9880 * config/avr/avr.c (avr_set_current_function): Don't error if
9881 naked, OS_task or OS_main are specified at the same time.
9882 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9884 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9886 * common/config/avr/avr-common.c (avr_option_optimization_table):
9887 Switch on -mmain-is-OS_task for optimizing compilations.
9889 2018-01-09 Richard Biener <rguenther@suse.de>
9891 PR tree-optimization/83572
9892 * graphite.c: Include cfganal.h.
9893 (graphite_transform_loops): Connect infinite loops to exit
9894 and remove fake edges at the end.
9896 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9898 * ipa-inline.c (edge_badness): Revert accidental checkin.
9900 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9903 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9904 symbols; not inline clones.
9906 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9909 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9910 hard registers. Formatting fixes.
9912 PR preprocessor/83722
9913 * gcc.c (try_generate_repro): Pass
9914 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9915 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9918 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9919 Kito Cheng <kito.cheng@gmail.com>
9921 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9922 (riscv_leaf_function_p): Delete.
9923 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9925 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9927 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9929 (do_ifelse): New function.
9930 (do_isel): New function.
9931 (do_sub3): New function.
9932 (do_add3): New function.
9933 (do_load_mask_compare): New function.
9934 (do_overlap_load_compare): New function.
9935 (expand_compare_loop): New function.
9936 (expand_block_compare): Call expand_compare_loop() when appropriate.
9937 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9939 (-mblock-compare-inline-loop-limit): New option.
9941 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9944 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9945 Reverse order of second and third operands in first alternative.
9946 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9947 of first and second elements in UNSPEC_VPERMR vector.
9948 (altivec_expand_vec_perm_le): Likewise.
9950 2018-01-08 Jeff Law <law@redhat.com>
9952 PR rtl-optimizatin/81308
9953 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9954 (process_switch): If group_case_labels makes a change, then set
9956 (pass_convert_switch::execute): If a switch is converted, then
9957 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9959 PR rtl-optimization/81308
9960 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9963 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9965 PR target/83663 - Revert r255946
9966 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9967 generation for cases where splatting a value is not useful.
9968 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9969 across a vec_duplicate and a paradoxical subreg forming a vector
9970 mode to a vec_concat.
9972 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9974 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9975 -march=armv8.3-a variants.
9976 * config/arm/t-multilib: Likewise.
9977 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9979 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9981 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9983 (cceq_ior_compare_complement): Give it a name so I can use it, and
9984 change boolean_or_operator predicate to boolean_operator so it can
9985 be used to generate a crand.
9986 (eqne): New code iterator.
9987 (bd/bd_neg): New code_attrs.
9988 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9989 a single define_insn.
9990 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9991 decrement (bdnzt/bdnzf/bdzt/bdzf).
9992 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9993 with the new names of the branch decrement patterns, and added the
9994 names of the branch decrement conditional patterns.
9996 2018-01-08 Richard Biener <rguenther@suse.de>
9998 PR tree-optimization/83563
9999 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10002 2018-01-08 Richard Biener <rguenther@suse.de>
10004 PR middle-end/83713
10005 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10007 2018-01-08 Richard Biener <rguenther@suse.de>
10009 PR tree-optimization/83685
10010 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10011 references to abnormals.
10013 2018-01-08 Richard Biener <rguenther@suse.de>
10016 * dwarf2out.c (output_indirect_strings): Handle empty
10017 skeleton_debug_str_hash.
10018 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10020 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10022 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10023 (emit_store_direct): Likewise.
10024 (arc_trampoline_adjust_address): Likewise.
10025 (arc_asm_trampoline_template): New function.
10026 (arc_initialize_trampoline): Use asm_trampoline_template.
10027 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10028 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10029 * config/arc/arc.md (flush_icache): Delete pattern.
10031 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10033 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10034 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10037 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10040 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10041 by not USED_FOR_TARGET.
10042 (make_pass_resolve_sw_modes): Likewise.
10044 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10046 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10049 2018-01-08 Richard Biener <rguenther@suse.de>
10051 PR middle-end/83580
10052 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10054 2018-01-08 Richard Biener <rguenther@suse.de>
10056 PR middle-end/83517
10057 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10059 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10061 PR middle-end/81897
10062 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10063 basic blocks with a small number of successors.
10064 (convert_control_dep_chain_into_preds): Improve handling of
10066 (dump_predicates): Split apart into...
10067 (dump_pred_chain): ...here...
10068 (dump_pred_info): ...and here.
10069 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10070 (can_chain_union_be_invalidated_p): Improve check for invalidation
10072 (uninit_uses_cannot_happen): Avoid unnecessary if
10073 convert_control_dep_chain_into_preds yielded nothing.
10075 2018-01-06 Martin Sebor <msebor@redhat.com>
10077 PR tree-optimization/83640
10078 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10079 subtracting negative offset from size.
10080 (builtin_access::overlap): Adjust offset bounds of the access to fall
10081 within the size of the object if possible.
10083 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10085 PR rtl-optimization/83699
10086 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10087 extract_bit_field_as_subreg to cases in which the extracted
10088 value is also a vector.
10090 * lra-constraints.c (process_alt_operands): Test for the equivalence
10091 substitutions when detecting a possible reload cycle.
10093 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10096 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10097 by default if flag_selective_schedling{,2}. Formatting fixes.
10099 PR rtl-optimization/83682
10100 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10101 if it has non-VECTOR_MODE element mode.
10102 (vec_duplicate_p): Likewise.
10104 PR middle-end/83694
10105 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10106 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10108 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10111 * config/i386/i386-builtin.def
10112 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10113 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10114 Require also OPTION_MASK_ISA_AVX512F in addition to
10115 OPTION_MASK_ISA_GFNI.
10116 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10117 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10118 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10119 to OPTION_MASK_ISA_GFNI.
10120 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10121 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10122 OPTION_MASK_ISA_AVX512BW.
10123 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10124 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10125 addition to OPTION_MASK_ISA_GFNI.
10126 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10127 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10128 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10129 to OPTION_MASK_ISA_GFNI.
10130 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10131 a requirement for all ISAs rather than any of them with a few
10133 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10135 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10136 bitmasks to be enabled with 3 exceptions, instead of requiring any
10137 enabled ISA with lots of exceptions.
10138 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10139 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10140 Change avx512bw in isa attribute to avx512f.
10141 * config/i386/sgxintrin.h: Add license boilerplate.
10142 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10143 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10144 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10145 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10147 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10148 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10149 temporarily sse2 rather than sse if not enabled already.
10152 * config/i386/sse.md (VI248_VLBW): Rename to ...
10153 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10154 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10155 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10156 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10157 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10158 mode iterator instead of VI248_VLBW.
10160 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10162 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10163 (record_modified): Skip clobbers; add debug output.
10164 (param_change_prob): Use sreal frequencies.
10166 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10168 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10169 punt for user-aligned variables.
10171 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10173 * tree-chrec.c (chrec_contains_symbols): Return true for
10176 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10179 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10180 of (x|y) == x for BICS pattern.
10182 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10184 PR tree-optimization/83605
10185 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10186 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10189 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10191 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10192 * config/epiphany/rtems.h: New file.
10194 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10195 Uros Bizjak <ubizjak@gmail.com>
10198 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10199 QIreg_operand instead of register_operand predicate.
10200 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10201 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10202 comments instead of -fmitigate[-_]rop.
10204 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10207 * cgraphunit.c (symbol_table::compile): Switch to text_section
10208 before calling assembly_start debug hook.
10209 * run-rtl-passes.c (run_rtl_passes): Likewise.
10212 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10214 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10215 range_int_cst_p rather than !symbolic_range_p before calling
10216 extract_range_from_multiplicative_op_1.
10218 2018-01-04 Jeff Law <law@redhat.com>
10220 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10221 redundant test in assertion.
10223 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10225 * doc/rtl.texi: Document machine_mode wrapper classes.
10227 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10229 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10230 using tree_to_uhwi.
10232 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10234 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10235 the VEC_PERM_EXPR fold to fail.
10237 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10240 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10241 to switched_sections.
10243 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10246 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10247 test for d.testing.
10249 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10252 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10253 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10255 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10258 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10259 is BLKmode and bitpos not zero or mode change is needed.
10261 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10264 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10267 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10270 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10271 instead of MULT rtx. Update all corresponding splitters.
10272 (*saddl_se): Ditto.
10273 (*ssub<modesuffix>): Ditto.
10274 (*ssubl_se): Ditto.
10275 (*cmp_sadd_di): Update split patterns.
10276 (*cmp_sadd_si): Ditto.
10277 (*cmp_sadd_sidi): Ditto.
10278 (*cmp_ssub_di): Ditto.
10279 (*cmp_ssub_si): Ditto.
10280 (*cmp_ssub_sidi): Ditto.
10281 * config/alpha/predicates.md (const23_operand): New predicate.
10282 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10283 Look for ASHIFT, not MULT inner operand.
10284 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10286 2018-01-04 Martin Liska <mliska@suse.cz>
10288 PR gcov-profile/83669
10289 * gcov.c (output_intermediate_file): Add version to intermediate
10291 * doc/gcov.texi: Document new field 'version' in intermediate
10292 file format. Fix location of '-k' option of gcov command.
10294 2018-01-04 Martin Liska <mliska@suse.cz>
10297 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10299 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10301 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10303 2018-01-03 Martin Sebor <msebor@redhat.com>
10305 PR tree-optimization/83655
10306 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10307 checking calls with invalid arguments.
10309 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10311 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10312 (vectorizable_mask_load_store): Delete.
10313 (vectorizable_call): Return false for masked loads and stores.
10314 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10315 instead of gimple_assign_rhs1.
10316 (vectorizable_load): Handle IFN_MASK_LOAD.
10317 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10321 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10323 (vectorizable_mask_load_store): ...here.
10324 (vectorizable_load): ...and here.
10326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10328 * tree-vect-stmts.c (vect_build_all_ones_mask)
10329 (vect_build_zero_merge_argument): New functions, split out from...
10330 (vectorizable_load): ...here.
10332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10334 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10336 (vectorizable_mask_load_store): ...here.
10337 (vectorizable_store): ...and here.
10339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10341 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10343 (vectorizable_mask_load_store): ...here.
10345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10347 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10348 (vect_model_store_cost): Take a vec_load_store_type instead of a
10350 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10351 (vect_model_store_cost): Take a vec_load_store_type instead of a
10353 (vectorizable_mask_load_store): Update accordingly.
10354 (vectorizable_store): Likewise.
10355 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10359 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10360 IFN_MASK_LOAD calls here rather than...
10361 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10364 Alan Hayward <alan.hayward@arm.com>
10365 David Sherwood <david.sherwood@arm.com>
10367 * expmed.c (extract_bit_field_1): For vector extracts,
10368 fall back to extract_bit_field_as_subreg if vec_extract
10371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10372 Alan Hayward <alan.hayward@arm.com>
10373 David Sherwood <david.sherwood@arm.com>
10375 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10376 they are variable or constant sized.
10377 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10378 slots for constant-sized data.
10380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10381 Alan Hayward <alan.hayward@arm.com>
10382 David Sherwood <david.sherwood@arm.com>
10384 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10385 handling COND_EXPRs with boolean comparisons, try to find a better
10386 basis for the mask type than the boolean itself.
10388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10390 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10391 is calculated and how it can be overridden.
10392 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10393 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10395 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10399 Alan Hayward <alan.hayward@arm.com>
10400 David Sherwood <david.sherwood@arm.com>
10402 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10403 Remove the mode argument.
10404 (aarch64_simd_valid_immediate): Remove the mode and inverse
10406 * config/aarch64/iterators.md (bitsize): New iterator.
10407 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10408 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10409 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10410 aarch64_simd_valid_immediate.
10411 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10412 (aarch64_reg_or_bic_imm): Likewise.
10413 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10414 with an insn_type enum and msl with a modifier_type enum.
10415 Replace element_width with a scalar_mode. Change the shift
10416 to unsigned int. Add constructors for scalar_float_mode and
10417 scalar_int_mode elements.
10418 (aarch64_vect_float_const_representable_p): Delete.
10419 (aarch64_can_const_movi_rtx_p)
10420 (aarch64_simd_scalar_immediate_valid_for_move)
10421 (aarch64_simd_make_constant): Update call to
10422 aarch64_simd_valid_immediate.
10423 (aarch64_advsimd_valid_immediate_hs): New function.
10424 (aarch64_advsimd_valid_immediate): Likewise.
10425 (aarch64_simd_valid_immediate): Remove mode and inverse
10426 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10427 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10428 and aarch64_float_const_representable_p on the result.
10429 (aarch64_output_simd_mov_immediate): Remove mode argument.
10430 Update call to aarch64_simd_valid_immediate and use of
10431 simd_immediate_info.
10432 (aarch64_output_scalar_simd_mov_immediate): Update call
10435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10436 Alan Hayward <alan.hayward@arm.com>
10437 David Sherwood <david.sherwood@arm.com>
10439 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10440 (mode_nunits): Likewise CONST_MODE_NUNITS.
10441 * machmode.def (ADJUST_NUNITS): Document.
10442 * genmodes.c (mode_data::need_nunits_adj): New field.
10443 (blank_mode): Update accordingly.
10444 (adj_nunits): New variable.
10445 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10447 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10448 listed in adj_nunits.
10449 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10450 listed in adj_nunits. Don't emit case statements for such modes.
10451 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10452 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10453 nothing if adj_nunits is nonnull.
10454 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10455 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10456 (emit_mode_fbit): Update use of print_maybe_const_decl.
10457 (emit_move_size): Likewise. Treat the array as non-const
10459 (emit_mode_adjustments): Handle adj_nunits.
10461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10463 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10464 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10465 (VECTOR_MODES): Use it.
10466 (make_vector_modes): Take the prefix as an argument.
10468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10469 Alan Hayward <alan.hayward@arm.com>
10470 David Sherwood <david.sherwood@arm.com>
10472 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10473 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10474 for MODE_VECTOR_BOOL.
10475 * machmode.def (VECTOR_BOOL_MODE): Document.
10476 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10477 (make_vector_bool_mode): New function.
10478 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10480 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10481 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10483 * stor-layout.c (int_mode_for_mode): Likewise.
10484 * tree.c (build_vector_type_for_mode): Likewise.
10485 * varasm.c (output_constant_pool_2): Likewise.
10486 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10487 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10488 for MODE_VECTOR_BOOL.
10489 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10490 of mode class checks.
10491 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10492 instead of a list of mode class checks.
10493 (expand_vector_scalar_condition): Likewise.
10494 (type_for_widest_vector_mode): Handle BImode as an inner 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 * machmode.h (mode_size): Change from unsigned short to
10502 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10503 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10504 or if measurement_type is not polynomial.
10505 (fixed_size_mode::includes_p): Check for constant-sized modes.
10506 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10507 return a poly_uint16 rather than an unsigned short.
10508 (emit_mode_size): Change the type of mode_size from unsigned short
10509 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10510 (emit_mode_adjustments): Cope with polynomial vector sizes.
10511 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10513 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10515 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10516 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10517 * caller-save.c (setup_save_areas): Likewise.
10518 (replace_reg_with_saved_mem): Likewise.
10519 * calls.c (emit_library_call_value_1): Likewise.
10520 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10521 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10522 (gen_lowpart_for_combine): Likewise.
10523 * convert.c (convert_to_integer_1): Likewise.
10524 * cse.c (equiv_constant, cse_insn): Likewise.
10525 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10526 (cselib_subst_to_values): Likewise.
10527 * dce.c (word_dce_process_block): Likewise.
10528 * df-problems.c (df_word_lr_mark_ref): Likewise.
10529 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10530 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10531 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10532 (rtl_for_decl_location): Likewise.
10533 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10534 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10535 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10536 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10537 (expand_expr_real_1): Likewise.
10538 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10539 (pad_below): Likewise.
10540 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10541 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10542 * ira.c (get_subreg_tracking_sizes): Likewise.
10543 * ira-build.c (ira_create_allocno_objects): Likewise.
10544 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10545 (ira_sort_regnos_for_alter_reg): Likewise.
10546 * ira-costs.c (record_operand_costs): Likewise.
10547 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10548 (resolve_simple_move): Likewise.
10549 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10550 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10551 (lra_constraints): Likewise.
10552 (CONST_POOL_OK_P): Reject variable-sized modes.
10553 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10554 (add_pseudo_to_slot, lra_spill): Likewise.
10555 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10556 * optabs-query.c (get_best_extraction_insn): Likewise.
10557 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10558 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10559 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10560 * recog.c (offsettable_address_addr_space_p): Likewise.
10561 * regcprop.c (maybe_mode_change): Likewise.
10562 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10563 * regrename.c (build_def_use): Likewise.
10564 * regstat.c (dump_reg_info): Likewise.
10565 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10566 (find_reloads, find_reloads_subreg_address): Likewise.
10567 * reload1.c (eliminate_regs_1): Likewise.
10568 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10569 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10570 (simplify_binary_operation_1, simplify_subreg): Likewise.
10571 * targhooks.c (default_function_arg_padding): Likewise.
10572 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10573 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10574 (verify_gimple_assign_ternary): Likewise.
10575 * tree-inline.c (estimate_move_cost): Likewise.
10576 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10577 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10578 (get_address_cost_ainc): Likewise.
10579 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10580 (vect_supportable_dr_alignment): Likewise.
10581 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10582 (vectorizable_reduction): Likewise.
10583 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10584 (vectorizable_operation, vectorizable_load): Likewise.
10585 * tree.c (build_same_sized_truth_vector_type): Likewise.
10586 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10587 * var-tracking.c (emit_note_insn_var_location): Likewise.
10588 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10589 (ADDR_VEC_ALIGN): Likewise.
10591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10592 Alan Hayward <alan.hayward@arm.com>
10593 David Sherwood <david.sherwood@arm.com>
10595 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10597 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10598 or if measurement_type is polynomial.
10599 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10600 * combine.c (make_extraction): Likewise.
10601 * dse.c (find_shift_sequence): Likewise.
10602 * dwarf2out.c (mem_loc_descriptor): Likewise.
10603 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10604 (extract_bit_field, extract_low_bits): Likewise.
10605 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10606 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10607 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10608 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10609 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10610 * reload.c (find_reloads): Likewise.
10611 * reload1.c (alter_reg): Likewise.
10612 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10613 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10614 * tree-if-conv.c (predicate_mem_writes): Likewise.
10615 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10616 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10617 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10618 * valtrack.c (dead_debug_insert_temp): Likewise.
10619 * varasm.c (mergeable_constant_section): Likewise.
10620 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10622 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10623 Alan Hayward <alan.hayward@arm.com>
10624 David Sherwood <david.sherwood@arm.com>
10626 * expr.c (expand_assignment): Cope with polynomial mode sizes
10627 when assigning to a CONCAT.
10629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10630 Alan Hayward <alan.hayward@arm.com>
10631 David Sherwood <david.sherwood@arm.com>
10633 * machmode.h (mode_precision): Change from unsigned short to
10635 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10637 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10638 or if measurement_type is not polynomial.
10639 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10640 in which the mode is already known to be a scalar_int_mode.
10641 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10642 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10644 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10645 for GET_MODE_PRECISION.
10646 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10647 for GET_MODE_PRECISION.
10648 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10650 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10651 (expand_field_assignment, make_extraction): Likewise.
10652 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10653 (get_last_value): Likewise.
10654 * convert.c (convert_to_integer_1): Likewise.
10655 * cse.c (cse_insn): Likewise.
10656 * expr.c (expand_expr_real_1): Likewise.
10657 * lra-constraints.c (simplify_operand_subreg): Likewise.
10658 * optabs-query.c (can_atomic_load_p): Likewise.
10659 * optabs.c (expand_atomic_load): Likewise.
10660 (expand_atomic_store): Likewise.
10661 * ree.c (combine_reaching_defs): Likewise.
10662 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10663 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10664 * tree.h (type_has_mode_precision_p): Likewise.
10665 * ubsan.c (instrument_si_overflow): Likewise.
10667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10668 Alan Hayward <alan.hayward@arm.com>
10669 David Sherwood <david.sherwood@arm.com>
10671 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10672 polynomial numbers of units.
10673 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10674 (valid_vector_subparts_p): New function.
10675 (build_vector_type): Remove temporary shim and take the number
10676 of units as a poly_uint64 rather than an int.
10677 (build_opaque_vector_type): Take the number of units as a
10678 poly_uint64 rather than an int.
10679 * tree.c (build_vector_from_ctor): Handle polynomial
10680 TYPE_VECTOR_SUBPARTS.
10681 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10682 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10683 (build_vector_from_val): If the number of units is variable,
10684 use build_vec_duplicate_cst for constant operands and
10685 VEC_DUPLICATE_EXPR otherwise.
10686 (make_vector_type): Remove temporary is_constant ().
10687 (build_vector_type, build_opaque_vector_type): Take the number of
10688 units as a poly_uint64 rather than an int.
10689 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10691 * cfgexpand.c (expand_debug_expr): Likewise.
10692 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10693 (store_constructor, expand_expr_real_1): Likewise.
10694 (const_scalar_mask_from_tree): Likewise.
10695 * fold-const-call.c (fold_const_reduction): Likewise.
10696 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10697 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10698 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10699 (fold_relational_const): Likewise.
10700 (native_interpret_vector): Likewise. Change the size from an
10701 int to an unsigned int.
10702 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10703 TYPE_VECTOR_SUBPARTS.
10704 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10705 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10706 duplicating a non-constant operand into a variable-length vector.
10707 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10708 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10709 * ipa-icf.c (sem_variable::equals): Likewise.
10710 * match.pd: Likewise.
10711 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10712 * print-tree.c (print_node): Likewise.
10713 * stor-layout.c (layout_type): Likewise.
10714 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10715 * tree-cfg.c (verify_gimple_comparison): Likewise.
10716 (verify_gimple_assign_binary): Likewise.
10717 (verify_gimple_assign_ternary): Likewise.
10718 (verify_gimple_assign_single): Likewise.
10719 * tree-pretty-print.c (dump_generic_node): Likewise.
10720 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10721 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10722 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10723 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10724 (vect_shift_permute_load_chain): Likewise.
10725 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10726 (expand_vector_condition, optimize_vector_constructor): Likewise.
10727 (lower_vec_perm, get_compute_type): Likewise.
10728 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10729 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10730 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10731 (vect_recog_mask_conversion_pattern): Likewise.
10732 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10733 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10734 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10735 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10736 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10737 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10738 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10739 (supportable_widening_operation): Likewise.
10740 (supportable_narrowing_operation): Likewise.
10741 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10743 * varasm.c (output_constant): Likewise.
10745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10746 Alan Hayward <alan.hayward@arm.com>
10747 David Sherwood <david.sherwood@arm.com>
10749 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10750 so that both the length == 3 and length != 3 cases set up their
10751 own permute vectors. Add comments explaining why we know the
10752 number of elements is constant.
10753 (vect_permute_load_chain): Likewise.
10755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10756 Alan Hayward <alan.hayward@arm.com>
10757 David Sherwood <david.sherwood@arm.com>
10759 * machmode.h (mode_nunits): Change from unsigned char to
10761 (ONLY_FIXED_SIZE_MODES): New macro.
10762 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10763 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10764 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10766 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10767 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10768 or if measurement_type is not polynomial.
10769 * genmodes.c (ZERO_COEFFS): New macro.
10770 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10772 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10773 Use ZERO_COEFFS when emitting initializers.
10774 * data-streamer.h (bp_pack_poly_value): New function.
10775 (bp_unpack_poly_value): Likewise.
10776 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10777 for GET_MODE_NUNITS.
10778 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10779 for GET_MODE_NUNITS.
10780 * tree.c (make_vector_type): Remove temporary shim and make
10781 the real function take the number of units as a poly_uint64
10782 rather than an int.
10783 (build_vector_type_for_mode): Handle polynomial nunits.
10784 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10785 * emit-rtl.c (const_vec_series_p_1): Likewise.
10786 (gen_rtx_CONST_VECTOR): Likewise.
10787 * fold-const.c (test_vec_duplicate_folding): Likewise.
10788 * genrecog.c (validate_pattern): Likewise.
10789 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10790 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10791 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10792 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10793 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10794 * rtlanal.c (subreg_get_info): Likewise.
10795 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10796 (vect_grouped_load_supported): Likewise.
10797 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10798 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10799 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10800 (simplify_const_unary_operation, simplify_binary_operation_1)
10801 (simplify_const_binary_operation, simplify_ternary_operation)
10802 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10803 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10804 instead of CONST_VECTOR_NUNITS.
10805 * varasm.c (output_constant_pool_2): Likewise.
10806 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10807 explicit-encoded elements in the XVEC for variable-length vectors.
10809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10811 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10814 Alan Hayward <alan.hayward@arm.com>
10815 David Sherwood <david.sherwood@arm.com>
10817 * coretypes.h (fixed_size_mode): Declare.
10818 (fixed_size_mode_pod): New typedef.
10819 * builtins.h (target_builtins::x_apply_args_mode)
10820 (target_builtins::x_apply_result_mode): Change type to
10821 fixed_size_mode_pod.
10822 * builtins.c (apply_args_size, apply_result_size, result_vector)
10823 (expand_builtin_apply_args_1, expand_builtin_apply)
10824 (expand_builtin_return): Update accordingly.
10826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10828 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10829 * cselib.c (cselib_hash_rtx): Likewise.
10830 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10831 CONST_VECTOR encoding.
10833 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10834 Jeff Law <law@redhat.com>
10837 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10838 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10839 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10840 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10843 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10844 explicitly probe *sp in a noreturn function if there were any callee
10845 register saves or frame pointer is needed.
10847 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10850 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10851 BLKmode for ternary, binary or unary expressions.
10854 * var-tracking.c (delete_vta_debug_insn): New inline function.
10855 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10856 insns from get_insns () to NULL instead of each bb separately.
10857 Use delete_vta_debug_insn. No longer static.
10858 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10859 delete_vta_debug_insns callers.
10860 * rtl.h (delete_vta_debug_insns): Declare.
10861 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10862 instead of variable_tracking_main.
10864 2018-01-03 Martin Sebor <msebor@redhat.com>
10866 PR tree-optimization/83603
10867 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10868 arguments past the endof the argument list in functions declared
10869 without a prototype.
10870 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10871 Avoid checking when arguments are null.
10873 2018-01-03 Martin Sebor <msebor@redhat.com>
10876 * doc/extend.texi (attribute const): Fix a typo.
10877 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10878 issuing -Wsuggest-attribute for void functions.
10880 2018-01-03 Martin Sebor <msebor@redhat.com>
10882 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10883 offset_int::from instead of wide_int::to_shwi.
10884 (maybe_diag_overlap): Remove assertion.
10885 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10886 * gimple-ssa-sprintf.c (format_directive): Same.
10887 (parse_directive): Same.
10888 (sprintf_dom_walker::compute_format_length): Same.
10889 (try_substitute_return_value): Same.
10891 2018-01-03 Jeff Law <law@redhat.com>
10893 PR middle-end/83654
10894 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10895 non-constant residual for zero at runtime and avoid probing in
10896 that case. Reorganize code for trailing problem to mirror handling
10899 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10901 PR tree-optimization/83501
10902 * tree-ssa-strlen.c (get_string_cst): New.
10903 (handle_char_store): Call get_string_cst.
10905 2018-01-03 Martin Liska <mliska@suse.cz>
10907 PR tree-optimization/83593
10908 * tree-ssa-strlen.c: Include tree-cfg.h.
10909 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10910 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10911 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10913 (strlen_dom_walker::before_dom_children): Call
10914 gimple_purge_dead_eh_edges. Dump tranformation with details
10916 (strlen_dom_walker::before_dom_children): Update call by adding
10917 new argument cleanup_eh.
10918 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10920 2018-01-03 Martin Liska <mliska@suse.cz>
10923 * cif-code.def (VARIADIC_THUNK): New enum value.
10924 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10927 2018-01-03 Jan Beulich <jbeulich@suse.com>
10929 * sse.md (mov<mode>_internal): Tighten condition for when to use
10930 vmovdqu<ssescalarsize> for TI and OI modes.
10932 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10934 Update copyright years.
10936 2018-01-03 Martin Liska <mliska@suse.cz>
10939 * ipa-visibility.c (function_and_variable_visibility): Skip
10940 functions with noipa attribure.
10942 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10944 * gcc.c (process_command): Update copyright notice dates.
10945 * gcov-dump.c (print_version): Ditto.
10946 * gcov.c (print_version): Ditto.
10947 * gcov-tool.c (print_version): Ditto.
10948 * gengtype.c (create_file): Ditto.
10949 * doc/cpp.texi: Bump @copying's copyright year.
10950 * doc/cppinternals.texi: Ditto.
10951 * doc/gcc.texi: Ditto.
10952 * doc/gccint.texi: Ditto.
10953 * doc/gcov.texi: Ditto.
10954 * doc/install.texi: Ditto.
10955 * doc/invoke.texi: Ditto.
10957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10959 * vector-builder.h (vector_builder::m_full_nelts): Change from
10960 unsigned int to poly_uint64.
10961 (vector_builder::full_nelts): Update prototype accordingly.
10962 (vector_builder::new_vector): Likewise.
10963 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10964 (vector_builder::operator ==): Likewise.
10965 (vector_builder::finalize): Likewise.
10966 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10967 Take the number of elements as a poly_uint64 rather than an
10969 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10970 from unsigned int to poly_uint64.
10971 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10972 (vec_perm_indices::new_vector): Likewise.
10973 (vec_perm_indices::length): Likewise.
10974 (vec_perm_indices::nelts_per_input): Likewise.
10975 (vec_perm_indices::input_nelts): Likewise.
10976 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10977 number of elements per input as a poly_uint64 rather than an
10978 unsigned int. Use the original encoding for variable-length
10979 vectors, rather than clamping each individual element.
10980 For the second and subsequent elements in each pattern,
10981 clamp the step and base before clamping their sum.
10982 (vec_perm_indices::series_p): Handle polynomial element counts.
10983 (vec_perm_indices::all_in_range_p): Likewise.
10984 (vec_perm_indices_to_tree): Likewise.
10985 (vec_perm_indices_to_rtx): Likewise.
10986 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10987 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10988 (tree_vector_builder::new_binary_operation): Handle polynomial
10989 element counts. Return false if we need to know the number
10990 of elements at compile time.
10991 * fold-const.c (fold_vec_perm): Punt if the number of elements
10992 isn't known at compile time.
10994 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10996 * vec-perm-indices.h (vec_perm_builder): Change element type
10997 from HOST_WIDE_INT to poly_int64.
10998 (vec_perm_indices::element_type): Update accordingly.
10999 (vec_perm_indices::clamp): Handle polynomial element_types.
11000 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11001 (vec_perm_indices::all_in_range_p): Likewise.
11002 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11004 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11005 polynomial vec_perm_indices element types.
11006 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11007 * fold-const.c (fold_vec_perm): Likewise.
11008 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11009 * tree-vect-generic.c (lower_vec_perm): Likewise.
11010 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11011 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11012 element type to HOST_WIDE_INT.
11014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11015 Alan Hayward <alan.hayward@arm.com>
11016 David Sherwood <david.sherwood@arm.com>
11018 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11019 rather than an int. Use plus_constant.
11020 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11021 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11024 Alan Hayward <alan.hayward@arm.com>
11025 David Sherwood <david.sherwood@arm.com>
11027 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11028 a HOST_WIDE_INT to a poly_int64.
11030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11031 Alan Hayward <alan.hayward@arm.com>
11032 David Sherwood <david.sherwood@arm.com>
11034 * calls.c (load_register_parameters): Cope with polynomial
11035 mode sizes. Require a constant size for BLKmode parameters
11036 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11037 forces a parameter to be padded at the lsb end in order to
11038 fill a complete number of words, require the parameter size
11039 to be ordered wrt UNITS_PER_WORD.
11041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11042 Alan Hayward <alan.hayward@arm.com>
11043 David Sherwood <david.sherwood@arm.com>
11045 * reload1.c (spill_stack_slot_width): Change element type
11046 from unsigned int to poly_uint64_pod.
11047 (alter_reg): Treat mode sizes as polynomial.
11049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11050 Alan Hayward <alan.hayward@arm.com>
11051 David Sherwood <david.sherwood@arm.com>
11053 * reload.c (complex_word_subreg_p): New function.
11054 (reload_inner_reg_of_subreg, push_reload): Use it.
11056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11057 Alan Hayward <alan.hayward@arm.com>
11058 David Sherwood <david.sherwood@arm.com>
11060 * lra-constraints.c (process_alt_operands): Reject matched
11061 operands whose sizes aren't ordered.
11062 (match_reload): Refer to this check here.
11064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11065 Alan Hayward <alan.hayward@arm.com>
11066 David Sherwood <david.sherwood@arm.com>
11068 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11069 that the mode size is in the set {1, 2, 4, 8, 16}.
11071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11072 Alan Hayward <alan.hayward@arm.com>
11073 David Sherwood <david.sherwood@arm.com>
11075 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11076 Use plus_constant instead of gen_rtx_PLUS.
11078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11079 Alan Hayward <alan.hayward@arm.com>
11080 David Sherwood <david.sherwood@arm.com>
11082 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11083 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11084 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11085 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11086 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11087 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11088 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11089 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11090 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11091 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11093 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11094 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11095 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11096 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11097 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11098 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11099 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11100 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11101 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11102 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11104 * expr.c (emit_move_resolve_push): Treat the input and result
11105 of PUSH_ROUNDING as a poly_int64.
11106 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11107 (emit_push_insn): Likewise.
11108 * lra-eliminations.c (mark_not_eliminable): Likewise.
11109 * recog.c (push_operand): Likewise.
11110 * reload1.c (elimination_effects): Likewise.
11111 * rtlanal.c (nonzero_bits1): Likewise.
11112 * calls.c (store_one_arg): Likewise. Require the padding to be
11113 known at compile time.
11115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11116 Alan Hayward <alan.hayward@arm.com>
11117 David Sherwood <david.sherwood@arm.com>
11119 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11120 Use plus_constant instead of gen_rtx_PLUS.
11122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11123 Alan Hayward <alan.hayward@arm.com>
11124 David Sherwood <david.sherwood@arm.com>
11126 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11127 rather than an int.
11129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11130 Alan Hayward <alan.hayward@arm.com>
11131 David Sherwood <david.sherwood@arm.com>
11133 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11134 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11135 via stack temporaries. Treat the mode size as polynomial too.
11137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11138 Alan Hayward <alan.hayward@arm.com>
11139 David Sherwood <david.sherwood@arm.com>
11141 * expr.c (expand_expr_real_2): When handling conversions involving
11142 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11143 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11144 as a poly_uint64 too.
11146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11147 Alan Hayward <alan.hayward@arm.com>
11148 David Sherwood <david.sherwood@arm.com>
11150 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11153 Alan Hayward <alan.hayward@arm.com>
11154 David Sherwood <david.sherwood@arm.com>
11156 * combine.c (can_change_dest_mode): Handle polynomial
11157 REGMODE_NATURAL_SIZE.
11158 * expmed.c (store_bit_field_1): Likewise.
11159 * expr.c (store_constructor): Likewise.
11160 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11161 and polynomial REGMODE_NATURAL_SIZE.
11162 (gen_lowpart_common): Likewise.
11163 * reginfo.c (record_subregs_of_mode): Likewise.
11164 * rtlanal.c (read_modify_subreg_p): Likewise.
11166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11167 Alan Hayward <alan.hayward@arm.com>
11168 David Sherwood <david.sherwood@arm.com>
11170 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11171 numbers of elements.
11173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11174 Alan Hayward <alan.hayward@arm.com>
11175 David Sherwood <david.sherwood@arm.com>
11177 * match.pd: Cope with polynomial numbers of vector elements.
11179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11180 Alan Hayward <alan.hayward@arm.com>
11181 David Sherwood <david.sherwood@arm.com>
11183 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11184 in a POINTER_PLUS_EXPR.
11186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11187 Alan Hayward <alan.hayward@arm.com>
11188 David Sherwood <david.sherwood@arm.com>
11190 * omp-simd-clone.c (simd_clone_subparts): New function.
11191 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11192 (ipa_simd_modify_function_body): Likewise.
11194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11195 Alan Hayward <alan.hayward@arm.com>
11196 David Sherwood <david.sherwood@arm.com>
11198 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11199 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11200 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11201 (expand_vector_condition, vector_element): Likewise.
11202 (subparts_gt): New function.
11203 (get_compute_type): Use subparts_gt.
11204 (count_type_subparts): Delete.
11205 (expand_vector_operations_1): Use subparts_gt instead of
11206 count_type_subparts.
11208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11209 Alan Hayward <alan.hayward@arm.com>
11210 David Sherwood <david.sherwood@arm.com>
11212 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11213 (vect_compile_time_alias): ...this new function. Do the calculation
11214 on poly_ints rather than trees.
11215 (vect_prune_runtime_alias_test_list): Update call accordingly.
11217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11218 Alan Hayward <alan.hayward@arm.com>
11219 David Sherwood <david.sherwood@arm.com>
11221 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11223 (vect_schedule_slp_instance): Likewise.
11225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11226 Alan Hayward <alan.hayward@arm.com>
11227 David Sherwood <david.sherwood@arm.com>
11229 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11230 constant and extern definitions for variable-length vectors.
11231 (vect_get_constant_vectors): Note that the number of units
11232 is known to be constant.
11234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11235 Alan Hayward <alan.hayward@arm.com>
11236 David Sherwood <david.sherwood@arm.com>
11238 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11239 of units as polynomial. Choose between WIDE and NARROW based
11242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11243 Alan Hayward <alan.hayward@arm.com>
11244 David Sherwood <david.sherwood@arm.com>
11246 * tree-vect-stmts.c (simd_clone_subparts): New function.
11247 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11250 Alan Hayward <alan.hayward@arm.com>
11251 David Sherwood <david.sherwood@arm.com>
11253 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11254 vectors as polynomial. Use build_index_vector for
11255 IFN_GOMP_SIMD_LANE.
11257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11258 Alan Hayward <alan.hayward@arm.com>
11259 David Sherwood <david.sherwood@arm.com>
11261 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11262 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11263 for variable-length vectors.
11264 (vectorizable_mask_load_store): Treat the number of units as
11265 polynomial, asserting that it is constant if the condition has
11266 already been enforced.
11267 (vectorizable_store, vectorizable_load): Likewise.
11269 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11270 Alan Hayward <alan.hayward@arm.com>
11271 David Sherwood <david.sherwood@arm.com>
11273 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11274 of units as polynomial. Punt if we can't tell at compile time
11275 which vector contains the final result.
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 * tree-vect-loop.c (vectorizable_induction): Treat the number
11282 of units as polynomial. Punt on SLP inductions. Use an integer
11283 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11284 cast of such a series for variable-length floating-point
11287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11288 Alan Hayward <alan.hayward@arm.com>
11289 David Sherwood <david.sherwood@arm.com>
11291 * tree.h (build_index_vector): Declare.
11292 * tree.c (build_index_vector): New function.
11293 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11294 of units as polynomial, forcibly converting it to a constant if
11295 vectorizable_reduction has already enforced the condition.
11296 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11297 to create a {1,2,3,...} vector.
11298 (vectorizable_reduction): Treat the number of units as polynomial.
11299 Choose vectype_in based on the largest scalar element size rather
11300 than the smallest number of units. Enforce the restrictions
11303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11304 Alan Hayward <alan.hayward@arm.com>
11305 David Sherwood <david.sherwood@arm.com>
11307 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11308 number of units as polynomial.
11310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11311 Alan Hayward <alan.hayward@arm.com>
11312 David Sherwood <david.sherwood@arm.com>
11314 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11315 * target.def (autovectorize_vector_sizes): Return the vector sizes
11316 by pointer, using vector_sizes rather than a bitmask.
11317 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11318 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11319 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11321 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11322 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11323 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11324 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11325 * omp-general.c (omp_max_vf): Likewise.
11326 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11327 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11328 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11329 * tree-vect-slp.c (vect_slp_bb): Likewise.
11330 * doc/tm.texi: Regenerate.
11331 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11333 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11334 the vector size as a poly_uint64 rather than an unsigned int.
11335 (current_vector_size): Change from an unsigned int to a poly_uint64.
11336 (get_vectype_for_scalar_type): Update accordingly.
11337 * tree.h (build_truth_vector_type): Take the size and number of
11338 units as a poly_uint64 rather than an unsigned int.
11339 (build_vector_type): Add a temporary overload that takes
11340 the number of units as a poly_uint64 rather than an unsigned int.
11341 * tree.c (make_vector_type): Likewise.
11342 (build_truth_vector_type): Take the number of units as a poly_uint64
11343 rather than an unsigned int.
11345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11346 Alan Hayward <alan.hayward@arm.com>
11347 David Sherwood <david.sherwood@arm.com>
11349 * target.def (get_mask_mode): Take the number of units and length
11350 as poly_uint64s rather than unsigned ints.
11351 * targhooks.h (default_get_mask_mode): Update accordingly.
11352 * targhooks.c (default_get_mask_mode): Likewise.
11353 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11354 * doc/tm.texi: Regenerate.
11356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11357 Alan Hayward <alan.hayward@arm.com>
11358 David Sherwood <david.sherwood@arm.com>
11360 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11361 * omp-general.c (omp_max_vf): Likewise.
11362 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11363 (expand_omp_simd): Handle polynomial safelen.
11364 * omp-low.c (omplow_simd_context): Add a default constructor.
11365 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11366 (lower_rec_simd_input_clauses): Update accordingly.
11367 (lower_rec_input_clauses): Likewise.
11369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11370 Alan Hayward <alan.hayward@arm.com>
11371 David Sherwood <david.sherwood@arm.com>
11373 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11374 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11375 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11376 (vect_analyze_slp_cost): Likewise.
11377 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11378 (vect_model_load_cost): Likewise.
11380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11381 Alan Hayward <alan.hayward@arm.com>
11382 David Sherwood <david.sherwood@arm.com>
11384 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11385 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11386 from an unsigned int * to a poly_uint64_pod *.
11387 (calculate_unrolling_factor): New function.
11388 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11391 Alan Hayward <alan.hayward@arm.com>
11392 David Sherwood <david.sherwood@arm.com>
11394 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11395 from an unsigned int to a poly_uint64.
11396 (_loop_vec_info::slp_unrolling_factor): Likewise.
11397 (_loop_vec_info::vectorization_factor): Change from an int
11399 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11400 (vect_get_num_vectors): New function.
11401 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11402 (vect_get_num_copies): Use vect_get_num_vectors.
11403 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11404 to an unsigned int *.
11405 (vect_analyze_data_refs): Change min_vf from an int * to a
11407 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11408 than an unsigned HOST_WIDE_INT.
11409 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11410 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11411 to an unsigned int *.
11412 (vect_analyze_data_ref_dependences): Likewise.
11413 (vect_compute_data_ref_alignment): Handle polynomial vf.
11414 (vect_enhance_data_refs_alignment): Likewise.
11415 (vect_prune_runtime_alias_test_list): Likewise.
11416 (vect_shift_permute_load_chain): Likewise.
11417 (vect_supportable_dr_alignment): Likewise.
11418 (dependence_distance_ge_vf): Take the vectorization factor as a
11419 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11420 (vect_analyze_data_refs): Change min_vf from an int * to a
11422 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11423 vfm1 as a poly_uint64 rather than an int. Make the same change
11424 for the returned bound_scalar.
11425 (vect_gen_vector_loop_niters): Handle polynomial vf.
11426 (vect_do_peeling): Likewise. Update call to
11427 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11428 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11430 * tree-vect-loop.c (vect_determine_vectorization_factor)
11431 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11432 (vect_get_known_peeling_cost): Likewise.
11433 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11434 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11435 (vect_transform_loop): Likewise. Use the lowest possible VF when
11436 updating the upper bounds of the loop.
11437 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11438 rather than an int.
11439 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11440 polynomial unroll factors.
11441 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11442 (vect_make_slp_decision): Likewise.
11443 (vect_supported_load_permutation_p): Likewise, and polynomial
11445 (vect_analyze_slp_cost): Handle polynomial vf.
11446 (vect_slp_analyze_node_operations): Likewise.
11447 (vect_slp_analyze_bb_1): Likewise.
11448 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11449 than an unsigned HOST_WIDE_INT.
11450 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11451 (vectorizable_load): Handle polynomial vf.
11452 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11454 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11457 Alan Hayward <alan.hayward@arm.com>
11458 David Sherwood <david.sherwood@arm.com>
11460 * match.pd: Handle bit operations involving three constants
11461 and try to fold one pair.
11463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11465 * tree-vect-loop-manip.c: Include gimple-fold.h.
11466 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11467 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11468 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11469 Add a path that uses a step of VF instead of 1, but disable it
11471 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11472 and niters_no_overflow parameters. Update calls to
11473 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11474 Create a new SSA name if the latter choses to use a ste other
11475 than zero, and return it via niters_vector_mult_vf_var.
11476 * tree-vect-loop.c (vect_transform_loop): Update calls to
11477 vect_do_peeling, vect_gen_vector_loop_niters and
11478 slpeel_make_loop_iterate_ntimes.
11479 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11480 (vect_gen_vector_loop_niters): Update declarations after above changes.
11482 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11484 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11485 128-bit round to integer instructions.
11486 (ceil<mode>2): Likewise.
11487 (btrunc<mode>2): Likewise.
11488 (round<mode>2): Likewise.
11490 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11492 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11493 unaligned VSX load/store on P8/P9.
11494 (expand_block_clear): Allow the use of unaligned VSX
11495 load/store on P8/P9.
11497 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11499 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11501 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11502 swap associated with both a load and a store.
11504 2018-01-02 Andrew Waterman <andrew@sifive.com>
11506 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11507 * config/riscv/riscv.md (clear_cache): Use it.
11509 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11511 * web.c: Remove out-of-date comment.
11513 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11515 * expr.c (fixup_args_size_notes): Check that any existing
11516 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11517 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11518 (emit_single_push_insn): ...here.
11520 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11522 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11523 (const_vector_encoded_nelts): New function.
11524 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11525 (const_vector_int_elt, const_vector_elt): Declare.
11526 * emit-rtl.c (const_vector_int_elt_1): New function.
11527 (const_vector_elt): Likewise.
11528 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11529 of CONST_VECTOR_ELT.
11531 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11533 * expr.c: Include rtx-vector-builder.h.
11534 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11535 directly on the tree encoding.
11536 (const_vector_from_tree): Likewise.
11537 * optabs.c: Include rtx-vector-builder.h.
11538 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11539 sequence of "u" values.
11540 * vec-perm-indices.c: Include rtx-vector-builder.h.
11541 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11542 directly on the vec_perm_indices encoding.
11544 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11546 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11547 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11548 * rtx-vector-builder.h: New file.
11549 * rtx-vector-builder.c: Likewise.
11550 * rtl.h (rtx_def::u2): Add a const_vector field.
11551 (CONST_VECTOR_NPATTERNS): New macro.
11552 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11553 (CONST_VECTOR_DUPLICATE_P): Likewise.
11554 (CONST_VECTOR_STEPPED_P): Likewise.
11555 (CONST_VECTOR_ENCODED_ELT): Likewise.
11556 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11557 (unwrap_const_vec_duplicate): Likewise.
11558 (const_vec_series_p): Check for a non-duplicated vector encoding.
11559 Say that the function only returns true for integer vectors.
11560 * emit-rtl.c: Include rtx-vector-builder.h.
11561 (gen_const_vec_duplicate_1): Delete.
11562 (gen_const_vector): Call gen_const_vec_duplicate instead of
11563 gen_const_vec_duplicate_1.
11564 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11565 (gen_const_vec_duplicate): Use rtx_vector_builder.
11566 (gen_const_vec_series): Likewise.
11567 (gen_rtx_CONST_VECTOR): Likewise.
11568 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11569 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11570 Build a new vector rather than modifying a CONST_VECTOR in-place.
11571 (handle_special_swappables): Update call accordingly.
11572 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11573 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11574 Build a new vector rather than modifying a CONST_VECTOR in-place.
11575 (handle_special_swappables): Update call accordingly.
11577 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11579 * simplify-rtx.c (simplify_const_binary_operation): Use
11580 CONST_VECTOR_ELT instead of XVECEXP.
11582 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11584 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11585 the selector elements to be different from the data elements
11586 if the selector is a VECTOR_CST.
11587 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11588 ssizetype for the selector.
11590 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11592 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11593 before testing each element individually.
11594 * tree-vect-generic.c (lower_vec_perm): Likewise.
11596 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11598 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11599 * selftest-run-tests.c (selftest::run_tests): Call it.
11600 * vector-builder.h (vector_builder::operator ==): New function.
11601 (vector_builder::operator !=): Likewise.
11602 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11603 (vec_perm_indices::all_from_input_p): New function.
11604 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11605 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11606 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11607 instead of reading the VECTOR_CST directly. Detect whether both
11608 vector inputs are the same before constructing the vec_perm_indices,
11609 and update the number of inputs argument accordingly. Use the
11610 utility functions added above. Only construct sel2 if we need to.
11612 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11614 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11615 the broadcast of the low byte.
11616 (expand_mult_highpart): Use an explicit encoding for the permutes.
11617 * optabs-query.c (can_mult_highpart_p): Likewise.
11618 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11619 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11620 (vectorizable_bswap): Likewise.
11621 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11622 explicit encoding for the power-of-2 permutes.
11623 (vect_permute_store_chain): Likewise.
11624 (vect_grouped_load_supported): Likewise.
11625 (vect_permute_load_chain): Likewise.
11627 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11629 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11630 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11631 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11632 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11633 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11634 (vect_gen_perm_mask_any): Likewise.
11636 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11638 * int-vector-builder.h: New file.
11639 * vec-perm-indices.h: Include int-vector-builder.h.
11640 (vec_perm_indices): Redefine as an int_vector_builder.
11641 (auto_vec_perm_indices): Delete.
11642 (vec_perm_builder): Redefine as a stand-alone class.
11643 (vec_perm_indices::vec_perm_indices): New function.
11644 (vec_perm_indices::clamp): Likewise.
11645 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11646 (vec_perm_indices::new_vector): New function.
11647 (vec_perm_indices::new_expanded_vector): Update for new
11648 vec_perm_indices class.
11649 (vec_perm_indices::rotate_inputs): New function.
11650 (vec_perm_indices::all_in_range_p): Operate directly on the
11651 encoded form, without computing elided elements.
11652 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11653 encoding. Update for new vec_perm_indices class.
11654 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11655 the given vec_perm_builder.
11656 (expand_vec_perm_var): Update vec_perm_builder constructor.
11657 (expand_mult_highpart): Use vec_perm_builder instead of
11658 auto_vec_perm_indices.
11659 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11660 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11661 or double series encoding as appropriate.
11662 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11663 vec_perm_indices instead of auto_vec_perm_indices.
11664 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11665 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11666 (vect_permute_store_chain): Likewise.
11667 (vect_grouped_load_supported): Likewise.
11668 (vect_permute_load_chain): Likewise.
11669 (vect_shift_permute_load_chain): Likewise.
11670 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11671 (vect_transform_slp_perm_load): Likewise.
11672 (vect_schedule_slp_instance): Likewise.
11673 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11674 (vectorizable_mask_load_store): Likewise.
11675 (vectorizable_bswap): Likewise.
11676 (vectorizable_store): Likewise.
11677 (vectorizable_load): Likewise.
11678 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11679 vec_perm_indices instead of auto_vec_perm_indices. Use
11680 tree_to_vec_perm_builder to read the vector from a tree.
11681 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11682 vec_perm_builder instead of a vec_perm_indices.
11683 (have_whole_vector_shift): Use vec_perm_builder and
11684 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11685 truncation to calc_vec_perm_mask_for_shift.
11686 (vect_create_epilog_for_reduction): Likewise.
11687 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11688 from auto_vec_perm_indices to vec_perm_indices.
11689 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11690 instead of changing individual elements.
11691 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11692 the vector in d.perm.
11693 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11694 from auto_vec_perm_indices to vec_perm_indices.
11695 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11696 instead of changing individual elements.
11697 (arm_vectorize_vec_perm_const): Use new_vector to install
11698 the vector in d.perm.
11699 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11700 Update vec_perm_builder constructor.
11701 (rs6000_expand_interleave): Likewise.
11702 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11703 (rs6000_expand_interleave): Likewise.
11705 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11707 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11708 to qimode could truncate the indices.
11709 * optabs.c (expand_vec_perm_var): Likewise.
11711 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11713 * Makefile.in (OBJS): Add vec-perm-indices.o.
11714 * vec-perm-indices.h: New file.
11715 * vec-perm-indices.c: Likewise.
11716 * target.h (vec_perm_indices): Replace with a forward class
11718 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11719 * optabs.h: Include vec-perm-indices.h.
11720 (expand_vec_perm): Delete.
11721 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11722 (expand_vec_perm_const): Declare.
11723 * target.def (vec_perm_const_ok): Replace with...
11724 (vec_perm_const): ...this new hook.
11725 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11726 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11727 * doc/tm.texi: Regenerate.
11728 * optabs.def (vec_perm_const): Delete.
11729 * doc/md.texi (vec_perm_const): Likewise.
11730 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11731 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11732 expand_vec_perm for constant permutation vectors. Assert that
11733 the mode of variable permutation vectors is the integer equivalent
11734 of the mode that is being permuted.
11735 * optabs-query.h (selector_fits_mode_p): Declare.
11736 * optabs-query.c: Include vec-perm-indices.h.
11737 (selector_fits_mode_p): New function.
11738 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11739 is defined, instead of checking whether the vec_perm_const_optab
11740 exists. Use targetm.vectorize.vec_perm_const instead of
11741 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11742 fit in the vector mode before using a variable permute.
11743 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11744 vec_perm_indices instead of an rtx.
11745 (expand_vec_perm): Replace with...
11746 (expand_vec_perm_const): ...this new function. Take the selector
11747 as a vec_perm_indices rather than an rtx. Also take the mode of
11748 the selector. Update call to shift_amt_for_vec_perm_mask.
11749 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11750 Use vec_perm_indices::new_expanded_vector to expand the original
11751 selector into bytes. Check whether the indices fit in the vector
11752 mode before using a variable permute.
11753 (expand_vec_perm_var): Make global.
11754 (expand_mult_highpart): Use expand_vec_perm_const.
11755 * fold-const.c: Includes vec-perm-indices.h.
11756 * tree-ssa-forwprop.c: Likewise.
11757 * tree-vect-data-refs.c: Likewise.
11758 * tree-vect-generic.c: Likewise.
11759 * tree-vect-loop.c: Likewise.
11760 * tree-vect-slp.c: Likewise.
11761 * tree-vect-stmts.c: Likewise.
11762 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11764 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11765 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11766 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11767 (aarch64_vectorize_vec_perm_const): ...this new function.
11768 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11769 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11770 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11771 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11772 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11773 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11774 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11776 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11777 check for NEON modes.
11778 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11779 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11780 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11781 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11783 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11784 the old VEC_PERM_CONST conditions.
11785 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11786 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11787 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11788 (ia64_vectorize_vec_perm_const_ok): Merge into...
11789 (ia64_vectorize_vec_perm_const): ...this new function.
11790 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11791 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11792 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11793 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11794 * config/mips/mips.c (mips_expand_vec_perm_const)
11795 (mips_vectorize_vec_perm_const_ok): Merge into...
11796 (mips_vectorize_vec_perm_const): ...this new function.
11797 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11798 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11799 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11800 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11801 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11802 (rs6000_expand_vec_perm_const): Delete.
11803 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11805 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11806 (altivec_expand_vec_perm_const_le): Take each operand individually.
11807 Operate on constant selectors rather than rtxes.
11808 (altivec_expand_vec_perm_const): Likewise. Update call to
11809 altivec_expand_vec_perm_const_le.
11810 (rs6000_expand_vec_perm_const): Delete.
11811 (rs6000_vectorize_vec_perm_const_ok): Delete.
11812 (rs6000_vectorize_vec_perm_const): New function.
11813 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11814 an element count and rtx array.
11815 (rs6000_expand_extract_even): Update call accordingly.
11816 (rs6000_expand_interleave): Likewise.
11817 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11818 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11819 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11820 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11821 (rs6000_expand_vec_perm_const): Delete.
11822 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11823 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11824 (altivec_expand_vec_perm_const_le): Take each operand individually.
11825 Operate on constant selectors rather than rtxes.
11826 (altivec_expand_vec_perm_const): Likewise. Update call to
11827 altivec_expand_vec_perm_const_le.
11828 (rs6000_expand_vec_perm_const): Delete.
11829 (rs6000_vectorize_vec_perm_const_ok): Delete.
11830 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11831 reference to the SPE evmerge intructions.
11832 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11833 an element count and rtx array.
11834 (rs6000_expand_extract_even): Update call accordingly.
11835 (rs6000_expand_interleave): Likewise.
11836 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11837 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11839 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11841 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11843 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11844 vector mode and that that mode matches the mode of the data
11846 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11847 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11848 directly using expand_vec_perm_1 when forcing selectors into
11850 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11852 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11854 * optabs-query.h (can_vec_perm_p): Delete.
11855 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11856 * optabs-query.c (can_vec_perm_p): Split into...
11857 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11858 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11859 particular selector is valid.
11860 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11861 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11862 (vect_grouped_load_supported): Likewise.
11863 (vect_shift_permute_load_chain): Likewise.
11864 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11865 (vect_transform_slp_perm_load): Likewise.
11866 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11867 (vectorizable_bswap): Likewise.
11868 (vect_gen_perm_mask_checked): Likewise.
11869 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11870 implementations of variable permutation vectors into account
11871 when deciding which selector to use.
11872 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11873 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11874 with a false third argument.
11875 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11876 to test whether the constant selector is valid and can_vec_perm_var_p
11877 to test whether a variable selector is valid.
11879 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11881 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11882 * optabs-query.c (can_vec_perm_p): Likewise.
11883 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11884 instead of vec_perm_indices.
11885 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11886 (vect_gen_perm_mask_checked): Likewise,
11887 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11888 (vect_gen_perm_mask_checked): Likewise,
11890 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11892 * optabs-query.h (qimode_for_vec_perm): Declare.
11893 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11894 (qimode_for_vec_perm): ...this new function.
11895 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11897 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11899 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11900 does not have a conditional at the top.
11902 2018-01-02 Richard Biener <rguenther@suse.de>
11904 * ipa-inline.c (big_speedup_p): Fix expression.
11906 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11909 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11912 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11916 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11917 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11918 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11919 cond_taken_branch_cost 3->4.
11921 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11923 PR tree-optimization/83581
11924 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11925 TODO_cleanup_cfg if any changes have been made.
11927 PR middle-end/83608
11928 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11929 convert_modes if target mode has the right side, but different mode
11932 PR middle-end/83609
11933 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11934 last argument when extracting from CONCAT. If either from_real or
11935 from_imag is NULL, use expansion through memory. If result is not
11936 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11937 the parts directly to inner mode, if even that fails, use expansion
11940 PR middle-end/83623
11941 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11942 check for bswap in mode rather than HImode and use that in expand_unop
11945 Copyright (C) 2018 Free Software Foundation, Inc.
11947 Copying and distribution of this file, with or without modification,
11948 are permitted in any medium without royalty provided the copyright
11949 notice and this notice are preserved.