1 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3 PR tree-optimization/82665
4 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5 pointer subtraction where arguments come from a memchr call.
7 2018-05-01 Jakub Jelinek <jakub@redhat.com>
9 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
10 --push-state --as-needed and --pop-state instead of --as-needed and
11 --no-as-needed if ld supports it.
12 * configure: Regenerated.
15 * doc/install.texi2html: Replace _002d with - and _002a with * in
16 generated html files using sed.
18 2018-04-30 David Malcolm <dmalcolm@redhat.com>
21 * gcc-rich-location.c (blank_line_before_p): New function.
22 (use_new_line): New function.
23 (gcc_rich_location::add_fixit_insert_formatted): New function.
25 (gcc_rich_location::add_fixit_insert_formatted): New function.
27 2018-04-30 David Malcolm <dmalcolm@redhat.com>
29 * selftest.c (assert_streq): Rename "expected" and "actual" to
30 "val1" and "val2". Extend NULL-handling to cover both inputs
31 symmetrically, while still requiring both to be non-NULL for a pass.
32 * selftest.h (assert_streq): Rename "expected" and "actual" to
34 (ASSERT_EQ): Likewise.
35 (ASSERT_EQ_AT): Likewise.
36 (ASSERT_KNOWN_EQ): Likewise.
37 (ASSERT_KNOWN_EQ_AT): Likewise.
38 (ASSERT_NE): Likewise.
39 (ASSERT_MAYBE_NE): Likewise.
40 (ASSERT_MAYBE_NE_AT): Likewise.
41 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
42 the assertion to pass.
43 (ASSERT_STREQ_AT): Likewise.
45 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
47 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
48 interaction with -pie.
50 2018-04-30 David Malcolm <dmalcolm@redhat.com>
52 * selftest.h: Fix alphabetization of per-source-file selftest
55 2018-04-30 Jason Merrill <jason@redhat.com>
57 PR c++/61982 - dead stores to destroyed objects.
58 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
61 2018-04-30 Jason Merrill <jason@redhat.com>
63 * tree.c (build_clobber): New.
65 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
67 2018-04-30 David Malcolm <dmalcolm@redhat.com>
69 * diagnostic-show-locus.c (layout::layout): Update for
70 location_get_source_line returning a char_span.
71 (struct char_span): Move to input.h.
72 (struct correction): Update for fields in char_span becoming
74 (struct source_line): Update for location_get_source_line
75 returning a char_span.
76 (layout::print_line): Likewise.
77 * edit-context.c (edited_file::print_content): Likewise.
78 (edited_file::print_diff_hunk): Likewise.
79 (edited_file::print_run_of_changed_lines): Likewise.
80 (edited_file::get_num_lines): Likewise.
81 (edited_line::edited_line): Likewise.
82 * final.c (asm_show_source): Likewise.
83 * input.c (location_get_source_line): Convert return type
84 from const char * to char_span, losing the final "line_len"
86 (dump_location_info): Update for the above.
87 (get_substring_ranges_for_loc): Likewise. Use a char_span
88 when handling the literal within the line.
89 (test_reading_source_line): Update for location_get_source_line
90 returning a char_span.
91 * input.h (class char_span): Move here from
92 diagnostic-show-locus.c, converting from a struct to a class.
93 Make data members private.
94 (char_span::operator bool): New.
95 (char_span::length): New.
96 (char_span::get_buffer): New.
97 (char_span::operator[]): New.
98 (char_span::subspan): Make const.
99 (char_span::xstrdup): New.
100 (location_get_source_line): Convert return type from const char *
101 to char_span, losing the final "line_size" param.
103 2018-04-30 Jan Hubicka <jh@suse.cz>
105 * lto-wrapper.c (ltrans_priorities): New static var.
107 (run_gcc): Read priorities and if doing parallel build order
108 the Makefile by them.
110 2018-04-30 David Malcolm <dmalcolm@redhat.com>
112 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
114 2018-04-30 Richard Biener <rguenther@suse.de>
116 * tree-cfg.c (verify_address): Remove base argument, add
117 flag whether to check TREE_ADDRESSABLE and do that.
118 (verify_expr): Remove.
119 (verify_types_in_gimple_reference): Add pieces from verify_expr.
120 (verify_gimple_assign_single): Likewise.
121 (verify_gimple_switch): Likewise.
122 (verify_expr_location_1): Dereference tp once. Add (disabled)
123 piece from verify_expr.
124 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
126 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
128 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
130 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
132 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
133 (small_data_pattern): Likewise.
134 (arc_rewrite_small_data): Likewise.
135 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
136 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
137 (get_symbol_alignment): New function.
138 (legitimate_small_data_address_p): Likewise.
139 (legitimate_scaled_address): Update, call
140 legitimate_small_data_address_p.
141 (output_sdata): New static variable.
142 (arc_print_operand): Update how we handle small data operands.
143 (arc_print_operand_address): Likewise.
144 (arc_legitimate_address_p): Update, use
145 legitimate_small_data_address_p.
146 (arc_rewrite_small_data_p): Remove.
147 (arc_rewrite_small_data_1): Likewise.
148 (arc_rewrite_small_data): Likewise.
149 (small_data_pattern): Likewise.
150 (compact_sda_memory_operand): Update to use
151 legitimate_small_data_address_p and get_symbol_alignment.
152 (prepare_move_operands): Don't rewite sdata pattern.
153 (prepare_extend_operands): Remove.
154 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
156 (zero_extendqisi2): Likewise.
157 (zero_extendhisi2): Likewise.
158 (extendqihi2): Likewise.
159 (extendqisi2): Likewise.
160 (extendhisi2): Likewise.
164 * config/arc/constraints.md (Usd): Change it to memory constraint.
166 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
168 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
169 as source of std instructions.
170 * config/arc/arc.md (movsi_insn): Update pattern predicate to
171 allow 6-bit constants as source for store instructions.
172 (movdi_insn): Update instruction pattern to allow 6-bit constants
173 as source for store instructions.
175 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
177 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
179 2018-04-30 Nathan Sidwell <nathan@acm.org>
180 Sandra Loosemore <sandra@codesourcery.com>
182 * dumpfile.c (dump_open): Allow '-' for stdout.
183 * doc/invoke.texi (Developer Options): Document dump filename
184 determination early. Document stdin/stdout selection.
186 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
188 Microblaze Target: PIC data text relative
190 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
191 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
193 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
194 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
195 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
196 New addressing mode for data-text relative position indepenedent code.
197 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
198 'ADDRESS_SYMBOLIC_TXT_REL'.
199 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
200 (microblaze_legitimate_pic_operand): Exclude function calls from
201 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
202 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
204 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
205 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
206 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
207 for 'address + offset'.
208 (microblaze_expand_prologue): Add new function prologue call for
210 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
211 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
212 table in case of TARGET_PIC_DATA_TEXT_REL.
213 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
214 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
215 Add new macros 'UNSPEC_TEXT',
216 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
217 + exclude function calls from 'UNSPEC_PLT' in case of data text
219 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
220 new target hook for generating address diff vector tables in case of
222 * doc/tm.texi : Regenerate.
223 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
224 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
225 of addr diff vector generation.
226 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
227 target hook definition.
228 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
229 Add default function for generate_pic_addr_diff_vec -> flag_pic.
230 * doc/invoke.texi (Add new pic option): Add new microblaze pic
231 option for data text relative.
233 2018-04-30 Richard Biener <rguenther@suse.de>
235 * tree-chrec.h (evolution_function_is_constant_p): Remove
237 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
239 2018-04-30 Richard Biener <rguenther@suse.de>
242 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
244 2018-04-30 Richard Biener <rguenther@suse.de>
246 PR tree-optimization/28364
247 PR tree-optimization/85275
248 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
249 copying first exit test.
251 2018-04-28 Mark Wielaard <mark@klomp.org>
253 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
255 (dwarf_AT): Handle DW_AT_addr_base.
256 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
258 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
261 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
262 (*ashl<dwi>3_doubleword_mask_1): Ditto.
263 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
264 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
266 2018-04-28 Richard Biener <rguenther@suse.de>
268 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
269 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
270 to reflect use. Only add interesting stmts.
272 2018-04-27 Martin Jambor <mjambor@suse.cz>
275 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
276 the jump function allows for passing through aggregate values.
278 2018-04-27 David Malcolm <dmalcolm@redhat.com>
280 * input.h (in_system_header_at): Convert from macro to inline
282 (from_macro_expansion_at): Likewise.
283 (from_macro_definition_at): Likewise.
285 2018-04-27 Jeff Law <law@redhat.com>
287 * config.gcc: Mark tile* targets as deprecated/obsolete.
289 2018-04-27 Richard Biener <rguenther@suse.de>
291 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
294 2018-04-27 Richard Biener <rguenther@suse.de>
296 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
298 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
300 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
301 with Yd constraint. Set "preferred_for_speed" attribute from
302 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
304 (*movdi_internal): Ditto.
305 (movti_interunit splitters): Remove
306 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
307 (movdi_interunit splitters): Ditto.
308 * config/i386/constraints.md (Ye): Remove.
309 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
311 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
314 * config/aarch64/constraints.md (Usg): Limit to 31.
317 2018-04-27 Jakub Jelinek <jakub@redhat.com>
319 PR tree-optimization/85529
320 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
321 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
322 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
323 zero extension or masking of the MSB bit.
324 (optimize_range_tests): Add FIRST_BB argument, pass it through
325 to optimize_range_tests_var_bound.
326 (maybe_optimize_range_tests, reassociate_bb): Adjust
327 optimize_range_tests callers.
329 2018-04-26 Richard Biener <rguenther@suse.de>
330 Jakub Jelinek <jakub@redhat.com>
332 * cgraph.h (symbol_table): Just declare debug method here.
333 * symtab.c (symbol_table::debug): Define.
335 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
337 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
339 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
341 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
342 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
343 (*movdi_internal): Substitute Yi and Yj constraint with x
344 and Ym and Yn constraint with y constraint. Update "isa"
345 attribute and set "preferred_for_speed" attribute from
346 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
347 (*movsi_internal): Ditto.
348 (*movdf_internal): Ditto.
349 (*movsf_internal): Ditto.
350 (*zero_extendsidi2): Ditto.
351 * config/i386/sse.md (vec_set<mode>_0): Ditto.
352 (sse2_loadld): Ditto.
353 (*vec_extract<ssevecmodelower>_0): Ditto.
354 (*vec_extractv4si_0_zext_sse4): Ditto.
355 (vec_concatv2di): Ditto.
356 (*vec_dup<mode>): Ditto.
357 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
358 * config/i386/constraints.md (Yi): Remove.
363 2018-04-26 Nathan Sidwell <nathan@acm.org>
365 * dumpfile.c (dump_open): New.
366 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
367 (dump_finish): Detect stdio/stderr by value not name.
369 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
371 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
373 2018-04-26 Tom de Vries <tom@codesourcery.com>
376 * config/nvptx/nvptx.c (verify_neutering_jumps)
377 (verify_neutering_labels): New function
378 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
380 2018-04-26 Tom de Vries <tom@codesourcery.com>
383 * config/nvptx/nvptx.c (needs_neutering_p): New function.
384 (nvptx_single): Use needs_neutering_p to skip over insns that do not
387 2018-04-26 Richard Biener <rguenther@suse.de>
388 Tom de Vries <tom@codesourcery.com>
391 * lto-streamer-out.c (output_function): Fixup loops if required to match
392 discovery done in the reader.
394 2018-04-26 Richard Biener <rguenther@suse.de>
396 PR tree-optimization/85116
397 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
398 have a loop exit from the single latch predecessor. Remove
399 case of header with just condition.
400 (ch_base::copy_headers): Exclude infinite loops from any
402 (pass_ch::execute): Record exits.
404 2018-04-26 Richard Biener <rguenther@suse.de>
406 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
407 prologue cost vector and pass it to vect_get_load_cost.
408 (vect_get_peeling_costs_all_drs): Likewise.
409 (vect_peeling_hash_get_lowest_cost): Likewise.
410 (vect_enhance_data_refs_alignment): Likewise.
412 2018-04-26 Richard Biener <rguenther@suse.de>
415 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
416 checking of integer<->pointer conversions.
417 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
418 sign-/zero-extending pointer types.
419 (expand_omp_for_static_chunk): Likewise.
421 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
422 Jean Lee <xiaoyur347@gmail.com>
424 * config/mips/mips.c (mips_asan_shadow_offset): New function.
425 (TARGET_ASAN_SHADOW_OFFSET): Define.
426 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
427 true for -fsanitize=address.
429 2018-04-25 Mark Wielaard <mark@klomp.org>
431 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
434 2018-04-25 Jakub Jelinek <jakub@redhat.com>
436 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
437 than "alu", remove explicit "memory" and "imm_disp" attributes.
438 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
441 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
442 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
445 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
448 * config/i386/i386.c (ix86_expand_builtin): Change memory
449 operand to XI, extend p0 to Pmode.
450 * config/i386/i386.md: Change unspec volatile and operand
451 1 mode to XI, change operand 0 mode to P.
453 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
455 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
456 GET_MODE_MASK before any checking.
457 (nds32_can_use_bset_p): Likewise.
458 (nds32_can_use_btgl_p): Likewise.
460 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
462 * config/nds32/nds32-doubleword.md: New define_split pattern for
463 illegal register number.
465 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
467 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
469 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
471 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
473 2018-04-25 Richard Biener <rguenther@suse.de>
475 * lto-streamer.h (LTO_major_version): Bump to 8.
477 2018-04-25 Jakub Jelinek <jakub@redhat.com>
479 * BASE-VER: Set to 9.0.0.
481 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
483 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
484 in __abskf2 and __powikf2.
486 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
489 * config/aarch64/constraints.md (Usg, Usj): New constraints.
490 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
491 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
492 Use the above on operand 2. Reindent.
493 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
495 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
498 * common/config/i386/i386-common.c (ix86_handle_option): Don't
500 * config/i386/i386.opt (mcet): Removed.
501 * doc/install.texi: Remove -mcet documentation.
502 * doc/invoke.texi: Likewise.
504 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
507 * doc/install.texi: Remove -mcet from bootstrap-cet.
509 2018-04-24 Jakub Jelinek <jakub@redhat.com>
512 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
513 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
517 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
518 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
519 containing a CONST_VECTOR.
521 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
523 * doc/install.texi: Update newlib dependency for nvptx.
525 2018-04-24 Jakub Jelinek <jakub@redhat.com>
528 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
529 instead of INTVAL when shifting x left.
531 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
533 PR tree-optimization/85478
534 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
535 vect_grouped_store_supported for single element vectors.
537 2018-04-24 Richard Biener <rguenther@suse.de>
540 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
541 load cost increase to the case of non-constant step.
543 2018-04-24 Jakub Jelinek <jakub@redhat.com>
546 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
547 destination if any_malformed_asm.
549 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
552 * expr.c (store_field): In the bitfield case, if the value comes from
553 a function call and is returned in registers by means of a PARALLEL,
554 do not change the mode of the temporary unless BLKmode and VOIDmode.
556 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
558 PR rtl-optimization/85423
559 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
560 dependencies to debug insns when the previous insn is non-debug.
562 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
564 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
565 enums into a single definition.
566 (fls): Fix predicates and printing.
569 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
571 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
572 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
573 and short u6 immediate.
574 (check_if_valid_sleep_operand): Remove.
575 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
577 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
579 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
580 flag_always_save_lp condition.
581 * config/nds32/nds32.opt (malways-save-lp): New option.
583 2018-04-22 Shiva Chen <shiva0217@gmail.com>
585 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
586 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
587 * config/nds32/nds32.h
588 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
589 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
591 2018-04-22 Shiva Chen <shiva0217@gmail.com>
593 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
594 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
596 2018-04-22 Shiva Chen <shiva0217@gmail.com>
597 Chung-Ju Wu <jasonwucj@gmail.com>
599 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
601 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
602 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
604 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
606 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
608 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
610 * config/nds32/nds32-protos.h (nds32_data_alignment,
611 nds32_local_alignment): Declare.
612 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
613 nds32_local_alignment): New functions.
614 (TARGET_CONSTANT_ALIGNMENT): Define.
615 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
617 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
619 * config/nds32/nds32.c
620 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
621 (TARGET_MODES_TIEABLE_P): Likewise.
623 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
625 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
628 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
629 Chung-Ju Wu <jasonwucj@gmail.com>
631 * config/nds32/constants.md (unspec_volatile_element): Add enum values
632 for unaligned access.
633 * config/nds32/nds32-intrinsic.c: Implementation of expanding
635 * config/nds32/nds32-intrinsic.md: Likewise.
636 * config/nds32/nds32_intrinsic.h: Likewise.
637 * config/nds32/nds32.h (nds32_builtins): Likewise.
638 * config/nds32/nds32.opt (munaligned-access): New option.
639 * config/nds32/nds32.c (nds32_asm_file_start): Display
640 flag_unaligned_access status.
642 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
644 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
645 -mno-relax is present.
646 * config/riscv/linux.h (LINK_SPEC): Ditto.
648 2018-04-20 Martin Sebor <msebor@redhat.com>
651 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
653 (gimple_fold_builtin_stxcpy_chk): Same.
654 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
656 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
659 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
660 __powikf2 when long double is IEEE 128-bit.
662 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
664 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
665 step to make sure stack always aligned.
667 2018-04-20 Carl Love <cel@us.ibm.com>
670 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
673 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
674 Tom de Vries <tom@codesourcery.com>
677 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
678 Emit insns for calls too.
679 (nvptx_find_par): Always look for worker-level predecessor insn.
680 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
682 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
683 (nvptx_process_pars): Propagate frames for calls.
685 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
688 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
690 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
691 (ix86_handle_option): Don't handle OPT_mibt.
692 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
694 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
696 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
697 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
698 (ix86_target_macros): Define __CET__ with flag_cf_protection
700 * config/i386/i386.c (isa2_opts): Remove -mibt.
701 * config/i386/i386.h (TARGET_IBT): Removed.
702 (TARGET_IBT_P): Likewise.
703 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
704 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
705 * config/i386/i386.opt (mcet): Update help message.
708 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
709 -mcet as an alias for -mshstk.
711 2018-04-20 Richard Biener <rguenther@suse.de>
714 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
715 complexity by forcing a single use of the multiply operand.
717 2018-04-20 Martin Jambor <mjambor@suse.cz>
720 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
721 recursion dependency to only apply to non-clones.
723 2018-04-20 Martin Jambor <mjambor@suse.cz>
726 * ipa-cp.c (create_specialized_node): Check that clones of
727 self-recursive edges exist during IPA-CP.
729 2018-04-19 Toon Moene <toon@moene.org>
731 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
734 2018-04-19 Jakub Jelinek <jakub@redhat.com>
736 PR tree-optimization/85467
737 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
738 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
739 VECTOR_CST element to type.
741 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
744 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
745 * config/i386/i386.md (builtin_setjmp_setup): Removed.
746 (builtin_longjmp): Likewise.
747 (save_stack_nonlocal): New pattern.
748 (restore_stack_nonlocal): Likewise.
750 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
753 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
754 Replace ASM_OUTPUT_LABEL with fprintf.
756 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
759 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
760 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
761 * config/i386/i386-c.c (ix86_target_macros_internal): Also
762 define __IBT__ and __SHSTK__ for -fcf-protection.
763 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
765 (ix86_trampoline_init): Likewise.
766 (x86_output_mi_thunk): Likewise.
767 (ix86_notrack_prefixed_insn_p): Likewise.
768 (ix86_option_override_internal): Don't disallow -fcf-protection.
769 * config/i386/i386.md (rdssp<mode>): Also enable for
771 (incssp<mode>): Likewise.
772 (nop_endbr): Likewise.
773 * config/i386/i386.opt (mcet): Change help message to built-in
777 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
778 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
779 enable CET built-in functions.
781 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
783 * common/config/i386/i386-common.c
784 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
785 OPTION_MASK_ISA_MOVDIRI_UNSET,
786 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
787 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
788 * config.gcc (movdirintrin.h): New header.
789 * config/i386/cpuid.h (bit_MOVDIRI,
790 bit_MOVDIR64B): New bits.
791 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
793 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
794 (VOID, PVOID, PCVOID)): New function types.
795 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
796 __builtin_ia32_directstoreu_u64,
797 __builtin_ia32_movdir64b): New builtins.
798 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
799 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
801 (ix86_valid_target_attribute_inner_p): Ditto.
802 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
803 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
804 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
805 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
806 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
807 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
808 (movdiri<mode>, movdir64b_<mode>): New.
809 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
810 * config/i386/immintrin.h: Include movdirintrin.h.
811 * config/i386/movdirintrin.h: New file.
812 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
814 2018-04-19 Richard Biener <rguenther@suse.de>
817 * cfg.c (clear_bb_flags): When loop state says we have
818 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
820 2018-04-19 Richard Biener <rguenther@suse.de>
822 PR tree-optimization/84737
823 * tree-vect-data-refs.c (vect_copy_ref_info): New function
824 copying restrict info.
825 (vect_setup_realignment): Use it.
826 * tree-vectorizer.h (vect_copy_ref_info): Declare.
827 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
828 the first DR to all generated stores.
829 (vectorizable_load): Likewise for loads.
831 2018-04-19 Jakub Jelinek <jakub@redhat.com>
833 PR tree-optimization/85446
834 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
835 the integral and pointer types to have the same precision.
837 * doc/install.texi: Document --disable-cet being the default and
840 2018-04-18 Martin Liska <mliska@suse.cz>
842 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
845 2018-04-18 Martin Liska <mliska@suse.cz>
848 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
851 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
852 arguments if they are comparable.
854 2018-04-18 Martin Liska <mliska@suse.cz>
857 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
860 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
863 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
866 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
867 ENDBR after calling __morestack.
869 2018-04-18 David Malcolm <dmalcolm@redhat.com>
872 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
873 by using gcc_base_ver to generate a gcc_driver_version, and use
874 it when generating GCC_DRIVER_NAME.
875 * configure: Regenerate.
877 2018-04-18 Jakub Jelinek <jakub@redhat.com>
880 * config.gcc: Obsolete powerpc*-*-*spe*.
882 2018-04-17 Jakub Jelinek <jakub@redhat.com>
885 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
886 (stabstr_D): Change type of unum from unsigned int to
887 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
890 2018-04-17 Jim Wilson <jimw@sifive.com>
893 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
894 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
895 Set arg_pointer_offset after using pretend_args_size.
897 2018-04-17 Jakub Jelinek <jakub@redhat.com>
899 PR rtl-optimization/85431
900 * dse.c (record_store): Ignore zero width stores.
903 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
904 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
905 __builtin_stack_restore rather than after it.
906 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
907 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
908 argument instead of virtual_dynamic_stack_rtx.
910 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
912 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
914 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
915 Add note to error message to explain internal mapping of overloaded
916 built-in function name to non-overloaded built-in function name.
917 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
920 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
923 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
924 where the inputs overlap with the output.
926 2018-04-17 Jakub Jelinek <jakub@redhat.com>
928 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
929 (=v, v) alternative and explicit "memory" attribute.
930 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
931 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
933 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
934 "sselog1" type instead of "sselog".
935 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
936 "sselog". Remove explicit "memory" attribute.
937 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
938 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
940 (vec_extract_hi_v32hi): Merge all alternatives into one, use
941 "sselog1" type instead of "sselog". Remove explicit "memory"
943 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
944 use "sselog1" type instead of "sselog". Remove explicit "memory"
946 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
947 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
949 (vec_extract_hi_v64qi): Merge all alternatives into one, use
950 "sselog1" type instead of "sselog". Remove explicit "memory"
952 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
953 use "sselog1" type instead of "sselog". Remove explicit "memory"
957 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
960 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
963 2018-04-17 Martin Jambor <mjambor@suse.cz>
966 * ipa-cp.c (create_specialized_node): Call
967 expand_all_artificial_thunks if necessary.
969 2018-04-17 Martin Liska <mliska@suse.cz>
972 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
973 in message, remote space in between '_G' and '('.
975 2018-04-17 Jakub Jelinek <jakub@redhat.com>
978 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
979 avx512f_vmcmp<mode>3<round_saeonly_name>,
980 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
981 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
982 avx512f_rndscale<mode><round_saeonly_name>,
983 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
984 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
985 Use %<iptr>2 instead of %2 for -masm=intel.
986 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
987 avx512f_vcvttss2usi<round_saeonly_name>,
988 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
990 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
991 avx512f_vcvttsd2usi<round_saeonly_name>,
992 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
993 Use %q1 instead of %1 for -masm=intel.
994 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
995 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
996 of %3 for -masm=intel.
997 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
999 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1001 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1003 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1004 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1006 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1008 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1009 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1010 %g1 and one with %0 and %1.
1011 (avx512er_vmrcp28<mode><round_saeonly_name>,
1012 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1014 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1015 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1016 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1017 of %0 and %{%4%} for -masm=intel.
1018 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1019 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1020 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1021 order of %0 and %{%5%}%{z%} for -masm=intel.
1023 2018-04-17 Jan Hubicka <jh@suse.cz>
1026 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1028 2018-04-17 Martin Liska <mliska@suse.cz>
1031 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1032 for target_clone error message. Make default implementation
1033 clone to be a local declaration.
1034 (separate_attrs): Add new argument and check for an empty
1036 (expand_target_clones): Handle it.
1037 (ipa_target_clone): Make redirection just for target_clones
1040 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1041 Tom de Vries <tom@codesourcery.com>
1044 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1045 tiled basic blocks without omp continue statements.
1047 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1050 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1051 vec_extract expression as having side effects to make sure it gets
1054 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1057 * config/i386/i386.c (get_builtin_code_for_version): Check
1060 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1063 * gcc/config.gcc: Support "skylake".
1064 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1066 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1067 (processor_target_table): Add "skylake".
1068 (ix86_option_override_internal): Add "skylake".
1069 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1070 PROCESSOR_CANNONLAKE.
1071 (get_builtin_code_for_version): Fix priority for
1072 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1073 PROCESSOR_SKYLAKE-AVX512.
1074 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1075 (processor_type): Add PROCESSOR_SKYLAKE.
1077 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1078 Jason Merrill <jason@redhat.com>
1081 * convert.c (convert_to_integer_1): Use direct recursion for
1082 enumeral types and types with a precision less than the number
1083 of bits in their mode.
1085 2018-04-16 Julia Koval <julia.koval@intel.com>
1088 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1089 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1091 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1094 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1095 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1096 and -mno-direct-move.
1098 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1101 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1102 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1103 Ensure negative shifts result in {0}.
1105 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1107 PR rtl-optimization/79916
1108 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1109 regs (if any) to define how to gnerate SD moves when LRA is in
1112 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1114 PR rtl-optimization/85393
1115 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1116 * except.c (expand_dw2_landing_pad_for_region): Make static.
1117 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1118 a label and unconditional jump to old_bb, rather than
1119 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1122 PR rtl-optimization/85376
1123 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1124 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1125 instead of a specific value.
1127 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1128 Bin Cheng <bin.cheng@arm.com>
1130 PR tree-optimization/82965
1131 PR tree-optimization/83991
1132 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1133 by_profile_only parameter.
1134 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1135 information if the loop was predicted to iterate too many times.
1136 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1138 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1141 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1144 2018-04-13 Martin Liska <mliska@suse.cz>
1145 Jakub Jelinek <jakub@redhat.com>
1148 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1149 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1150 * builtins.c (expand_builtin_memory_copy_args): Use
1151 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1152 handle dest_addr == pc_rtx.
1154 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1157 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1158 asked to not generate direct moves.
1159 (fix_trunc<mode>si2_stfiwx): Similar.
1160 (fix_trunc<mode>si2_internal): Similar.
1162 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1165 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1166 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1167 lookup if dest in some wider mode is known to be const0_rtx and
1168 if so, record permanent equivalence for it to be ZERO_EXTEND of
1169 the narrower mode destination.
1171 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1173 * lto-streamer-out.c (output_function): Revert 259346.
1174 * omp-expand.c (expand_oacc_for): Likewise.
1176 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1178 PR rtl-optimization/85354
1179 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1180 * sel-sched.c (sel_global_init): ... here.
1182 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1185 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1186 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1187 mode for PE-COFF targets.
1188 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1189 (i386_pe_asm_lto_end): Likewise.
1190 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1191 (TARGET_ASM_LTO_END): Likewise.
1192 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1193 (i386_pe_asm_lto_start): New function.
1194 (i386_pe_asm_lto_end): Likewise.
1196 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1197 Richard Biener <rguenther@suse.de>
1200 * lto-streamer-out.c (output_function): Fix CFG loop state before
1202 * omp-expand.c (expand_oacc_for): Handle calls to internal
1203 functions like regular functions.
1205 2018-04-12 Richard Biener <rguenther@suse.de>
1208 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1209 for the early LTO debug to properly generate references to it
1210 during DIE emission. Do not re-use that for the skeleton for
1212 (dwarf2out_early_finish): Likewise.
1214 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1217 * config/i386/sse.md
1218 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1219 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1220 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1221 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1222 and output is a reg, avoid creating invalid lowpart subreg, but
1223 instead split into a 512-bit move. Don't split if not AVX512VL,
1224 input is xmm16+ reg and output is a mem.
1225 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1226 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1227 xmm16+ reg and output is a mem.
1229 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1231 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1232 also for flag_dwarf2_cfi_asm.
1234 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1236 PR rtl-optimization/85342
1237 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1238 a bool scalar var inside of the loop instead. Don't try to update
1239 recog_data.operand after failed apply_change_group.
1241 2018-04-12 Tom de Vries <tom@codesourcery.com>
1244 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1245 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1246 array with flexible array member as array without given dimension.
1247 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1248 argument for undefined param to true.
1250 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1253 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1254 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1255 from PowerPC section.
1256 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1257 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1258 help text that is too long.
1259 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1260 help text that is too long.
1261 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1262 help text that is too long.
1264 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1266 * config/alpha/alpha.md (stack_probe_internal): Rename
1267 from "probe_stack". Update all callers.
1269 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1271 PR rtl-optimization/84566
1272 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1273 sched_macro_fuse_insns.
1275 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1278 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1279 (compute_block_dependences): ... from here.
1281 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1283 PR tree-optimization/85331
1284 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1285 from int to HOST_WIDE_INT.
1287 2018-04-11 Martin Jambor <mjambor@suse.cz>
1290 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1291 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1292 not the same as the source val.
1293 (cgraph_edge_brings_value_p): New parameter.
1294 (gather_edges_for_value): Pass destination value to
1295 cgraph_edge_brings_value_p.
1296 (perhaps_add_new_callers): Likewise.
1297 (get_info_about_necessary_edges): Likewise and exclude values brought
1298 only by self-recursive edges.
1299 (create_specialized_node): Redirect only clones of self-calling edges.
1300 (+self_recursive_pass_through_p): New function.
1301 (find_more_scalar_values_for_callers_subset): Use it.
1302 (find_aggregate_values_for_callers_subset): Likewise.
1303 (known_aggs_to_agg_replacement_list): Removed.
1304 (decide_whether_version_node): Re-calculate known constants for all
1305 remaining context clones.
1307 2018-04-11 Richard Biener <rguenther@suse.de>
1310 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1311 from early DWARF output.
1312 (dwarf2out_early_finish): Output line info unconditionally into
1313 early DWARF and add reference to it.
1315 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1318 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1319 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1320 other than V2DFmode using iptr mode attribute.
1321 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1323 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1325 PR rtl-optimization/84659
1326 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1328 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1331 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1333 (output_loc_list): Pass address of a dummy size variable even in the
1334 locview handling loop.
1335 (index_location_lists): Add comment on why skip_loc_list_entry can't
1338 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1341 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1344 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1347 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1348 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1349 and -mstring-compare-inline-limit.
1351 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1354 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1355 for stack clash protection in a register whenever we need it to be in
1358 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1360 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1361 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1363 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1366 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1368 (mlong-double-): Ditto.
1369 * config/rs6000/sysv4.opt (msdata=): Ditto.
1370 (mtls-size=): Ditto.
1372 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1374 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1375 erroneous entries for
1376 "vector int vec_ldl (int, long int *)", and
1377 "vector unsigned int vec_ldl (int, unsigned long int *)".
1378 Add comments and entries for
1379 "vector bool char vec_ldl (int, bool char *)",
1380 "vector bool short vec_ldl (int, bool short *)",
1381 "vector bool int vec_ldl (int, bool int *)",
1382 "vector bool long long vec_ldl (int, bool long long *)",
1383 "vector pixel vec_ldl (int, pixel *)",
1384 "vector long long vec_ldl (int, long long *)",
1385 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1386 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1387 type tree bool_long_long_type_node and correct definition of
1388 bool_V2DI_type_node to make reference to this new type tree.
1389 (rs6000_mangle_type): Replace erroneous reference to
1390 bool_long_type_node with bool_long_long_type_node.
1391 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1392 comments to emphasize sign distinctions for char and int types and
1393 replace RS6000_BTI_bool_long constant with
1394 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1395 use of RS6000_BTI_pixel.
1396 (bool_long_type_node): Remove this macro definition.
1397 (bool_long_long_type_node): New macro definition
1399 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1401 PR rtl-optimization/85300
1402 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1403 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1404 simplify_unary_operation fails.
1406 2018-04-10 Martin Liska <mliska@suse.cz>
1408 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1409 cgraph_edge and ipa_ref.
1411 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1415 * config/i386/sse.md
1416 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1417 computation of the VEC_MERGE selector from mask.
1418 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1419 Fix decoding of the VEC_MERGE selector into mask.
1421 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1423 PR tree-optimization/85286
1424 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1426 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1428 * final.c (final_1): Set insn_last_address as well as
1429 insn_current_address.
1431 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1434 * explow.c (emit_stack_probe): Call validize_mem on memory location
1435 before passing it to gen_probe_stack. Create address operand and
1436 legitimize it for the probe_stack_address case.
1438 2018-04-09 Jan Hubicka <jh@suse.cz>
1441 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1442 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1443 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1445 (free_lang_data): Rebuild type inheritance graph.
1447 2018-04-09 Martin Sebor <msebor@redhat.com>
1449 * invoke.texi (-finline-small-functions): Mention other optimization
1451 (-findirect-inlining, -fpartial-inlining): Same.
1452 (-finline-functions-called-once): Same.
1453 (-freorder-blocks-and-partition): Same.
1455 2018-04-09 Jan Hubicka <jh@suse.cz>
1458 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1459 jumps; choose last target that matches the criteria (i.e.
1460 no partition changes for non-crossing jumps).
1461 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1462 support for redirecting crossing jumps to non-crossing.
1464 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1466 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1467 also for naked functions.
1469 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1471 * config/arc/arc.md (add_shift): New pattern.
1472 (add_shift2): Likewise.
1473 (sub_shift): Likewise.
1474 (sub_shift_cmp0_noout): Likewise.
1475 (compare_si_ashiftsi): Likewise.
1476 (xbfu_cmp0_noout): New combine pattern.
1477 (xbfu_cmp0"): Likewise.
1478 (movsi_set_cc_insn): Place the predicable variant first.
1479 (commutative_binary_cmp0_noout): Remove clobber.
1480 (commutative_binary_cmp0): New pattern.
1481 (noncommutative_binary_cmp0): Likewise.
1482 (noncommutative_binary_cmp0_noout): Likewise.
1483 (noncommutative_binary_comparison_result_used): Removed.
1484 (rsub_cmp0): New pattern.
1485 (rsub_cmp0_noout): Likewise.
1486 (extzvsi): Changed, keep only meaningful variants.
1487 (SQH, SEZ): New iterators.
1488 (SQH_postfix): New mode attribute.
1489 (SEZ_prefix): New code attribute.
1490 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1491 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1492 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1494 (noncommutative_operator): Check the availability of barrel
1497 2018-04-09 Richard Biener <rguenther@suse.de>
1499 PR tree-optimization/85284
1500 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1501 Only use the niter constraining form of simple_iv when the exit
1504 2018-04-09 Tom de Vries <tom@codesourcery.com>
1507 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1508 (define_expand "*memory_barrier"): New define_expand.
1509 (define_insn "memory_barrier"): New insn.
1511 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1513 PR rtl-optimization/80463
1514 PR rtl-optimization/83972
1515 PR rtl-optimization/83480
1517 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1518 correct producer for the insn.
1519 (tidy_control_flow): Fixup seqnos in case of debug insns.
1521 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1523 PR rtl-optimization/83913
1525 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1526 different sched-times when merging exprs.
1528 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1530 PR rtl-optimization/83962
1532 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1533 tidy_fallthru_edge and tidy_control_flow.
1535 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1537 PR rtl-optimization/83530
1539 * sel-sched.c (force_next_insn): New global variable.
1540 (remove_insn_for_debug): When force_next_insn is true, also leave only
1541 next insn in the ready list.
1542 (sel_sched_region): When the region wasn't scheduled, make another pass
1543 over it with force_next_insn set to 1.
1545 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1547 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1549 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1550 for interrupt control.
1551 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1552 functions for interrupt control.
1553 * config/nds32/nds32-intrinsic.md: Likewise.
1554 * config/nds32/nds32_intrinsic.h: Likewise.
1555 * config/nds32/nds32.h (nds32_builtins): Likewise.
1557 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1559 * config/nds32/nds32.c (nds32_init_machine_status,
1560 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1561 strict_aligned_p field.
1562 (nds32_expand_to_rtl_hook): New function.
1563 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1564 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1566 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1567 Chung-Ju Wu <jasonwucj@gmail.com>
1569 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1570 * config/nds32/nds32-n7.md: New file.
1571 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1572 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1574 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1575 * config/nds32/nds32.md (pipeline_model): Add n7.
1576 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1577 * config/nds32/pipelines.md: Include n7 settings.
1579 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1580 Chung-Ju Wu <jasonwucj@gmail.com>
1582 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1583 * config/nds32/nds32-e8.md: New file.
1584 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1585 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1587 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1588 * config/nds32/nds32.md (pipeline_model): Add e8.
1589 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1590 * config/nds32/pipelines.md: Include e8 settings.
1592 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1593 Chung-Ju Wu <jasonwucj@gmail.com>
1595 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1596 * config/nds32/nds32-n8.md: New file.
1597 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1598 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1600 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1601 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1602 * config/nds32/nds32.md (pipeline_model): Add n8.
1603 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1604 * config/nds32/pipelines.md: Include n8 settings.
1606 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1607 Chung-Ju Wu <jasonwucj@gmail.com>
1609 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1610 * config/nds32/nds32-n9-2r1w.md: New file.
1611 * config/nds32/nds32-n9-3r2w.md: New file.
1612 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1613 nds32_register_ports): New or modify for cpu n9.
1614 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1616 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1617 * config/nds32/nds32-utils.c: New file.
1618 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1619 TARGET_MUL_SLOW): Define.
1620 * config/nds32/nds32.md (pipeline_model): New attribute.
1621 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1622 New options that support cpu n9.
1623 * config/nds32/pipelines.md: Include n9 settings.
1624 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1626 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1628 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1629 information if necessary.
1630 (output_cond_branch_compare_zero): Likewise.
1631 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1632 (nds32_target_alignment): Refine for alignment.
1633 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1634 (FUNCTION_BOUNDARY): Modify.
1635 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1637 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1639 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1641 * config/nds32/constants.md (unspec_volatile_element): Add values for
1642 TLB operation and data prefetch.
1643 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1644 functions for TLB operation and data prefetch.
1645 * config/nds32/nds32-intrinsic.md: Likewise.
1646 * config/nds32/nds32_intrinsic.h: Likewise.
1647 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1648 (nds32_print_operand): Likewise.
1649 * config/nds32/nds32.h (nds32_builtins): Likewise.
1651 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1652 Andrew Pinski <pinsika@gcc.gnu.org>
1655 * match.pd: Use constant_boolean_node of correct type instead of
1656 boolean_true_node or boolean_false_node for simplifying
1657 pointer comparisons to zero.
1659 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1661 PR tree-optimization/80021
1662 * tree.c (verify_type_variant): Make error call in verify_variant_match
1663 translatable and remove final full stop.
1665 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1667 * config/nds32/constants.md (unspec_volatile_element): Add
1668 UNSPEC_VOLATILE_EH_RETURN.
1669 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1670 nds32_output_stack_pop): Support dwarf exception handling process.
1671 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1672 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1673 exception handling process.
1674 (nds32_compute_stack_frame): Likewise.
1675 (nds32_return_addr_rtx): Likewise.
1676 (nds32_initial_elimination_offset): Likewise.
1677 (nds32_expand_prologue): Likewise.
1678 (nds32_expand_epilogue): Likewise.
1679 (nds32_dynamic_chain_address): New function.
1680 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1682 (DYNAMIC_CHAIN_ADDRESS): Define.
1683 (EH_RETURN_DATA_REGNO): Define.
1684 (EH_RETURN_STACKADJ_RTX): Define.
1685 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1686 patterns for dwarf exception handling.
1688 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1690 * config/nds32/nds32.h: Clean up obsolete macros.
1692 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1694 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1695 Add enum values for particular instructions.
1696 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1697 particular intrinsic functions.
1698 * config/nds32/nds32-intrinsic.md: Likewise.
1699 * config/nds32/nds32_intrinsic.h: Likewise.
1700 * config/nds32/nds32.h (nds32_builtins): Likewise.
1701 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1702 (btst, ave): New patterns for particular instructions.
1704 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1706 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1707 Add enum values for atomic load/store and memory sync.
1708 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1710 * config/nds32/nds32-intrinsic.md: Likewise.
1711 * config/nds32/nds32_intrinsic.h: Likewise.
1712 * config/nds32/nds32.h (nds32_builtins): Likewise.
1714 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1716 PR tree-optimization/85257
1717 * fold-const.c (native_encode_vector): If not all elts could fit
1718 and off is -1, return 0 rather than offset.
1719 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1720 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1721 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1722 adjust buffer in native_interpret_expr call.
1724 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1726 * config/nds32/constants.md (unspec_volatile_element): Add cache
1727 control enum values.
1728 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1729 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1730 * config/nds32/nds32.c (nds32_cctl_names): New.
1731 (nds32_print_operand): Handle cache control register names.
1732 * config/nds32/nds32.h (nds32_builtins): New enum values.
1733 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1735 * config/nds32/nds32.md (type): Add mmu.
1736 * config/nds32/pipelines.md (simple_insn): Add mmu.
1738 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1740 * config/nds32/nds32.md (type): Remove call.
1741 * config/nds32/pipelines.md (simple_insn): Likewise.
1743 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1745 * config/nds32/constants.md (unspec_volatile_element): Add
1746 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1747 UNSPEC_VOLATILE_FMFCFG.
1748 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1749 description for fmfcfg and fmfcsr.
1750 (bdesc_1arg): Add fmtcsr.
1751 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1752 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1753 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1754 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1755 unspec_fmfcfg): New patterns.
1756 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1757 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1758 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1759 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1760 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1761 __nds32__fmfcfg): Define.
1763 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1765 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1766 intrinsic register names.
1767 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1768 intrinsic register enum values and macros.
1770 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1772 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1773 for load/store addressing form.
1774 (nds32_print_operand_address): Likewise.
1776 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1779 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1780 based on LABEL_REF. Remove useless assertion.
1781 (pic_address_needs_scratch): Fix formatting.
1782 (sparc_legitimize_pic_address): Minor tweaks.
1783 (sparc_delegitimize_address): Adjust assertion accordingly.
1784 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1785 into symbolic_operand.
1786 (movsi_high_pic_label_ref): Likewise.
1787 (movsi_lo_sum_pic_label_ref): Likewise.
1788 (movdi_pic_label_ref): Likewise.
1789 (movdi_high_pic_label_ref): Likewise.
1790 (movdi_lo_sum_pic_label_ref): Likewise.
1792 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1794 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1795 custom LIB_SPEC setup.
1797 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1798 Kito Cheng <kito.cheng@gmail.com>
1800 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1801 * config/riscv/freebsd.h: New.
1803 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1805 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1806 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1809 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1810 Kito Cheng <kito.cheng@gmail.com>
1812 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1813 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1814 * config/nds32/nds32-protos.h (nds32_output_call,
1815 nds32_output_return): Declare.
1816 * config/nds32/nds32.md: Refine all the call and return patterns.
1818 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1821 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1822 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1824 PR rtl-optimization/84872
1825 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1826 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1829 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1831 * expr.c (copy_blkmode_to_reg): Revert 254862.
1832 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1834 2018-04-06 Richard Biener <rguenther@suse.de>
1837 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1838 after seeing a component reference with an adjacent field. Treat
1839 refs to arrays at struct end of external decls similar to
1840 refs to unconstrained commons.
1842 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1845 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1846 look through SAVE_EXPRs with non-side-effects argument. Adjust
1848 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1851 2018-04-06 Richard Biener <rguenther@suse.de>
1854 * alias.c (find_base_term): New wrapper around find_base_term
1855 unwinding CSELIB_VAL_PTR changes.
1856 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1859 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1861 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1863 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1864 constant definitions.
1865 ("nop"): lr 0,0 -> nopr r0
1866 ("nop_lr0", "nop_lr1"): New insn definitions.
1868 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1870 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1871 NDS32_V3PUSH_AVAILABLE_P macro.
1873 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1874 Chung-Ju Wu <jasonwucj@gmail.com>
1876 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1877 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1878 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1879 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1880 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1881 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1882 * config/nds32/constraints.md: New constraints and checking for hard
1883 float configuration.
1884 * config/nds32/iterators.md: New mode iterator and attribute for hard
1885 float configuration.
1886 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1888 * config/nds32/nds32-fpu.md: New file.
1889 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1890 deal with hard float code generation.
1891 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1893 (abi_type, float_reg_number): New enum type.
1894 * config/nds32/nds32-predicates.c: New predicates for hard float.
1895 * config/nds32/nds32-protos.h: Declare functions for hard float.
1896 * config/nds32/nds32.c: Implementation for hard float configuration.
1897 * config/nds32/nds32.h: Definitions for hard float configuration.
1898 * config/nds32/nds32.md: Include hard float machine description and
1899 modify patterns for hard float configuration.
1900 * config/nds32/nds32.opt: New options for hard float configuration.
1901 * config/nds32/predicates.md: New predicates for hard float
1904 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1906 * common/config/nds32/nds32-common.c
1907 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1909 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1912 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1913 CONSTRUCTOR_ELT (ctor, ...)->value.
1915 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1918 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1920 2018-04-05 Tom de Vries <tom@codesourcery.com>
1923 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1926 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1927 Kito Cheng <kito.cheng@gmail.com>
1929 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1930 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1931 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1932 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1934 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1935 Kito Cheng <kito.cheng@gmail.com>
1937 * config/nds32/constraints.md (Ufe): New memory constraint.
1938 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1939 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1940 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1942 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1943 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1945 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1947 * config/nds32/nds32.md: Use optimize_size in the condition for
1948 alu-shift instructions.
1950 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1952 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1954 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1956 * config/nds32/nds32.md (negsi2): Refine pattern.
1958 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1959 Chung-Ju Wu <jasonwucj@gmail.com>
1961 * config/nds32/iterators.md (shift_rotate): New code iterator.
1962 (shift): New code attribute.
1963 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1964 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1965 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1966 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1967 bit-wise operations.
1968 (andsi3, *andsi3): Ditto.
1969 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1970 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1971 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1972 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1973 nds32_ior_operand, nds32_xor_operand): New predicates.
1975 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1977 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1978 (addsi3, subsi3): ... this.
1980 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1982 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1984 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1986 * config/nds32/nds32.md: Adjust indention.
1988 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1990 * config/nds32/nds32.md (feature): New attribute.
1992 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1994 * config/nds32/nds32.md (subtype): New attribute.
1996 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1999 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2000 expansion to perform a bitwise AND of the argument followed by a
2001 boolean negation of the result.
2003 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2005 PR rtl-optimization/84878
2006 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2007 the basic block. Assert the use reference is not artificial and that
2008 it has an associated insn.
2010 2018-04-04 Michael Matz <matz@suse.de>
2012 * builtins.c (compute_objsize): Pass correct operand
2013 to array_at_struct_end_p.
2015 2018-04-04 Richard Biener <rguenther@suse.de>
2018 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2019 from contexts for DINFO_LEVEL_TERSE and below.
2021 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2023 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2024 resiter_operand condition.
2025 * config/nds32/nds32.md (*move<mode>): Ditto.
2027 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2028 Monk Chiang <sh.chiang04@gmail.com>
2030 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2032 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2034 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2036 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2037 Kito Cheng <kito.cheng@gmail.com>
2039 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2040 nds32_cond_code_str, output_cond_branch,
2041 output_cond_branch_compare_zero, nds32_expand_cbranch,
2042 nds32_expand_cstore, nds32_expand_movcc,
2043 nds32_output_cbranchsi4_equality_zero,
2044 nds32_output_cbranchsi4_equality_reg,
2045 nds32_output_cbranchsi4_equality_reg_or_const_int,
2046 nds32_output_cbranchsi4_greater_less_zero: New functions.
2047 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2048 nds32_expand_cstore, nds32_expand_movcc,
2049 nds32_output_cbranchsi4_equality_zero,
2050 nds32_output_cbranchsi4_equality_reg,
2051 nds32_output_cbranchsi4_equality_reg_or_const_int,
2052 nds32_output_cbranchsi4_greater_less_zero): Declare.
2053 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2054 nds32_rimm11s_operand): New predicates.
2055 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2056 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2059 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2061 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2062 * config/nds32/nds32.md: Ditto.
2063 * config/nds32/pipelines.md: Ditto.
2065 2018-04-04 Richard Biener <rguenther@suse.de>
2067 PR tree-optimization/85168
2068 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2069 propagating abnormals.
2071 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2073 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2075 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2076 Kito Cheng <kito.cheng@gmail.com>
2078 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2079 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2080 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2081 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2082 * config/nds32/nds32.md (sibcall_internal): New.
2083 (sibcall_register): Remove.
2084 (sibcall_immediate): Remove.
2085 (sibcall_value_internal): New.
2086 (sibcall_value_register): Remove.
2087 (sibcall_value_immediate): Remove.
2088 * config/nds32/predicates.md (nds32_general_register_operand): New.
2089 (nds32_call_address_operand): New.
2091 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2093 PR rtl-optimization/85167
2094 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2095 bb_defs if *split_p, instead preinitialize it to NULL.
2097 PR tree-optimization/85156
2098 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2099 evaluating the argument multiple times.
2101 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2103 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2105 (_mm_cvtpd_ps): Likewise.
2106 (_mm_cvttpd_epi32): Likewise.
2107 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2108 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2109 vector, pixel, and bool following altivec.h include.
2111 2018-04-03 Martin Sebor <msebor@redhat.com>
2113 * doc/extend.texi (Common Function Attributes): Clarify.
2114 (const attribute): Likewise.
2115 (pure attribute): Likewise.
2117 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2120 * config/i386/i386.c (ix86_expand_vector_set): Use
2121 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2123 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2125 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2126 instructions when changing rounding bits to preserve precision bits
2127 in the x87 control word.
2129 2018-04-03 Martin Liska <mliska@suse.cz>
2131 PR tree-optimization/82491
2132 * rtl.h (strip_offset_and_add): Replace += suboffset with
2133 poly_uint64 () + suboffset.
2135 2018-03-29 Martin Liska <mliska@suse.cz>
2136 Martin Jambor <mjambor@suse.cz>
2139 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2140 param_type is not an integral or pointer type.
2142 2018-04-03 Richard Biener <rguenther@suse.de>
2144 * sese.h (recompute_all_dominators): Remove.
2146 2018-04-02 Martin Sebor <msebor@redhat.com>
2148 * doc/invoke.texi (-Wrestrict): Fix typos.
2150 2018-04-02 Jim Wilson <jimw@sifive.com>
2152 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2153 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2154 (<optab>di3, <optab>si3_extend): Likewise.
2155 (<optab>si3_mask, <optab>si3_mask_1): New.
2156 (<optab>di3_mask, <optab>di3_mask_1): New.
2157 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2158 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2159 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2161 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2163 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2166 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2168 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2169 (nds32_canonicalize_comparison): New function.
2171 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2172 Kito Cheng <kito.cheng@gmail.com>
2173 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2175 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2176 * config/nds32/constants.md (unspec_volatile_element): Add
2177 UNSPEC_VOLATILE_RELAX_GROUP.
2178 * config/nds32/nds32-relax-opt.c: New file.
2179 * config/nds32/nds32-predicates.c
2180 (nds32_symbol_load_store_p): New function.
2181 * config/nds32/nds32-protos.h
2182 (nds32_symbol_load_store_p): Declare function.
2183 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2184 * config/nds32/nds32.c
2185 (nds32_register_pass): New function to register pass.
2186 (nds32_register_passes): New function to register passes.
2187 * config/nds32/nds32.md (relax_group): New pattern.
2188 * config/nds32/nds32.opt (mrelax-hint): New option.
2189 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2191 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2193 * config/nds32/t-nds32: Modify files dependency.
2195 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2197 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2198 (PROFILE_HOOK): Define its implementation.
2200 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2202 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2203 type and 32-bit size.
2205 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2208 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2209 (V_128_256): New mode iterator.
2210 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2211 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2212 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2214 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2215 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2217 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2220 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2221 NaN inputs correctly.
2223 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2226 * config/rs6000/vsx.md (??r): New mode attribute.
2227 (*vsx_mov<mode>_64bit): Use it.
2228 (*vsx_mov<mode>_32bit): Likewise.
2230 2018-03-30 Martin Sebor <msebor@redhat.com>
2232 PR tree-optimization/84818
2233 * builtins.c (check_access): Use warning_n.
2235 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2238 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2240 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2243 2018-03-30 Julia Koval <julia.koval@intel.com>
2246 * x86-tune.def (movx, partial_reg_dependency): Enable for
2249 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2252 * lra-constraints.c (process_alt_operands): Move setting
2253 this_alternative_matches below.
2255 2018-03-29 Martin Liska <mliska@suse.cz>
2258 * doc/invoke.texi: Document how LTO works with debug info.
2259 Describe auto-load support of binutils. Mention 'x86-64'
2260 as valid option value of -march option.
2262 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2264 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2267 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2268 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2269 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2272 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2275 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2276 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2277 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2278 for RS6000_BTM_POWERPC64.
2279 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2280 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2281 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2286 2018-03-28 Carl Love <cel@us.ibm.com>
2289 2017-09-27 Carl Love <cel@us.ibm.com>
2291 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2292 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2293 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2296 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2298 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2299 instead of __vector bool.
2300 (_mm_max_pu8): Likewise.
2301 (_mm_min_pi16): Likewise.
2303 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2306 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2307 (DIVWEUO): Likewise.
2309 (DIVDEUO): Likewise.
2310 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2311 DIVWEUO and DIVDEUO.
2312 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2313 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2314 (div_extend): Likewise.
2315 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2317 (__builtin_divweuo): Likewise.
2318 (__builtin_divdeo): Likewise.
2319 (__builtin_divdeuo): Likewise.
2321 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2324 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2325 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2327 PR tree-optimization/82004
2328 * gimple-match-head.c (optimize_pow_to_exp): New function.
2329 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2330 Don't fold to exp if optimize_pow_to_exp is false.
2332 2018-03-28 Martin Liska <mliska@suse.cz>
2335 * calls.c (initialize_argument_information): Fix trailing space.
2336 * common.opt: Fix typo and provide better explanation for
2337 -fsanitize-coverage option.
2338 * config/i386/i386.opt: Fix typo.
2340 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2341 Martin Liska <mliska@suse.cz>
2344 * gimplify.c (asan_poison_variable): Don't do the check for
2345 gimplify_omp_ctxp here.
2346 (gimplify_decl_expr): Do it here.
2347 (gimplify_target_expr): Likewise.
2349 2018-03-28 Martin Liska <mliska@suse.cz>
2352 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2353 chkp_type_bounds_count if MPX is not enabled.
2355 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2357 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2359 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2362 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2363 function to create the function decl for complex long double
2364 multiply and divide for -mabi=ieeelongdouble.
2365 (init_float128_ieee): Call it.
2367 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2370 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2371 -fcf-protection=branch -mibt.
2372 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2374 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2377 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2379 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2382 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2383 extern array declarations.
2385 2018-03-27 Richard Biener <rguenther@suse.de>
2388 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2389 explicit single_use checks.
2391 2018-03-27 Richard Biener <rguenther@suse.de>
2393 PR tree-optimization/85082
2394 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2397 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2399 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2400 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2401 Turn on fasynchronous-unwind-tables and funwind-tables.
2403 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2406 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2407 (*bmi_blsr_<mode>_ccz): Ditto.
2409 2018-03-26 Tom de Vries <tom@codesourcery.com>
2411 PR tree-optimization/85063
2412 * omp-general.c (offloading_function_p): New function. Factor out
2414 * omp-offload.c (pass_omp_target_link::gate): ... here.
2415 * omp-general.h (offloading_function_p): Declare.
2416 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2417 with attribute omp declare target for offloading functions.
2419 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2421 PR tree-optimization/84005
2422 * tree-data-ref.h (get_base_for_alignment): Declare.
2423 * tree-data-ref.c (get_base_for_alignment_1): New function.
2424 (get_base_for_alignment): Likewise.
2425 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2426 get_base_for_alignment to find a suitable base object, instead
2427 of always using drb->base_address.
2429 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2432 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2433 known size by default.
2435 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2438 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2439 and non BLKmode operands.
2441 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2444 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2445 Clean up attributes.
2447 2018-03-23 Richard Biener <rguenther@suse.de>
2450 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2451 we are going to emit early debug for LTO.
2453 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2456 * function.c (match_asm_constraints_1): Don't optimize if input
2457 doesn't satisfy general_operand predicate for output's mode.
2460 * alias.c (write_dependence_p): Don't require for x_canonicalized
2461 non-VOIDmode if x has VOIDmode.
2464 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2465 just don't try to optimize it rather than assert it never happens.
2467 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2469 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2470 macro expansions for definition of ST_INTERNAL_<mode> and
2471 LD_INTERNAL_<mode> builtins.
2472 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2474 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2476 (altivec_expand_st_builtin): Likewise.
2477 (altivec_expand_builtin): Remove calls to deleted functions.
2478 (rs6000_address_for_altivec): Delete this function.
2479 * config/rs6000/vector.md: Remove expands for
2480 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2482 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2485 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2486 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2487 re-computing once computed.
2488 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2489 (arm_init_machine_status): Initialize
2490 machine->static_chain_stack_bytes.
2492 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2495 * doc/extend.texi: Add four new prototypes for vec_ld.
2496 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2497 definitions for more logical presentation.
2498 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2499 entries for V1TI variants of __builtin_altivec_ld builtin.
2500 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2501 handling of V1TI variant of LVX icode pattern.
2502 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2503 (rs6000_gimple_fold_builtin): Likewise.
2504 (altivec_init_builtins): Add code to define
2505 __builtin_altivec_lvx_v1ti function.
2507 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2510 * function.c (match_asm_constraints_1): Don't do the optimization
2511 if input isn't a REG, SUBREG, MEM or constant.
2513 2018-03-22 Tom de Vries <tom@codesourcery.com>
2515 PR tree-optimization/84956
2516 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2517 bb_has_abnormal_pred.
2519 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2522 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2523 DECL_INITIAL (decl) to decl at the end.
2524 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2527 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2529 * doc/extend.texi (__builtin_tgmath): Document when complex
2530 integer types are treated as _Complex _Float64.
2532 2018-03-21 Tom de Vries <tom@codesourcery.com>
2534 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2536 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2538 PR tree-optimization/84960
2539 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2540 if it is ENTRY block, move them into single succ of ENTRY in that case.
2542 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2544 PR tree-optimization/84811
2545 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2546 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2547 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2548 (known_subrange_p): Update accordingly. Cast each value involved
2549 in the size comparison, rather than casting the result of the
2552 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2554 PR tree-optimization/84982
2555 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2556 by flipping the least significant bit rather than all bits from
2557 bitpos to bitpos + bitsize - 1.
2559 2018-03-21 Nathan Sidwell <nathan@acm.org>
2561 * doc/extend.texi (Deprecated Features): Remove mention of
2562 long-deleted deprecations.
2564 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2567 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2568 * configure: Regenerate.
2570 2018-03-21 Tom de Vries <tom@codesourcery.com>
2572 PR tree-optimization/83126
2573 * tree-parloops.c (num_phis): New function.
2574 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2576 2018-03-21 Nathan Sidwell <nathan@acm.org>
2578 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2579 mention anon-struct/union members and trailing attributes.
2581 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2583 PR tree-optimization/84969
2584 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2585 builtin memset partitions if they set different rhs values.
2587 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2589 PR rtl-optimization/84989
2590 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2591 VEC_DUPLICATE with scalar result mode.
2593 2018-03-21 Martin Liska <mliska@suse.cz>
2596 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2597 not intended return statement.
2599 2018-03-21 Martin Liska <mliska@suse.cz>
2602 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2603 (chkp_find_bound_slots_1): Limit number of iterations.
2605 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2608 * Minor grammar fixes for x86 options.
2610 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2613 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2614 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2616 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2619 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2620 (altivec_lvx_<mode>_1op): Likewise.
2621 (altivec_stvx_<mode>_2op): Likewise.
2622 (altivec_stvx_<mode>_1op): Likewise.
2623 (altivec_lvx_<VM2:mode>): New define_expand.
2624 (altivec_stvx_<VM2:mode>): Likewise.
2625 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2626 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2627 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2628 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2629 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2630 (rs6000_gen_lvx): Likewise.
2631 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2632 (altivec_expand_stv_builtin): Likewise.
2633 (altivec_expand_builtin): Likewise.
2634 * config/rs6000/vector.md: Likewise.
2636 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2639 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2642 2018-03-20 Richard Biener <rguenther@suse.de>
2645 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2646 sign-conversions as zero, fall back to standard scalar_stmt
2649 2018-03-20 Martin Liska <mliska@suse.cz>
2652 * predict.c (rebuild_frequencies): Handle case when we have
2653 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2655 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2658 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2659 flag_section_anchors.
2660 * varasm.c (use_blocks_for_decl_p): Remove hack for
2661 dw2_force_const_mem.
2664 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2666 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2667 be created, use lowpart_subreg of operands[0] rather than operands[0]
2669 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2670 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2671 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2672 and n constraint instead of aarch64_shift_imm_di and Usd.
2673 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2674 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2676 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2679 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2680 to favor GPR over NEON registers.
2681 (<shift>di3_neon): Likewise.
2683 2018-03-20 Tom de Vries <tom@codesourcery.com>
2686 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2687 (nvptx_process_pars): Emit bar.sync asap and alap.
2689 2018-03-20 Tom de Vries <tom@codesourcery.com>
2692 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2693 seen_label if seen_label is already set.
2695 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2698 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2699 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2700 Use 1U instead of 1. Formatting fixes.
2703 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2704 instead of TREE_TYPE (s1) for the return value.
2706 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2708 PR tree-optimization/84946
2709 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2710 bitsize + bitsize in poly_uint64 rather than poly_int64.
2713 * dwarf2asm.c: Include fold-const.c.
2714 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2715 of decl rather than decl itself.
2717 PR rtl-optimization/84643
2718 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2720 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2723 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2724 calling assemble_variable.
2726 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2729 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2730 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2732 2018-03-19 Jim Wilson <jimw@sifive.com>
2735 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2736 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2737 (riscv_first_stack_step): Likewise.
2738 (riscv_option_override): Use STACK_BOUNDARY instead of
2740 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2742 (BIGGEST_ALIGNMENT): Set to 128.
2743 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2744 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2747 2018-03-19 Richard Biener <rguenther@suse.de>
2749 PR tree-optimization/84933
2750 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2751 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2753 2018-03-19 Richard Biener <rguenther@suse.de>
2755 PR tree-optimization/84859
2756 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2757 (cond_if_else_store_replacement): Perform sinking operation on
2758 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2759 Generalize what a BB with a single eligible store is.
2761 2018-03-19 Richard Biener <rguenther@suse.de>
2763 PR tree-optimization/84929
2764 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2765 chrec_is_positive against non-chrec arg.
2767 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2770 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2772 2018-03-18 Martin Liska <mliska@suse.cz>
2774 PR rtl-optimization/84635
2775 * regrename.c (build_def_use): Use matches_mode only when
2778 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2780 PR tree-optimization/84913
2781 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2782 vectorize chains of COND_EXPRs.
2784 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2786 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2788 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2790 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2792 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2794 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2796 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2797 Kito Cheng <kito.cheng@gmail.com>
2799 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2800 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2801 (nds32_adjust_reg_alloc_order): New function.
2802 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2804 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2806 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2807 nds32_print_operand, nds32_print_operand_address): Use
2808 HOST_WIDE_INT_PRINT_DEC instead.
2810 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2812 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2814 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2817 * config/i386/i386.c (initial_ix86_tune_features,
2818 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2820 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2821 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2822 rather than 1u << ix86_tune. Formatting fix.
2823 (ix86_option_override_internal): Change ix86_arch_mask from
2824 unsigned int to unsigned HOST_WIDE_INT, initialize to
2825 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2826 (ix86_function_specific_restore): Likewise.
2828 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2831 * postreload.c (reload_combine_recognize_pattern): Perform
2832 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2833 truncate_int_for_mode the result for the destination's mode.
2836 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2837 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2839 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2842 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2845 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2846 regno_allocno_class_array and sorted_pseudos.
2847 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2848 insns where regno is used.
2850 2018-03-16 Martin Liska <mliska@suse.cz>
2853 * multiple_target.c (create_dispatcher_calls): Redirect
2854 reference in the symbol table.
2856 2018-03-16 Martin Liska <mliska@suse.cz>
2859 * multiple_target.c (create_dispatcher_calls): Redirect also
2862 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2866 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2867 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2868 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2870 2018-03-16 Julia Koval <julia.koval@intel.com>
2872 * doc/invoke.texi (Skylake Server): Add CLWB.
2873 Cannonlake): Remove CLWB.
2875 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2877 PR tree-optimization/84841
2878 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2880 (FLOAT_ONE_CONST_TYPE): Define.
2881 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2882 (sort_by_operand_rank): Put entries with higher constant_type last
2883 rather than first to match comments.
2885 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2887 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2890 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2893 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2894 check and use address of target always.
2896 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2899 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2900 (indirect_thunk_bnd_needed): Likewise.
2901 (indirect_thunks_used): Likewise.
2902 (indirect_thunks_bnd_used): Likewise.
2903 (indirect_return_needed): New.
2904 (indirect_return_bnd_needed): Likewise.
2905 (output_indirect_thunk_function): Add a bool argument for
2907 (output_indirect_thunk_function): Don't generate alias for
2908 function return thunk.
2909 (ix86_code_end): Call output_indirect_thunk_function to generate
2910 function return thunks.
2911 (ix86_output_function_return): Set indirect_return_bnd_needed
2912 and indirect_return_needed instead of indirect_thunk_bnd_needed
2913 and indirect_thunk_needed.
2915 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2917 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2918 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2919 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2921 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2922 Paul Hua <paul.hua.gm@gmail.com>
2925 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2927 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2929 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2930 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2933 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2936 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2937 instead of GET_MODE_SIZE when comparing Units.
2939 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2942 * varasm.c (hash_section): Return an unchangeble hash value
2943 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2944 Return !aarch64_can_use_per_function_literal_pools_p ().
2946 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2949 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2950 rather than address of cmode as last argument to prepare_cmp_insn.
2952 2018-03-15 Julia Koval <julia.koval@intel.com>
2954 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2955 F_AVX512VNNI, F_AVX512BITALG): New.
2957 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2960 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2961 insn for floating-point loads and stores.
2963 2018-03-14 Carl Love <cel@us.ibm.com>
2965 * config/rs6000/rs6000-c.c: Add macro definitions for
2966 ALTIVEC_BUILTIN_VEC_PERMXOR.
2967 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2968 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2969 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2970 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2972 * config/doc/extend.texi: Add prototypes for vec_permxor.
2974 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2977 * diagnostic-show-locus.c (class layout_point): Convert m_line
2978 from int to linenum_type.
2979 (line_span::comparator): Use linenum "compare" function when
2980 comparing line numbers.
2981 (test_line_span): New function.
2982 (layout_range::contains_point): Convert param "row" from int to
2984 (layout_range::intersects_line_p): Likewise.
2985 (layout::will_show_line_p): Likewise.
2986 (layout::print_source_line): Likewise.
2987 (layout::should_print_annotation_line_p): Likewise.
2988 (layout::print_annotation_line): Likewise.
2989 (layout::print_leading_fixits): Likewise.
2990 (layout::annotation_line_showed_range_p): Likewise.
2991 (struct line_corrections): Likewise for field m_row.
2992 (line_corrections::line_corrections): Likewise for param "row".
2993 (layout::print_trailing_fixits): Likewise.
2994 (layout::get_state_at_point): Likewise.
2995 (layout::get_x_bound_for_row): Likewise.
2996 (layout::print_line): Likewise.
2997 (diagnostic_show_locus): Likewise for locals "last_line" and
2999 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3000 * input.c (selftest::test_linenum_comparisons): New function.
3001 (selftest::input_c_tests): Call it.
3002 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3003 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3004 * selftest.h (ASSERT_GT): New macro.
3005 (ASSERT_GT_AT): New macro.
3006 (ASSERT_LT): New macro.
3007 (ASSERT_LT_AT): New macro.
3009 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3011 PR rtl-optimization/84780
3012 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3014 2018-03-14 Martin Liska <mliska@suse.cz>
3016 * tree.c (record_node_allocation_statistics): Use
3017 get_stats_node_kind.
3018 (get_stats_node_kind): New function extracted from
3019 record_node_allocation_statistics.
3020 (free_node): Use get_stats_node_kind.
3022 2018-03-14 Richard Biener <rguenther@suse.de>
3024 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3025 that the value-set of ANTIC_IN doesn't grow.
3028 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3030 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3031 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3033 2018-03-14 Julia Koval <julia.koval@intel.com>
3035 * config.gcc (icelake-client, icelake-server): New.
3037 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3038 (initial_ix86_arch_features): Ditto.
3039 (PTA_SKYLAKE): Add SGX.
3040 (PTA_ICELAKE): Remove.
3041 (PTA_ICELAKE_CLIENT): New.
3042 (PTA_ICELAKE_SERVER): New.
3043 (ix86_option_override_internal): Split up icelake on icelake client and
3045 (get_builtin_code_for_version): Ditto.
3046 (fold_builtin_cpu): Ditto.
3047 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3048 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3049 * config/i386/i386.h (processor_type): Ditto.
3050 * doc/invoke.texi: Ditto.
3052 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3055 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3056 INTEGER_CST offset, add it together with bitpos / 8 and
3057 sign extend based on POINTER_SIZE.
3061 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3064 * config/i386/constraints.md (Yc): New register constraint.
3065 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3066 Use Yc constraint for alternative 2 of operand 0. Remove
3067 preferred_for_speed attribute.
3069 2018-03-14 Richard Biener <rguenther@suse.de>
3071 PR tree-optimization/84830
3072 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3073 with the old one to avoid oscillations.
3075 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3078 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3080 (assign_by_spills): Return a flag of reload assignment failure.
3081 Do not process the reload assignment failures. Do not spill other
3082 reload pseudos if they has the same reg class. Update n if
3084 (lra_assign): Add a return arg. Set up from the result of
3085 assign_by_spills call.
3086 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3087 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3088 usage_insns if it is not NULL.
3089 (spill_hard_reg_in_range): New function.
3090 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3091 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3092 function prototypes.
3093 (lra_assign): Change prototype.
3094 * lra.c (lra): Add code to deal with fails by splitting hard reg
3097 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3099 * config/riscv/riscv.opt (mrelax): New option.
3100 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3101 "norelax" when riscv_mrelax is disabled.
3102 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3104 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3107 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3108 reassociation for int modes.
3110 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3112 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3113 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3115 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3116 * config/aarch64/aarch64-sve.md
3117 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3118 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3119 (*extend<mode><Vwide>2): Rename to...
3120 (aarch64_sve_extend<mode><Vwide>2): ...this.
3121 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3122 renaming the old pattern to...
3123 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3125 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3126 define_expand, renaming the old pattern to...
3127 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3128 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3129 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3130 account when deciding which SVE instruction the optab should use.
3131 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3133 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3135 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3136 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3137 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3138 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3139 (tlsdesc_small_<mode>): Turn a define_expand and use
3140 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3141 (tlsdesc_small_advsimd_<mode>): ...this.
3142 (tlsdesc_small_sve_<mode>): New pattern.
3144 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3146 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3147 (UNSPEC_UMUL_HIGHPART): New constants.
3148 (MUL_HIGHPART): New int iteraor.
3149 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3150 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3152 (*<su>mul<mode>3_highpart): New define_insn.
3154 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3157 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3160 2018-03-13 Martin Liska <mliska@suse.cz>
3163 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3165 (sem_item_optimizer::~sem_item_optimizer): Release it.
3166 (sem_item_optimizer::merge_classes): Register variable aliases.
3167 (sem_item_optimizer::fixup_pt_set): New function.
3168 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3169 * ipa-icf.h: Declare new variables and functions.
3171 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3174 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3175 integer_pow2p@2 and test integer_pow2p in condition.
3176 (A < 0 ? C : 0): Similarly for @1.
3179 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3180 characters starting at p contain '\0' character, don't look beyond
3184 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3185 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3188 * reg-stack.c (change_stack): Change update_end var from int to
3189 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3190 also call set_block_for_insn on the newly added insns and rescan.
3193 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3194 on the last operand.
3197 * tree.c (stabilize_reference_1): Return save_expr (e) for
3198 STATEMENT_LIST even if it doesn't have side-effects.
3200 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3202 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3204 2018-03-12 Renlin Li <renlin.li@arm.com>
3206 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3207 aarch64_output_scalar_simd_mov_immediate.
3209 2018-03-12 Martin Sebor <msebor@redhat.com>
3211 PR tree-optimization/83456
3212 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3213 for perfectly overlapping calls to memcpy.
3214 (gimple_fold_builtin_memory_chk): Same.
3215 (gimple_fold_builtin_strcpy): Handle no-warning.
3216 (gimple_fold_builtin_stxcpy_chk): Same.
3217 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3219 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3221 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3222 parameter. Use it for SFmode.
3223 (rs6000_function_arg_advance_1): Adjust.
3224 (rs6000_function_arg): Adjust.
3225 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3227 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3229 PR rtl-optimization/84169
3230 PR rtl-optimization/84780
3231 * combine.c (can_combine_p): Check for a 2-insn combination whether
3232 the destination register is used between the two insns, too.
3234 2018-03-12 Richard Biener <rguenther@suse.de>
3236 PR tree-optimization/84803
3237 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3238 for refs DR analysis didn't process.
3240 2018-03-12 Richard Biener <rguenther@suse.de>
3242 PR tree-optimization/84777
3243 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3244 force-vectorize loops ignore whether we are optimizing for size.
3246 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3248 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3249 (TARGET_MD_ASM_ADJUST): Define.
3251 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3252 Kito Cheng <kito.cheng@gmail.com>
3253 Chung-Ju Wu <jasonwucj@gmail.com>
3255 * config/nds32/nds32.c (nds32_compute_stack_frame,
3256 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3257 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3258 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3259 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3260 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3261 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3262 * config/nds32/nds32.md (prologue, epilogue): Use macro
3263 NDS32_V3PUSH_AVAILABLE_P to do checking.
3265 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3268 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3269 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3270 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3271 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3272 addition of most attributes on !orig_type_die or the attribute not
3273 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3275 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3276 Chung-Ju Wu <jasonwucj@gmail.com>
3278 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3280 * config/nds32/nds32.opt (mvh): New option.
3282 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3283 Chung-Ju Wu <jasonwucj@gmail.com>
3285 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3287 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3288 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3291 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3292 Chung-Ju Wu <jasonwucj@gmail.com>
3294 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3296 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3297 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3299 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3300 Kito Cheng <kito.cheng@gmail.com>
3301 Chung-Ju Wu <jasonwucj@gmail.com>
3303 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3304 UNSPEC_FFMISM and UNSPEC_FLMISM.
3305 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3306 for ffb, ffmism and flmism.
3307 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3308 (unspec_ffmism): Ditto.
3309 (unspec_flmism): Ditto.
3310 (nds32_expand_builtin_impl): Check if string extension is available.
3311 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3312 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3314 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3317 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3320 * lra-assigns.c (assign_by_spills): Return a flag of reload
3321 assignment failure. Do not process the reload assignment
3322 failures. Do not spill other reload pseudos if they has the same
3324 (lra_assign): Add a return arg. Set up from the result of
3325 assign_by_spills call.
3326 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3327 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3328 usage_insns if it is not NULL.
3329 (spill_hard_reg_in_range): New function.
3330 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3331 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3332 function prototypes.
3333 (lra_assign): Change prototype.
3334 * lra.c (lra): Add code to deal with fails by splitting hard reg
3337 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3340 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3342 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3345 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3346 (dw_val_node): Add val_symbolic_view.
3347 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3348 (symview_upper_bound): New.
3349 (new_line_info_table): Initialize symviews_since_reset.
3350 (dwarf2out_source_line): Count symviews_since_reset and set
3351 symview_upper_bound.
3352 (dw_val_equal_p): Handle symview.
3353 (add_AT_symview): New.
3354 (print_dw_val): Handle symview.
3355 (attr_checksum, attr_checksum_ordered): Likewise.
3356 (same_dw_val_p, size_of_die): Likewise.
3357 (value_format, output_die): Likewise.
3358 (add_high_low_attributes): Use add_AT_symview for entry_view.
3359 (dwarf2out_finish): Reset symview_upper_bound, clear
3362 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3365 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3366 Add strict argument and use it.
3367 (rs6000_split_multireg_move): Update for new strict argument.
3368 (mem_operand_gpr): Disallow all non-offsettable addresses.
3369 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3371 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3374 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3375 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3376 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3379 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3380 decl, use remap_type if we want to use the type.
3382 2018-03-09 Martin Sebor <msebor@redhat.com>
3384 PR tree-optimization/84526
3385 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3387 (builtin_access::generic_overlap): Be prepared to handle non-array
3390 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3392 PR rtl-optimization/84682
3393 * lra-constraints.c (process_address_1): Check is_address flag
3394 for address constraints.
3395 (process_alt_operands): Likewise.
3396 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3397 preprocess_constraints.
3398 * recog.h (preprocess_constraints): Add oploc parameter.
3400 * recog.c (preprocess_constraints): Test address_operand for
3401 CT_ADDRESS constraints.
3403 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3406 * lra-assigns.c (assign_by_spills): Return a flag of reload
3407 assignment failure. Do not process the reload assignment
3408 failures. Do not spill other reload pseudos if they has the same
3410 (lra_assign): Add a return arg. Set up from the result of
3411 assign_by_spills call.
3412 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3413 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3414 usage_insns if it is not NULL.
3415 (spill_hard_reg_in_range): New function.
3416 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3417 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3418 function prototypes.
3419 (lra_assign): Change prototype.
3420 * lra.c (lra): Add code to deal with fails by splitting hard reg
3423 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3426 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3427 Accept complain bool parameter. Only emit errors if it is true.
3428 (arm_parse_cpu_option_name): Likewise.
3429 (arm_target_thumb_only): Adjust callers of the above.
3430 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3431 prototype to take a default true bool parameter.
3432 (arm_parse_arch_option_name): Likewise.
3434 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3435 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3439 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3440 * configure: Regenerate.
3441 * configure.ac ("linker --version-script option"): New.
3442 ("linker soname option"): New.
3444 2018-03-09 Richard Biener <rguenther@suse.de>
3446 PR tree-optimization/84775
3447 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3448 immediate uses of predicate stmts and mark them modified.
3451 PR tree-optimization/84178
3452 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3454 (version_loop_for_if_conversion): Delay update_ssa call.
3455 (tree_if_conversion): Delay update_ssa until after predicate
3458 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3461 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3462 when the function accesses prior frames.
3464 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3467 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3468 gen_llsym, otherwise call maybe_gen_llsym.
3471 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3472 has ',' character inside of it.
3474 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3477 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3478 as clobbering CC_REGNUM.
3480 2018-03-08 Richard Biener <rguenther@suse.de>
3483 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3484 (follow_copies_to_constant): Do not follow SSA names registered
3487 2018-03-08 Richard Biener <rguenther@suse.de>
3489 PR tree-optimization/84178
3490 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3492 (version_loop_for_if_conversion): Delay update_ssa call.
3493 (tree_if_conversion): Delay update_ssa until after predicate
3496 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3498 PR tree-optimization/84178
3499 * tree-if-conv.c (release_bb_predicate): Remove the
3500 the assertion that the stmts have NULL use_ops.
3501 Discard the statements, asserting that they haven't
3502 yet been added to a BB.
3504 2018-03-08 Richard Biener <rguenther@suse.de>
3506 PR tree-optimization/84746
3507 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3508 (phi_translate): Pass in destination ANTIC_OUT set.
3509 (phi_translate_1): Likewise. For a simplified result lookup
3510 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3511 (phi_translate_set): Adjust.
3512 (do_pre_regular_insertion): Likewise.
3513 (do_pre_partial_partial_insertion): Likewise.
3515 2018-03-08 Martin Liska <mliska@suse.cz>
3517 PR gcov-profile/84735
3518 * doc/gcov.texi: Document usage of profile files.
3519 * gcov-io.h: Document changes in the format.
3521 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3525 * dwarf2out.c (struct dw_line_info_table): Update comments for
3527 (FORCE_RESET_NEXT_VIEW): New.
3528 (FORCE_RESETTING_VIEW_P): New.
3529 (RESETTING_VIEW_P): Check for -1 too.
3530 (ZERO_VIEW_P): Likewise.
3531 (new_line_info_table): Force-reset next view.
3532 (dwarf2out_begin_function): Likewise.
3533 (dwarf2out_source_line): Simplify zero_view_p initialization.
3534 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3535 view directly. Omit view when omitting .loc at line 0.
3537 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3539 PR tree-optimization/84740
3540 * tree-switch-conversion.c (process_switch): Call build_constructors
3541 only if info.phi_count is non-zero.
3543 PR tree-optimization/84739
3544 * tree-tailcall.c (find_tail_calls): Check call arguments against
3545 DECL_ARGUMENTS (current_function_decl) rather than
3546 DECL_ARGUMENTS (func) when checking for tail recursion.
3548 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3550 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3551 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3552 Volker Reichelt's entry and add entries for people that perform
3553 GCC fuzzy testing and report numerous bugs.
3555 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3558 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3559 readonly data in sdata, if that is disabled.
3560 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3561 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3562 -mreadonly-in-sdata option.
3564 2018-03-07 Martin Sebor <msebor@redhat.com>
3566 PR tree-optimization/84468
3567 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3568 basic block when looking for nul assignment.
3570 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3573 * except.h (output_function_exception_table): Adjust prototype.
3574 * except.c (output_function_exception_table): Remove FNNAME parameter
3575 and add SECTION parameter. Ouput one part of the table at a time.
3576 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3577 the first part of the exception table and emit unwind directives.
3578 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3579 (i386_pe_seh_cold_init): Likewise.
3580 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3581 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3582 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3583 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3584 * config/i386/winnt.c: Include except.h.
3585 (struct seh_frame_state): Add reg_offset, after_prologue and
3586 in_cold_section fields.
3587 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3588 (i386_pe_seh_cold_init): New function.
3589 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3590 to seh->in_cold_section.
3591 (seh_emit_push): Record the offset of the push.
3592 (seh_emit_save): Record the offet of the save.
3593 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3594 Test seh->after_prologue to disregard the epilogue.
3595 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3596 (i386_pe_end_cold_function): New function.
3598 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3601 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3602 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3605 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3607 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3608 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3611 * multiple_target.c: Include tree-inline.h and intl.h.
3612 (expand_target_clones): Diagnose and fail if node->definition and
3613 !tree_versionable_function_p (node->decl).
3615 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3617 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3619 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3620 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3621 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3623 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3626 * combine.c (try_combine): Use reg_or_subregno instead of handling
3627 just paradoxical SUBREGs and REGs.
3629 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3631 * config/arc/arc.c (arc_finalize_pic): Remove function.
3632 (arc_must_save_register): We use single base PIC register, remove
3633 checks to save/restore the PIC register.
3634 (arc_expand_prologue): Likewise.
3635 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3637 (arc_verify_short): Likewise.
3638 (arc_attr_type): Likewise.
3639 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3640 (walk_stores): Likewise.
3641 (arc_address_cost): Make it static.
3642 (arc_verify_short): Likewise.
3643 (branch_dest): Likewise.
3644 (arc_attr_type): Likewise.
3645 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3646 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3647 (arc_final_prescan_insn): Remove inserting the nops due to
3648 hardware hazards. It is done in reorg step.
3649 (insn_length_variant_t): Remove.
3650 (insn_length_parameters_t): Likewise.
3651 (arc_insn_length_parameters): Likewise.
3652 (arc_get_insn_variants): Likewise.
3653 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3655 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3658 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3661 PR tree-optimization/84687
3662 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3664 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3666 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3668 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3669 Rename to ppc_speculation_barrier.
3670 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3671 __builtin_ppc_speculation_barrier.
3673 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3676 * combine.c (combine_simplify_rtx): Don't try to simplify if
3677 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3680 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3682 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3683 to 32 bytes when compiling for POWER9.
3685 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3688 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3689 regparm >= 3 with no arg reg available also for calls with
3690 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3693 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3695 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3697 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3700 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3702 2018-03-05 Richard Biener <rguenther@suse.de>
3704 PR tree-optimization/84486
3705 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3706 When inserting a __builtin_assume_aligned call set the LHS
3707 SSA name alignment info accordingly.
3709 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3711 PR tree-optimization/84114
3712 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3713 Avoid reassociation of FLOAT_MODE addition.
3715 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3717 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3718 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3719 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3720 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3721 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3722 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3723 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3725 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3726 __builtin_ia32_wbinvd): New builtins.
3727 (SPECIAL_ARGS2): New.
3728 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3729 (SPECIAL_ARGS2): New.
3730 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3731 (ix86_valid_target_attribute_inner_p): Ditto.
3732 (ix86_init_mmx_sse_builtins): Add special_args2.
3733 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3734 TARGET_WBNOINVD_P): New.
3735 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3736 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3737 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3738 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3739 * config/i386/pconfigintrin.h: New file.
3740 * config/i386/wbnoinvdintrin.h: Ditto.
3741 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3742 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3744 2018-03-05 Richard Biener <rguenther@suse.de>
3746 PR tree-optimization/84670
3747 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3749 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3750 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3751 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3752 doesn't grow if all successors (recursively) were visited at least
3755 2018-03-05 Richard Biener <rguenther@suse.de>
3757 PR tree-optimization/84650
3758 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3759 if executed in the loop pipeline.
3761 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3763 * doc/configfiles.texi (Configuration Files): Move info about
3764 conditionalizing $target-protos.h to...
3765 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3766 differs from $target-protos.h.
3768 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3769 Chung-Ju Wu <jasonwucj@gmail.com>
3771 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3772 * config/nds32/nds32-multiple.md (setmemsi): Define.
3773 * config/nds32/nds32-memory-manipulation.c
3774 (nds32_gen_dup_4_byte_to_word_value): New.
3775 (emit_setmem_word_loop): New.
3776 (emit_setmem_byte_loop): New.
3777 (nds32_expand_setmem_loop): New.
3778 (nds32_expand_setmem_loop_v3m): New.
3779 (nds32_expand_setmem_unroll): New.
3780 (nds32_expand_setmem): New.
3782 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3783 Chung-Ju Wu <jasonwucj@gmail.com>
3785 * config/nds32/nds32-memory-manipulation.c
3786 (nds32_emit_load_store): New.
3787 (nds32_emit_post_inc_load_store): New.
3788 (nds32_emit_mem_move): New.
3789 (nds32_emit_mem_move_block): New.
3790 (nds32_expand_movmemsi_loop_unknown_size): New.
3791 (nds32_expand_movmemsi_loop_known_size): New.
3792 (nds32_expand_movmemsi_loop): New.
3793 (nds32_expand_movmemsi_unroll): New.
3794 (nds32_expand_movmemqi): Rename ...
3795 (nds32_expand_movmemsi): ... to this.
3796 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3797 (movmemsi): ... to this.
3798 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3799 (nds32_expand_movmemsi): ... to this.
3801 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3802 Monk Chiang <sh.chiang04@gmail.com>
3803 Chung-Ju Wu <jasonwucj@gmail.com>
3805 * config/nds32/nds32-protos.h
3806 (nds32_expand_load_multiple): New arguments.
3807 (nds32_expand_store_multiple): Ditto.
3808 (nds32_valid_multiple_load_store): Rename ...
3809 (nds32_valid_multiple_load_store_p): ... to this.
3810 * config/nds32/nds32-memory-manipulation.c
3811 (nds32_expand_load_multiple): Refine implementation.
3812 (nds32_expand_store_multiple): Ditto.
3813 * config/nds32/nds32-multiple.md
3814 (load_multiple): Update nds32_expand_load_multiple interface.
3815 (store_multiple): Update nds32_expand_store_multiple interface.
3816 * config/nds32/nds32-predicates.c
3817 (nds32_valid_multiple_load_store): Rename ...
3818 (nds32_valid_multiple_load_store_p): ... to this and refine
3820 * config/nds32/predicates.md
3821 (nds32_load_multiple_and_update_address_operation): New predicate.
3822 (nds32_store_multiple_and_update_address_operation): New predicate.
3824 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3825 Chung-Ju Wu <jasonwucj@gmail.com>
3827 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3828 (combo): New attribute.
3829 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3831 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3833 * config/nds32/nds32.opt: Change -mcmodel= default value.
3835 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3836 Monk Chiang <sh.chiang04@gmail.com>
3837 Chung-Ju Wu <jasonwucj@gmail.com>
3839 * config/nds32/constants.md (unspec_element): New enum.
3840 * config/nds32/constraints.md (Umw): New constraint.
3841 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3842 * config/nds32/nds32-intrinsic.md: Likewise.
3843 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3844 (nds32_valid_smw_lwm_base_p): New.
3845 (nds32_output_smw_single_word): New.
3846 (nds32_output_lmw_single_word): New.
3847 (nds32_expand_unaligned_load): New.
3848 (nds32_expand_unaligned_store): New.
3849 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3850 (nds32_output_smw_single_word): Declare.
3851 (nds32_output_lmw_single_word): Declare.
3852 (nds32_expand_unaligned_load): Declare.
3853 (nds32_expand_unaligned_store): Declare.
3854 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3855 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3856 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3857 NDS32_BUILTIN_UASTORE_DW.
3858 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3861 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3862 Kito Cheng <kito.cheng@gmail.com>
3863 Chung-Ju Wu <jasonwucj@gmail.com>
3865 * config/nds32/nds32-intrinsic.c
3866 (nds32_expand_builtin_null_ftype_reg): Delete.
3867 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3868 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3869 (nds32_read_argument): New.
3870 (nds32_legitimize_target): Ditto.
3871 (nds32_legitimize_argument): Ditto.
3872 (nds32_check_constant_argument): Ditto.
3873 (nds32_expand_unop_builtin): Ditto.
3874 (nds32_expand_unopimm_builtin): Ditto.
3875 (nds32_expand_binop_builtin): Ditto.
3876 (nds32_builtin_decl_impl): Ditto.
3877 (builtin_description): Ditto.
3878 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3879 (nds32_init_builtins_impl): Ditto.
3880 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3881 (nds32_builtin_decl): New.
3882 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3883 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3885 2018-03-02 Jeff Law <law@redhat.com>
3887 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3888 (redundant_insn, fill_simple_delay_slots): Likewise.
3889 (fill_slots_from_thread): Likewise.
3890 * resource.c (mark_referenced_resources): Likewise.
3891 (mark_set_resources, find_dead_or_set_registers): Likewise.
3893 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3895 * substring-locations.h (format_warning_va): Formatting fix for
3897 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3899 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3900 * substring-locations.c: Include intl.h.
3901 (format_warning_va): Turned into small wrapper around
3902 format_warning_n_va, renamed to ...
3903 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3904 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3906 (format_warning_at_substring_n): New function.
3907 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3908 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3909 format_warning_at_substring with just a shorter name instead of
3910 const function pointer.
3911 (fmtwarn_n): New function.
3912 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3913 appropriate, get rid of all the fmtstr temporaries, move conditionals
3914 with G_() wrapped string literals directly into fmtwarn arguments,
3915 cast dir.len to (int), formatting fixes.
3917 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3919 * doc/invoke.texi: Remove "Cilk Plus" references.
3921 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3922 Richard Biener <rguenther@suse.de>
3925 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3926 for error or warning attributes if CALL_FROM_THUNK_P is set.
3929 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3932 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3933 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3936 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3937 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3939 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3940 instead of -1U in last predictors element's probability member.
3942 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3945 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3946 arguments if they are comparable.
3948 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3950 PR tree-optimization/84634
3951 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3952 masks and masked_loop_p with a single loop_masks, making sure it's
3953 null for bb vectorization.
3955 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3957 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3958 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3959 loop->force_vectorize to check whether there is no alias.
3961 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3964 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3966 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3968 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3969 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3970 instead of a loop around prev_real_insn.
3971 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3975 * config/i386/i386.c (ix86_print_operand): Use conditional
3976 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3979 2018-03-02 Richard Biener <rguenther@suse.de>
3981 PR tree-optimization/84427
3982 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3983 (bitmap_set_subtract_values): Rewrite to handle multiple
3986 (prune_clobbered_mems): Likewise.
3987 (phi_translate): Take edge instead of pred/phiblock.
3988 (phi_translate_1): Likewise.
3989 (phi_translate_set): Likewise. Insert all translated
3990 exprs for a value into the set, keeping possibly multiple
3991 expressions per value.
3992 (compute_antic_aux): Adjust for phi_translate changes.
3993 When intersecting union the expressions and prune those
3994 not in the final value set, keeping possibly multiple
3995 expressions per value. Do not use value-insertion
3996 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3997 all expressions. Add verification that the value-sets
3998 only shrink during iteration.
3999 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4000 (do_pre_regular_insertion): Likewise.
4001 (do_pre_partial_partial_insertion): Likewise.
4003 2018-03-02 Richard Biener <rguenther@suse.de>
4006 * config/darwin.c (saved_debug_info_level): New static global.
4007 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4008 (darwin_asm_lto_end): Restore debug info generation settings.
4010 2018-03-01 Martin Liska <mliska@suse.cz>
4013 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4016 2018-03-01 Richard Biener <rguenther@suse.de>
4019 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4020 type annotation with LTO when debug was disabled at compile-time.
4022 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4024 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4026 (mips_final_postscan_insn): Likewise.
4028 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4030 PR rtl-optimization/84528
4031 * alias.c (init_alias_target): Add commentary.
4032 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4033 a unique base value if the frame pointer is not eliminated
4034 to the stack pointer.
4036 2018-03-01 Tom de Vries <tom@codesourcery.com>
4038 PR rtl-optimization/83327
4039 * lra-int.h (hard_regs_spilled_into): Declare.
4040 * lra.c (hard_regs_spilled_into): Define.
4041 (init_reg_info): Init hard_regs_spilled_into.
4042 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4043 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4044 (process_bb_lives): Handle hard_regs_spilled_into.
4045 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4046 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4048 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4050 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4051 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4052 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4053 * config/rs6000/aix72.h: New file.
4055 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4057 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4058 instead of warning_at with conditional singular and plural messages
4062 * stor-layout.c (update_alignment_for_field): For
4063 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4064 && !DECL_PACKED (field), do the alignment update, just use
4065 only desired_align instead of MAX (type_align, desired_align)
4067 (place_field): Don't do known_align < desired_align handling
4068 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4069 is non-NULL, instead do it after rli->prev_field handling and
4070 only if not within a bitfield word. For DECL_PACKED (field)
4071 use type_align of BITS_PER_UNIT.
4073 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4075 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4076 superfluous parentheses and trailing spaces.
4078 2018-02-28 Richard Biener <rguenther@suse.de>
4080 PR tree-optimization/84584
4081 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4082 SCoPs with fake exit edge.
4084 2018-02-28 Martin Liska <mliska@suse.cz>
4087 * timevar.c (timer::print): Fix format to properly print 100%
4090 2018-02-28 Richard Biener <rguenther@suse.de>
4093 * genmatch.c (capture_info::walk_match): Do not mark
4094 captured expressions without operands as expr_p given
4095 they act more like predicates and should be subject to
4096 "lost tail" side-effect preserving.
4098 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4100 PR rtl-optimization/81611
4101 * auto-inc-dec.c (attempt_change): Move dead note from
4102 mem_insn if it's the next use of regno
4103 (find_address): Take address use of reg holding
4104 non-incremented value. Add parm to limit search to the named
4106 (merge_in_block): Attempt to use a mem insn that is the next
4107 use of the original regno.
4109 2018-02-27 Martin Sebor <msebor@redhat.com>
4112 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4113 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4115 2018-02-27 Martin Sebor <msebor@redhat.com>
4117 PR translation/84207
4118 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4119 n argument to unsigned HOST_WIDE_INT.
4120 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4121 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4122 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4123 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4125 2018-02-27 Richard Biener <rguenther@suse.de>
4127 PR tree-optimization/84512
4128 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4129 Do not use the estimate returned from record_stmt_cost for
4130 the scalar iteration cost but sum properly using add_stmt_cost.
4132 2018-02-27 Richard Biener <rguenther@suse.de>
4134 PR tree-optimization/84466
4135 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4136 Adjust last change to less strictly validate use operands.
4138 2018-02-27 Martin Liska <mliska@suse.cz>
4140 PR gcov-profile/84548
4141 * gcov.c (process_file): Allow partial overlap and consider it
4142 also as group functions.
4143 (output_lines): Properly calculate range of lines for a group.
4145 2018-02-27 Martin Liska <mliska@suse.cz>
4147 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4148 'ggc' suffixes. Change first column width.
4149 (timer::print): Fix formatting of the column.
4151 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4153 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4154 preserve inline entry blocks for the sake of debug inline
4155 entry point markers alone.
4156 (remove_unused_locals): Suggest in comments a better place to
4157 force the preservation of inline entry blocks that are
4158 otherwise unused, but do not preserve them.
4160 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4162 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4164 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4167 * config/i386/constraints.md (Bs): Replace
4168 ix86_indirect_branch_register with
4169 TARGET_INDIRECT_BRANCH_REGISTER.
4171 * config/i386/i386.md (indirect_jump): Likewise.
4172 (tablejump): Likewise.
4173 (*sibcall_memory): Likewise.
4174 (*sibcall_value_memory): Likewise.
4175 Peepholes of indirect call and jump via memory: Likewise.
4176 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4177 (*sibcall_value_GOT_32): Likewise.
4178 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4179 (GOT_memory_operand): Likewise.
4180 (call_insn_operand): Likewise.
4181 (sibcall_insn_operand): Likewise.
4182 (GOT32_symbol_operand): Likewise.
4183 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4185 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4187 PR rtl-optimization/83496
4188 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4189 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4190 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4191 redundant insn, if any.
4192 (relax_delay_slots): Likewise.
4193 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4195 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4197 PR tree-optimization/83965
4198 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4199 that grouped statements are part of a reduction chain. Return
4200 true if the statement is not marked as a reduction itself but
4202 (vect_recog_dot_prod_pattern): Don't check whether the statement
4203 is part of a group here.
4204 (vect_recog_sad_pattern): Likewise.
4205 (vect_recog_widen_sum_pattern): Likewise.
4207 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4210 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4212 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4215 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4217 (ix86_output_indirect_function_return): New prototype.
4218 (ix86_split_simple_return_pop_internal): Likewise.
4219 * config/i386/i386.c (indirect_return_via_cx): New.
4220 (indirect_return_via_cx_bnd): Likewise.
4221 (indirect_thunk_name): Handle return va CX_REG.
4222 (output_indirect_thunk_function): Create alias for
4223 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4224 (ix86_output_indirect_jmp): Remove the bool argument.
4225 (ix86_output_indirect_function_return): New function.
4226 (ix86_split_simple_return_pop_internal): Likewise.
4227 * config/i386/i386.md (*indirect_jump): Don't pass false
4228 to ix86_output_indirect_jmp.
4229 (*tablejump_1): Likewise.
4230 (simple_return_pop_internal): Change it to define_insn_and_split.
4231 Call ix86_split_simple_return_pop_internal to split it for
4233 (simple_return_indirect_internal): Call
4234 ix86_output_indirect_function_return instead of
4235 ix86_output_indirect_jmp.
4237 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4240 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4241 memset and value initialization afterwards.
4243 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4245 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4247 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4250 * common/config/aarch64/aarch64-common.c
4251 (aarch_option_optimization_table[]): Switch
4252 off fomit-frame-pointer
4254 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4255 Chung-Ju Wu <jasonwucj@gmail.com>
4257 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4259 (store_multiple): Ditto.
4261 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4263 * config.gcc: Add --with-cpu support for nds32 target.
4264 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4265 * config/nds32/nds32.opt: Add -mcpu= option.
4267 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4269 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4270 isel=yes): Warn for these deprecated options.
4272 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4274 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4275 ISA_2_5_MASKS_EMBEDDED.
4277 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4279 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4280 p->max as pointers rather than using iterative_hash_expr.
4282 2018-02-23 Carl Love <cel@us.ibm.com>
4284 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4285 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4287 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4288 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4289 P8V_BUILTIN_VEC_VUNSIGNED2.
4291 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4294 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4295 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4297 * lra-constraints.c (curr_insn_transform): Set up
4298 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4300 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4301 * lra-eliminations.c (spill_pseudos): Ditto.
4302 (process_insn_for_elimination): Ditto.
4303 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4304 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4305 LRA_NON_CLOBBERED_ALT.
4307 2018-02-22 Martin Sebor <msebor@redhat.com>
4309 PR tree-optimization/84480
4310 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4311 to maybe_diag_stxncpy_trunc. Call it.
4312 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4313 from gimple_fold_builtin_strcpy. Print inlining stack.
4314 (handle_builtin_stxncpy): Print inlining stack.
4315 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4317 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4320 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4321 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4322 and -fcheck-pointer-bounds are used together.
4323 (indirect_thunk_prefix): New enum.
4324 (indirect_thunk_need_prefix): New function.
4325 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4326 "_nt" instead of "_bnd" for NOTRACK prefix.
4327 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4328 (output_indirect_thunk_function): Likewise.
4330 (ix86_code_end): Update output_indirect_thunk_function calls.
4331 (ix86_output_indirect_branch_via_reg): Replace
4332 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4333 (ix86_output_indirect_branch_via_push): Likewise.
4334 (ix86_output_function_return): Likewise.
4335 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4336 incompatible with -fcf-protection=branch and
4337 -fcheck-pointer-bounds.
4339 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4342 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4343 Change gcc_assert call to output_operand_lossage.
4345 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4347 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4349 2018-02-22 DJ Delorie <dj@redhat.com>
4350 Sebastian Perta <sebastian.perta@renesas.com>
4351 Oleg Endo <olegendo@gcc.gnu.org>
4353 * config/rx/rx.c (rx_rtx_costs): New function.
4354 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4356 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4358 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4360 2018-02-22 Martin Liska <mliska@suse.cz>
4363 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4364 Add "native" as a possible value.
4366 2018-02-22 Martin Liska <mliska@suse.cz>
4369 * config/i386/i386.c (ix86_option_override_internal):
4370 Add "native" as a possible value for -march and -mtune.
4372 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4375 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4376 to all type variants.
4378 PR tree-optimization/84503
4379 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4380 width as info->bitpos + info->bitsize - start.
4381 (merged_store_group::merge_overlapping): Simplify width computation.
4382 (check_no_overlap): New function.
4383 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4384 start + width and last_order of the group, fail if check_no_overlap
4386 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4387 to group if check_no_overlap fails.
4389 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4391 * config/rs6000/altivec.md: Delete contraint arguments to
4392 define_expand, define_split, and define_peephole2, and in
4393 define_insn_and_split if always unused.
4394 * config/rs6000/darwin.md: Ditto.
4395 * config/rs6000/dfp.md: Ditto.
4396 * config/rs6000/rs6000.md: Ditto.
4397 * config/rs6000/sync.md: Ditto.
4398 * config/rs6000/vector.md: Ditto.
4399 * config/rs6000/vsx.md: Ditto.
4401 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4403 * config/rs6000/altivec.md: Write output control strings as braced
4404 blocks instead of double-quoted strings.
4405 * config/rs6000/darwin.md: Ditto.
4406 * config/rs6000/rs6000.md: Ditto.
4407 * config/rs6000/vector.md: Ditto.
4408 * config/rs6000/vsx.md: Ditto.
4410 2018-02-21 Jason Merrill <jason@redhat.com>
4412 PR c++/84314 - ICE with templates and fastcall attribute.
4413 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4415 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4417 * ipa-cp.c (determine_versionability): Fix comment typos.
4419 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4422 * ipa-cp.c (determine_versionability): Do not version functions caling
4425 2018-02-21 Martin Liska <mliska@suse.cz>
4428 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4429 Add "native" as a possible value.
4430 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4431 the macro when native cpu detection is available.
4433 2018-02-21 Martin Liska <mliska@suse.cz>
4436 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4437 Add "native" as a possible value.
4438 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4439 when native cpu detection is available.
4441 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4442 Martin Sebor <msebor@redhat.com>
4444 PR tree-optimization/84478
4445 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4447 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4448 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4449 support which is conservatively correct, for 2 only stay conservative
4450 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4451 argument to the 2 argument get_range_strlen, adjust 6 arg
4452 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4454 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4455 (gimple_fold_builtin_strlen): Pass true as last argument to
4458 2018-02-20 Martin Sebor <msebor@redhat.com>
4461 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4462 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4463 (builtin_memref::builtin_memref): Factor out parts into
4464 set_base_and_offset and call it.
4466 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4469 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4470 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4471 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4472 search at the associated MODE_INT.
4474 2018-02-20 Jeff Law <law@redhat.com>
4477 PR tree-optimization/81592
4479 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4480 for range data rather than using global data.
4481 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4482 range data rather than using global data.
4483 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4484 pass it to children as needed.
4485 (struct directive::fmtresult): Similarly.
4486 (struct directive::set_width): Similarly.
4487 (struct directive::set_precision): Similarly.
4488 (format_integer, format_directive, parse_directive): Similarly.
4489 (format_none): Accept unnamed vr_values parameter.
4490 (format_percent, format_floating, format_character): Similarly.
4491 (format_string, format_plain): Similarly.
4492 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4493 the EVRP range analyzer for range data rather than using global data.
4494 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4495 gimple-ssa-evrp-analyze.h
4496 (class sprintf_dom_walker): Add after_dom_children member function.
4497 Add evrp_range_analyzer member.
4498 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4499 range analyzer as needed.
4500 (sprintf_dom_walker::after_dom_children): New member function.
4501 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4503 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4504 (evrp_range_analyzer::pop_to_marker): Likewise.
4506 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4508 PR tree-optimization/84419
4509 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4510 with the required type if its current type is compatible but
4513 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4516 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4517 after vectorization.
4519 2018-02-20 Martin Liska <mliska@suse.cz>
4522 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4523 possible values if we don't have a hint.
4525 2018-02-20 Martin Liska <mliska@suse.cz>
4529 * final.c (shorten_branches): Build align_tab array with one
4531 * opts.c (finish_options): Add alignment option limit check.
4532 (MAX_CODE_ALIGN): Likewise.
4533 (MAX_CODE_ALIGN_VALUE): Likewise.
4534 * doc/invoke.texi: Document maximum allowed option value for
4535 all -falign-* options.
4537 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4540 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4541 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4542 * var-tracking.c (emit_note_insn_var_location): Remove all references
4543 to NOTE_INSN_CALL_ARG_LOCATION.
4544 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4545 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4546 Use copy_rtx_if_shared.
4547 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4548 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4549 (dwarf2out_var_location): Remove handling of
4550 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4552 * final.c (final_scan_insn): Remove all references to
4553 NOTE_INSN_CALL_ARG_LOCATION.
4554 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4555 before dumping final insns.
4556 * except.c (emit_note_eh_region_end): Remove all references to
4557 NOTE_INSN_CALL_ARG_LOCATION.
4558 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4559 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4560 * config/arc/arc.c (hwloop_optimize): Likewise.
4561 * config/arm/arm.c (create_fix_barrier): Likewise.
4562 * config/s390/s390.c (s390_chunkify_start): Likewise.
4563 * config/sh/sh.c (find_barrier): Likewise.
4564 * config/i386/i386.c (rest_of_insert_endbranch,
4565 ix86_seh_fixup_eh_fallthru): Likewise.
4566 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4567 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4568 * config/frv/frv.c (frv_function_prologue): Likewise.
4569 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4571 (note_outside_basic_block_p): Remove all references to
4572 NOTE_INSN_CALL_ARG_LOCATION.
4573 * gengtype.c (adjust_field_rtx_def): Likewise.
4574 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4576 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4577 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4580 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4583 PR tree-optimization/84452
4584 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4585 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4588 2018-02-19 Martin Liska <mliska@suse.cz>
4591 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4593 2018-02-19 Martin Liska <mliska@suse.cz>
4594 Richard Sandiford <richard.sandiford@linaro.org>
4596 PR tree-optimization/82491
4597 * gimple-fold.c (get_base_constructor): Make earlier bail out
4600 2018-02-19 Carl Love <cel@us.ibm.com>
4602 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4603 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4605 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4606 P8V_BUILTIN_VEC_NEG.
4608 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4610 * config/rl78/rl78.md (movdf): New define expand.
4612 2018-02-19 Martin Liska <mliska@suse.cz>
4615 * doc/invoke.texi: Fix typo.
4616 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4618 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4620 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4621 handle rs6000_single_float and rs6000_double_float specially for
4624 2018-02-16 Jeff Law <law@redhat.com>
4626 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4629 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4632 * ipa-inline.c (inline_small_functions): Fix a typo.
4634 2018-02-16 Nathan Sidwell <nathan@acm.org>
4636 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4638 2018-02-16 Carl Love <cel@us.ibm.com>
4640 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4641 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4642 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4643 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4644 expansion to P8V_BUILTIN_VEC_FLOAT2.
4646 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4648 PR rtl-optimization/70023
4649 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4650 src_regno into account.
4652 2018-02-16 Carl Love <cel@us.ibm.com>
4654 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4655 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4656 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4657 * config/rs6000/rs6000.c: Remove case statements for
4658 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4659 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4660 and P9V_BUILTIN_VEC_VINSERT4B.
4661 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4662 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4663 * config/rs6000/vsx.md:
4664 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4667 2018-02-16 Carl Love <cel@us.ibm.com>
4669 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4671 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4673 * config/rs6000/rs6000-c.c: Add the definitions for
4674 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4675 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4676 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4677 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4678 definition for insert4b and define insn *insert3b_internal.
4679 * doc/extend.texi: Add documentation for vec_extract4b.
4681 2018-02-16 Nathan Sidwell <nathan@acm.org>
4683 * doc/extend.texi (Backwards Compatibility): Mention friend
4684 injection. Note for-scope is deprecated.
4685 * doc/invoke.texi (-ffriend-injection): Deprecate.
4687 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4689 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4690 that moved to I2, also allow destinations that are a paradoxical
4691 subreg (instead of a normal reg).
4693 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4696 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4699 2018-02-16 Richard Biener <rguenther@suse.de>
4701 PR tree-optimization/84037
4702 PR tree-optimization/84016
4704 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4705 Adjust vec_construct for the fact we need additional higher latency
4706 128bit inserts for AVX256 and AVX512 vector builds.
4707 (ix86_add_stmt_cost): Scale vector construction cost for
4710 2018-02-16 Richard Biener <rguenther@suse.de>
4712 PR tree-optimization/84417
4713 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4714 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4715 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4717 2018-02-16 Martin Liska <mliska@suse.cz>
4720 * internal-fn.def (ASAN_CHECK): Set proper flags.
4721 (ASAN_MARK): Likewise.
4723 2018-02-16 Julia Koval <julia.koval@intel.com>
4725 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4726 from PTA_CANNONLAKE.
4728 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4731 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4732 Use ++iter rather than iter++ for std::list iterators.
4733 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4734 defer deleting them until all nodes in the forest are processed. Do
4735 free even leaf nodes. Change to_process into auto_vec.
4738 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4739 * vec.h (vec_default_construct): Use memset instead of placement new
4740 if BROKEN_VALUE_INITIALIZATION is defined.
4741 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4742 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4745 PR rtl-optimization/83723
4746 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4747 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4748 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4750 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4752 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4754 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4756 PR rtl-optimization/81443
4757 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4758 from inner REGs to paradoxical SUBREGs.
4760 2018-02-16 Richard Biener <rguenther@suse.de>
4762 PR tree-optimization/84399
4763 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4764 For operands we can analyze at their definition make sure we can
4765 analyze them at each use as well.
4767 2018-02-16 Richard Biener <rguenther@suse.de>
4769 PR tree-optimization/84190
4770 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4771 volatile accesses if the decl isn't volatile.
4773 2018-02-15 Jason Merrill <jason@redhat.com>
4775 PR c++/84314 - ICE with templates and fastcall attribute.
4776 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4777 TYPE_CANONICAL on an existing type.
4779 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4781 PR tree-optimization/84383
4782 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4783 dstoff nor call operand_equal_p if dstbase is NULL.
4785 PR tree-optimization/84334
4786 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4787 also a CONSTANT_CLASS_P, punt.
4789 2018-02-14 Jim Wilson <jimw@sifive.com>
4791 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4792 first SMALL_OPERAND check. New local min_second_step. Move assert
4793 to where locals are set. Add TARGET_RVC support.
4794 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4796 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4798 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4800 2018-02-14 Martin Sebor <msebor@redhat.com>
4802 PR tree-optimization/83698
4803 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4804 arrays constrain the offset range to their bounds.
4805 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4806 (builtin_access::overlap): Avoid setting the size of overlap if it's
4808 (maybe_diag_overlap): Also consider arrays when deciding what values
4809 of offsets to include in diagnostics.
4811 2018-02-14 Martin Sebor <msebor@redhat.com>
4814 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4815 that correspond to the kind of a declaration.
4817 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4820 * config/pa/pa.md: Load address of PIC label using the linkage table
4821 if the label is nonlocal.
4823 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4825 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4826 warning message if user requests -maltivec=be.
4827 * doc/invoke.texi: Document deprecation of -maltivec=be.
4829 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4832 * config/rs6000/rs6000-c.c: Update definitions for
4833 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4834 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4836 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4839 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4840 add _get_ssp intrinsics. Remove argument from
4841 __builtin_ia32_rdssp[d|q].
4842 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4843 * config/i386/i386-builtin.def: Remove argument from
4844 __builtin_ia32_rdssp[d|q].
4845 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4846 ix86_expand_special_args_builtin for _rdssp[d|q].
4847 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4848 Clear register before usage.
4849 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4850 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4852 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4854 PR tree-optimization/84357
4855 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4856 operand 1 of an ARRAY_REF too.
4858 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4861 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4862 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4864 (set_of_reg): New struct.
4865 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4866 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4867 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4869 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4870 Split into bitclr, bitset, bitinvert patterns if appropriate.
4871 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4872 use rx_fuse_in_memory_bitop.
4873 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4874 to named insn, correct maximum insn length.
4876 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4879 * machmode.def: Define a complex mode for PARTIAL_INT.
4880 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4882 * doc/rtl.texi: Document CSPImode.
4883 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4885 (msp430_hard_regno_nregs_with_padding): Likewise.
4887 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4890 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4892 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4894 PR rtl-optimization/84169
4895 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4896 we generated a parallel as new i3 and we split that to new i2 and i3
4897 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4898 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4899 those to i2, not i1. Partially rewrite this scan code.
4901 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4904 * stor-layout.c (place_field): For variable length fields, adjust
4905 offset_align afterwards not just based on the field's alignment,
4906 but also on the size.
4909 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4910 of exps and logs in the use_exp2 case.
4912 2018-02-13 Jeff Law <law@redhat.com>
4914 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4917 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4920 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4924 * common.opt (gas-loc-support, gas-locview-support): New.
4925 (ginline-points, ginternal-reset-location-views): New.
4926 * doc/invoke.texi: Document them. Use @itemx where intended.
4927 (gvariable-location-views): Adjust.
4928 * target.def (reset_location_view): New.
4929 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4930 (TARGET_RESET_LOCATION_VIEW): New.
4931 * doc/tm.texi: Rebuilt.
4932 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4933 (dwarf2out_default_as_locview_support): New.
4934 (output_asm_line_debug_info): Use option variables.
4935 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4936 (output_loc_list): Likewise.
4937 (add_high_low_attributes): Check option variables.
4938 Don't output entry view attribute in strict mode.
4939 (gen_inlined_subroutine_die): Check option variables.
4940 (dwarf2out_inline_entry): Likewise.
4941 (init_sections_and_labels): Likewise.
4942 (dwarf2out_early_finish): Likewise.
4943 (maybe_reset_location_view): New, from...
4944 (dwarf2out_var_location): ... here. Call it.
4945 * debug.h (dwarf2out_default_as_loc_support): Declare.
4946 (dwarf2out_default_as_locview_support): Declare.
4947 * hooks.c (hook_int_rtx_insn_0): New.
4948 * hooks.h (hook_int_rtx_insn_0): Declare.
4949 * toplev.c (process_options): Take -gas-loc-support and
4950 -gas-locview-support from dwarf2out. Enable
4951 -gvariable-location-views by default only with locview
4952 assembler support. Enable -ginternal-reset-location-views by
4953 default only if the target defines the corresponding hook.
4954 Enable -ginline-points by default if location views are
4955 enabled; force it disabled if statement frontiers are
4957 * tree-inline.c (expand_call_inline): Check option variables.
4958 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4960 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4962 PR tree-optimization/84321
4963 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4964 handling. Also check whether the anti-range contains any values
4965 that satisfy the mask; switch to a VR_RANGE if not.
4967 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4970 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4972 2018-02-13 Martin Jambor <mjambor@suse.cz>
4975 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4976 of call statements, also set location of a load to a temporary.
4978 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4980 * config/rl78/rl78.c (add_vector_labels): New function.
4981 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4982 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4983 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4984 which checks that no arguments are passed.
4985 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4986 * doc/extend.texi: Documentation for the new attribute.
4988 2018-02-13 Andreas Schwab <schwab@suse.de>
4990 * config/riscv/linux.h (CPP_SPEC): Define.
4992 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4995 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4996 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4997 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4998 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4999 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5000 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5001 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5002 temporarily for AES and PCLMUL builtins.
5004 PR tree-optimization/84339
5005 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5006 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5010 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5011 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5012 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5014 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5016 * omp-simd-clone.h: New file.
5017 * omp-simd-clone.c: Include omp-simd-clone.h.
5018 (expand_simd_clones): No longer static.
5019 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5020 cgraph.h and omp-simd-clone.h.
5021 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5022 (vect_recog_widen_shift_pattern): Formatting fix.
5023 (vect_pattern_recog_1): Don't check optab for calls.
5026 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5027 operands[2] into a REG before using gen_lowpart on it.
5029 2018-02-12 Jeff Law <law@redhat.com>
5032 * config/sh/sh.c (find_barrier): Consider a sibling call
5035 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5036 successfully back substituting a reg.
5038 2018-02-12 Richard Biener <rguenther@suse.de>
5040 PR tree-optimization/84037
5041 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5042 parameter, move visited init to caller.
5043 (vect_slp_analyze_operations): Separate cost from validity
5044 check, initialize visited once for all instances.
5045 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5047 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5049 (vect_model_promotion_demotion_cost): Likewise.
5050 (vectorizable_bswap): Guard cost modeling with !slp_node
5051 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5053 (vectorizable_call): Likewise.
5054 (vectorizable_conversion): Likewise.
5055 (vectorizable_assignment): Likewise.
5056 (vectorizable_shift): Likewise.
5057 (vectorizable_operation): Likewise.
5058 (vectorizable_store): Likewise.
5059 (vectorizable_load): Likewise.
5060 (vectorizable_condition): Likewise.
5061 (vectorizable_comparison): Likewise.
5063 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5066 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5067 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5068 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5070 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5073 * params.def (inline-min-speedup): Increase from 8 to 15.
5074 (max-inline-insns-auto): Decrease from 40 to 30.
5075 * ipa-split.c (consider_split): Add some buffer for function to
5076 be considered inlining candidate.
5077 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5080 2018-02-12 Richard Biener <rguenther@suse.de>
5082 PR tree-optimization/84037
5083 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5084 matched stmts if we cannot swap the non-matched ones.
5086 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5088 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5089 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5090 _mm_maskz_scalef_round_ss): New intrinsics.
5091 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5092 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5093 __builtin_ia32_scalefss_round): Remove.
5094 (__builtin_ia32_scalefsd_mask_round,
5095 __builtin_ia32_scalefss_mask_round): New intrinsics.
5096 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5097 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5098 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5099 "<round_constraint>")): Changed to ...
5100 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5101 "<round_scalar_constraint>")): ... this.
5102 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5103 %0, %1, %2<round_op3>}"): Changed to ...
5104 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5105 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5106 %2<round_scalar_mask_op3>}"): ... this.
5107 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5109 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5111 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5112 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5113 (_mm_maskz_sqrt_round_ss): New intrinsics.
5114 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5115 (__builtin_ia32_sqrtsd_mask_round)
5116 (__builtin_ia32_sqrtss_mask_round): New builtins.
5117 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5118 (__builtin_ia32_sqrtss_round): Remove.
5119 (__builtin_ia32_sqrtsd_mask_round)
5120 (__builtin_ia32_sqrtss_mask_round): New builtins.
5121 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5122 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5123 ((match_operand:VF_128 1 "vector_operand"
5124 "xBm,<round_constraint>")): Changed to ...
5125 ((match_operand:VF_128 1 "vector_operand"
5126 "xBm,<round_scalar_constraint>")): ... this.
5127 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5128 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5129 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5130 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5131 %<iptr>1<round_scalar_mask_op3>}): ... this.
5132 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5133 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5135 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5138 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5139 Cast vec_cmpeq result to correct type.
5140 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5141 Cast vec_cmpgt result to correct type.
5143 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5145 * final.c (final_scan_insn_1): Renamed from...
5146 (final_scan_insn): ... this. New wrapper, to recover
5147 seen from the outermost call in recursive ones.
5148 * config/sparc/sparc.c (output_return): Drop seen from call.
5149 (output_sibcall): Likewise.
5150 * config/visium/visium.c (output_branch): Likewise.
5152 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5154 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5157 2018-02-10 Alan Modra <amodra@gmail.com>
5160 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5161 Specify LR as an input.
5163 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5166 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5167 remove_member_access_dummy_vars): New functions.
5168 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5169 lower_omp_1, execute_lower_omp): Use them.
5171 PR rtl-optimization/84308
5172 * shrink-wrap.c (spread_components): Release todo vector.
5174 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5176 PR rtl-optimization/57193
5177 * ira-color.c (struct allocno_color_data): Add member
5178 conflict_allocno_hard_prefs.
5179 (update_conflict_allocno_hard_prefs): New.
5180 (bucket_allocno_compare_func): Add a preference based on
5181 conflict_allocno_hard_prefs.
5182 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5183 (color_allocnos): Remove a dead code. Initiate
5184 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5186 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5189 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5190 constraint from =wa to wa. Avoid a subreg on the output operand,
5191 instead use a pseudo and subreg it in a move.
5192 (p9_xxbrd_<mode>): Changed to ...
5193 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5194 (p9_xxbrd_v2df): New expander.
5195 (p9_xxbrw_<mode>): Changed to ...
5196 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5197 (p9_xxbrw_v4sf): New expander.
5199 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5201 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5203 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5206 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5207 multiply in 32-bit mode.
5208 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5209 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5212 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5214 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5215 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5216 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5217 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5219 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5222 * dwarf2out.c (is_trivial_indirect_ref): New function.
5223 (dwarf2out_late_global_decl): Do not generate a location
5224 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5225 and that are not defined in the current unit.
5227 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5229 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5230 instead of a libcall for UNORDERED.
5232 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5235 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5236 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5238 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5241 * config/s390/s390.c (s390_set_current_function): Invoke
5242 s390_indirect_branch_settings also if fndecl didn't change.
5244 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5246 * config/rs6000/rs6000.md (blockage): Set length to zero.
5248 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5250 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5252 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5255 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5256 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5260 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5261 PARALLEL incoming that failed vt_get_decl_and_offset check.
5264 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5265 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5267 (get_variable_section): For decls in named .bss* sections pass true as
5268 second argument to bss_initializer_p.
5270 2018-02-09 Marek Polacek <polacek@redhat.com>
5271 Jakub Jelinek <jakub@redhat.com>
5274 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5275 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5276 Sync some changes from cxx_fold_indirect_ref.
5278 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5280 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5282 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5283 (BLOCK_INLINE_ENTRY_LABEL): New.
5284 (dwarf2out_var_location): Disregard inline entry markers.
5285 (inline_entry_data): New struct.
5286 (inline_entry_data_hasher): New hashtable type.
5287 (inline_entry_data_hasher::hash): New.
5288 (inline_entry_data_hasher::equal): New.
5289 (inline_entry_data_table): New variable.
5290 (add_high_low_attributes): Add DW_AT_entry_pc and
5291 DW_AT_GNU_entry_view attributes if a pending entry is found
5292 in inline_entry_data_table. Add old entry_pc attribute only
5293 if debug nonbinding markers are disabled.
5294 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5295 markers are enabled.
5296 (block_within_block_p, dwarf2out_inline_entry): New.
5297 (dwarf2out_finish): Check that no entries remained in
5298 inline_entry_data_table.
5299 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5300 (final_scan_insn, notice_source_line): Likewise.
5301 (rest_of_clean_state): Skip inline entry markers.
5302 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5304 * gimple.c (gimple_build_debug_inline_entry): New.
5305 * gimple.h (enum gimple_debug_subcode): Add
5306 GIMPLE_DEBUG_INLINE_ENTRY.
5307 (gimple_build_debug_inline_entry): Declare.
5308 (gimple_debug_inline_entry_p): New.
5309 (gimple_debug_nonbind_marker_p): Adjust.
5310 * insn-notes.def (INLINE_ENTRY): New.
5311 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5312 inline entry marker notes.
5313 (print_insn): Likewise.
5314 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5315 (INSN_DEBUG_MARKER_KIND): Likewise.
5316 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5317 * tree-inline.c (expand_call_inline): Build and insert
5318 debug_inline_entry stmt.
5319 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5320 inline entry blocks early, if nonbind markers are enabled.
5321 (dump_scope_block): Dump fragment info.
5322 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5323 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5324 (gimple_build_debug_inline_entry): New.
5325 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5326 Enable/disable inline entry points too.
5327 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5328 (DEBUG_INSN): Describe inline entry markers.
5330 * common.opt (gvariable-location-views): New.
5331 (gvariable-location-views=incompat5): New.
5332 * config.in: Rebuilt.
5333 * configure: Rebuilt.
5334 * configure.ac: Test assembler for view support.
5335 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5336 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5337 * dwarf2out.c (var_loc_view): New typedef.
5338 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5339 (dwarf2out_locviews_in_attribute): New.
5340 (dwarf2out_locviews_in_loclist): New.
5341 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5342 (enum dw_line_info_opcode): Add LI_adv_address.
5343 (struct dw_line_info_table): Add view.
5344 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5345 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5346 (zero_view_p): New variable.
5347 (ZERO_VIEW_P): New macro.
5348 (output_asm_line_debug_info): New.
5349 (struct var_loc_node): Add view.
5350 (add_AT_view_list, AT_loc_list): New.
5351 (add_var_loc_to_decl): Add view param. Test it against last.
5352 (new_loc_list): Add view params. Record them.
5353 (AT_loc_list_ptr): Handle loc and view lists.
5354 (view_list_to_loc_list_val_node): New.
5355 (print_dw_val): Handle dw_val_class_view_list.
5356 (size_of_die): Likewise.
5357 (value_format): Likewise.
5358 (loc_list_has_views): New.
5359 (gen_llsym): Set vl_symbol too.
5360 (maybe_gen_llsym, skip_loc_list_entry): New.
5361 (dwarf2out_maybe_output_loclist_view_pair): New.
5362 (output_loc_list): Output view list or entries too.
5363 (output_view_list_offset): New.
5364 (output_die): Handle dw_val_class_view_list.
5365 (output_dwarf_version): New.
5366 (output_compilation_unit_header): Use it.
5367 (output_skeleton_debug_sections): Likewise.
5368 (output_rnglists, output_line_info): Likewise.
5369 (output_pubnames, output_aranges): Update version comments.
5370 (output_one_line_info_table): Output view numbers in asm comments.
5371 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5372 Call maybe_gen_llsym.
5373 (loc_list_from_tree_1): Adjust.
5374 (add_AT_location_description): Create view list attribute if
5375 needed, check it's absent otherwise.
5376 (convert_cfa_to_fb_loc_list): Adjust.
5377 (maybe_emit_file): Call output_asm_line_debug_info for test.
5378 (dwarf2out_var_location): Reset views as needed. Precompute
5379 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5380 attribute. Set view.
5381 (new_line_info_table): Reset next view.
5382 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5383 (dwarf2out_source_line): Likewise. Output view resets and labels to
5384 the assembler, or select appropriate line info opcodes.
5385 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5386 (optimize_string_length): Catch it. Adjust.
5387 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5388 dw_val_class_view_list, and remove it if no longer needed.
5389 (hash_loc_list): Hash view numbers.
5390 (loc_list_hasher::equal): Compare them.
5391 (optimize_location_lists): Check whether a view list symbol is
5392 needed, and whether the locview attribute is present, and
5393 whether they match. Remove the locview attribute if no longer
5395 (index_location_lists): Call skip_loc_list_entry for test.
5396 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5397 Use output_dwarf_version.
5398 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5399 (struct dw_val_node): Add val_view_list.
5400 * final.c (SEEN_NEXT_VIEW): New.
5401 (set_next_view_needed): New.
5402 (clear_next_view_needed): New.
5403 (maybe_output_next_view): New.
5404 (final_start_function): Rename to...
5405 (final_start_function_1): ... this. Take pointer to FIRST,
5406 add SEEN parameter. Emit param bindings in the initial view.
5407 (final_start_function): Reintroduce SEEN-less interface.
5408 (final): Rename to...
5409 (final_1): ... this. Take SEEN parameter. Output final pending
5410 next view at the end.
5411 (final): Reintroduce seen-less interface.
5412 (final_scan_insn): Output pending next view before switching
5413 sections or ending a block. Mark the next view as needed when
5414 outputting variable locations. Notify debug backend of section
5415 changes, and of location view changes.
5416 (rest_of_handle_final): Adjust.
5417 * toplev.c (process_options): Autodetect value for debug variable
5418 location views option. Warn on incompat5 without -gdwarf-5.
5419 * doc/invoke.texi (gvariable-location-views): New.
5420 (gvariable-location-views=incompat5): New.
5421 (gno-variable-location-views): New.
5423 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5425 PR tree-optimization/84136
5426 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5427 that the result of find_edge is non-NULL.
5429 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5432 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5433 storing integer register in SImode. Fix cost of 256 and 512
5434 byte aligned SSE register store.
5436 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5438 * config/i386/i386.c (ix86_multiplication_cost): Fix
5439 multiplication cost for TARGET_AVX512DQ.
5441 2018-02-08 Marek Polacek <polacek@redhat.com>
5443 PR tree-optimization/84238
5444 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5447 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5449 PR tree-optimization/84265
5450 * tree-vect-stmts.c (vectorizable_store): Don't treat
5451 VMAT_CONTIGUOUS accesses as grouped.
5452 (vectorizable_load): Likewise.
5454 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5456 PR tree-optimization/81635
5457 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5458 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5459 (test_round_for_mask): New functions.
5460 (wide_int_cc_tests): Call test_round_for_mask.
5461 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5462 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5463 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5464 range returned by get_range_info.
5466 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5469 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5470 * symtab.c: Include builtins.h
5471 (symtab_node::output_to_lto_symbol_table_p): Move here
5472 from lto-streamer-out.c:output_symbol_p.
5473 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5474 (output_symbol_p): Move all logic to symtab.c
5475 (produce_symtab): Update.
5477 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5479 * config/s390/s390-opts.h (enum indirect_branch): Define.
5480 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5481 (s390_indirect_branch_via_thunk)
5482 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5483 (enum s390_indirect_branch_type): Define.
5484 * config/s390/s390.c (struct s390_frame_layout, struct
5485 machine_function): Remove.
5486 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5487 (indirect_branch_table_label_no, indirect_branch_table_name):
5489 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5490 (enum s390_indirect_branch_option): Define.
5491 (s390_return_addr_from_memory): New function.
5492 (s390_handle_string_attribute): New function.
5493 (s390_attribute_table): Add new attribute handler.
5494 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5495 (s390_indirect_branch_via_thunk): New function.
5496 (s390_indirect_branch_via_inline_thunk): New function.
5497 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5498 sibling call optimization for non z10 compiles.
5499 (s390_emit_call): Force indirect branch target to be a single
5500 register. Add r1 clobber for non-z10 compiles.
5501 (s390_emit_epilogue): Emit return jump via return_use expander.
5502 (s390_reorg): Handle JUMP_INSNs as execute targets.
5503 (s390_option_override_internal): Perform validity checks for the
5504 new command line options.
5505 (s390_indirect_branch_attrvalue): New function.
5506 (s390_indirect_branch_settings): New function.
5507 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5508 (s390_output_indirect_thunk_function): New function.
5509 (s390_code_end): Implement target hook.
5510 (s390_case_values_threshold): Implement target hook.
5511 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5513 * config/s390/s390.h (struct s390_frame_layout)
5514 (struct machine_function): Move here from s390.c.
5515 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5516 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5517 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5518 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5519 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5520 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5521 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5522 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5523 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5524 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5525 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5526 (mnemonic attribute): Add values which aren't recognized
5528 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5529 pattern for branch conversion. Fix mnemonic attribute.
5530 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5531 indirect branch via thunk if requested.
5532 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5533 ("*indirect_jump"): Disable for branch conversion using out of
5535 ("indirect_jump_via_thunk<mode>_z10")
5536 ("indirect_jump_via_thunk<mode>")
5537 ("indirect_jump_via_inlinethunk<mode>_z10")
5538 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5539 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5540 ("casesi_jump_via_inlinethunk<mode>_z10")
5541 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5542 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5543 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5544 ("*indirect2_jump"): Disable for branch conversion.
5545 ("casesi_jump"): Turn into expander and expand patterns for branch
5547 ("return_use"): New expander.
5548 ("*return"): Emit return via thunk and rename it to ...
5549 ("*return<mode>"): ... this one.
5550 * config/s390/s390.opt: Add new options and and enum for the
5553 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5555 * lra-constraints.c (match_reload): Unconditionally use
5556 gen_lowpart_SUBREG, rather than selecting between that
5557 and equivalent gen_rtx_SUBREG code.
5559 2018-02-08 Richard Biener <rguenther@suse.de>
5561 PR tree-optimization/84233
5562 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5563 changed flag instead of boguously re-using phi_inserted.
5565 2018-02-08 Martin Jambor <mjambor@suse.cz>
5567 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5568 static local variables.
5570 2018-02-08 Richard Biener <rguenther@suse.de>
5572 PR tree-optimization/84278
5573 * tree-vect-stmts.c (vectorizable_store): When looking for
5574 smaller vector types to perform grouped strided loads/stores
5575 make sure the mode is supported by the target.
5576 (vectorizable_load): Likewise.
5578 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5580 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5581 Increase LDP/STP opportunities by adding adjacent callee-saves.
5583 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5585 PR rtl-optimization/84068
5586 PR rtl-optimization/83459
5587 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5589 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5591 PR tree-optimization/84224
5592 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5593 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5596 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5599 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5600 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5601 count, remove one USE.
5603 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5605 * doc/install.texi (Configuration): Document the
5606 --with-long-double-format={ibm,ieee} PowerPC configuration
5610 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5611 Convert from define_expand to be define_insn_and_split. Rework
5612 float/double/_Float128 conversions to QI/HI/SImode to work with
5613 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5614 conversions to QI/HImode types did a store and then a load to
5615 truncate the value. For conversions to VSX registers, don't split
5616 the insn, instead emit the code directly. Use the code iterator
5617 any_fix to combine signed and unsigned conversions.
5618 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5619 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5620 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5621 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5622 (fix_<mode>di2_hw): Likewise.
5623 (fixuns_<mode>di2_hw): Likewise.
5624 (fix_<mode>si2_hw): Likewise.
5625 (fixuns_<mode>si2_hw): Likewise.
5626 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5627 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5628 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5629 fix<uns>_trunc<SFDF:mode>si2_p8.
5630 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5632 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5633 (fix<uns>_<mode>_mem): Likewise.
5634 (fctiw<u>z_<mode>_mem): Likewise.
5635 (fix<uns>_<mode>_mem): Likewise.
5636 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5637 the register allocator from doing a direct move to the GPRs to do
5638 a store, and instead use the ISA 3.0 store byte/half-word from
5639 vector register instruction. For IEEE 128-bit floating point,
5640 also optimize stores of 32-bit ints.
5641 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5643 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5645 * genextract.c (push_pathstr_operand): New function to support
5647 (walk_rtx): Call push_pathstr_operand.
5648 (print_path): Support [a-zA-Z].
5650 2018-02-07 Richard Biener <rguenther@suse.de>
5652 PR tree-optimization/84037
5653 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5654 (cse_and_gimplify_to_preheader): Declare.
5655 (vect_get_place_in_interleaving_chain): Likewise.
5656 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5658 (_loop_vec_info::~_loop_vec_info): Delete it.
5659 (cse_and_gimplify_to_preheader): New function.
5660 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5661 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5662 (vectorizable_load): Likewise. For grouped stores always base
5663 the IV on the first element.
5664 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5665 condition before gimplifying.
5667 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5669 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5670 *DIV_EXPR and *MOD_EXPR.
5672 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5675 * config/i386/i386.c (ix86_option_override_internal): Mask out
5676 the CF_SET bit when checking -fcf-protection.
5678 2018-02-07 Tom de Vries <tom@codesourcery.com>
5681 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5684 2018-02-07 Richard Biener <rguenther@suse.de>
5686 PR tree-optimization/84204
5687 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5690 PR tree-optimization/84205
5691 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5692 special-case isl_ast_op_zdiv_r.
5694 PR tree-optimization/84223
5695 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5696 Only add conditions from within the region.
5697 (gather_bbs::after_dom_children): Adjust.
5699 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5702 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5703 * config/avr/avr.md: Only post-reload split REG-REG moves if
5704 either register is GENERAL_REG_P.
5706 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5708 PR tree-optimization/84235
5709 * tree-ssa-scopedtables.c
5710 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5711 if the subtraction is performed in floating point type where NaNs are
5712 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5713 build 1. Formatting fix.
5715 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5718 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5719 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5720 and skip it regardless of bb boundaries. Use CALL_P macro,
5721 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5722 unnecessarily, formatting fix.
5724 2018-02-06 Michael Collison <michael.collison@arm.com>
5726 * config/arm/thumb2.md:
5727 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5728 (*thumb_mov_notscc): Ditto.
5730 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5733 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5734 unsigned_fix, not "s".
5736 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5738 * configure.ac (gcc_fn_eh_frame_ro): New function.
5739 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5740 correct .eh_frame permissions.
5741 * configure: Regenerate.
5743 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5745 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5748 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5750 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5751 Display warning message for -mno-speculate-indirect-jumps.
5753 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5755 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5757 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5759 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5761 PR tree-optimization/84225
5762 * tree-eh.c (find_trapping_overflow): Only call
5763 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5765 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5768 * config/i386/i386.c: Reimplement the check of possible options
5769 -mibt/-mshstk conbination. Change error messages.
5770 * doc/invoke.texi: Fix a typo: remove extra '='.
5772 2018-02-06 Marek Polacek <polacek@redhat.com>
5774 PR tree-optimization/84228
5775 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5777 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5780 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5781 emitted arch directives.
5782 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5783 __ARM_FEATURE_COPROC before changing architectures.
5785 2018-02-06 Richard Biener <rguenther@suse.de>
5787 * config/i386/i386.c (print_reg): Fix typo.
5788 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5790 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5792 * configure: Regenerate.
5794 2018-02-05 Martin Sebor <msebor@redhat.com>
5796 PR tree-optimization/83369
5797 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5800 2018-02-05 Martin Liska <mliska@suse.cz>
5802 * doc/invoke.texi: Cherry-pick upstream r323995.
5804 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5806 * ira.c (ira_init_register_move_cost): Adjust comment.
5808 2018-02-05 Martin Liska <mliska@suse.cz>
5810 PR gcov-profile/84137
5811 * doc/gcov.texi: Fix typo in documentation.
5813 2018-02-05 Martin Liska <mliska@suse.cz>
5815 PR gcov-profile/83879
5816 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5817 using dlopen functionality.
5819 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5821 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5822 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5823 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5824 _mm_maskz_range_round_ss): New intrinsics.
5825 (__builtin_ia32_rangesd128_round)
5826 (__builtin_ia32_rangess128_round): Remove.
5827 (__builtin_ia32_rangesd128_mask_round,
5828 __builtin_ia32_rangess128_mask_round): New builtins.
5829 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5830 __builtin_ia32_rangess128_round): Remove.
5831 (__builtin_ia32_rangesd128_mask_round,
5832 __builtin_ia32_rangess128_mask_round): New builtins.
5833 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5834 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5835 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5836 "<round_saeonly_constraint>")): Changed to ...
5837 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5838 "<round_saeonly_scalar_constraint>")): ... this.
5839 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5840 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5841 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5842 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5843 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5845 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5847 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5849 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5850 Remove all values except native, 8540 and 8548.
5852 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5854 * config/i386/i386.c (ix86_output_function_return): Pass
5855 INVALID_REGNUM, instead of -1, as invalid register number to
5856 indirect_thunk_name and output_indirect_thunk.
5858 2018-02-02 Julia Koval <julia.koval@intel.com>
5860 * config.gcc: Add -march=icelake.
5861 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5862 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5863 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5864 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5865 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5866 (processor_target_table): Add icelake.
5867 (ix86_option_override_internal): Handle new PTAs.
5868 (get_builtin_code_for_version): Handle icelake.
5869 (M_INTEL_COREI7_ICELAKE): New.
5870 (fold_builtin_cpu): Handle icelake.
5871 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5872 * doc/invoke.texi: Add -march=icelake.
5874 2018-02-02 Julia Koval <julia.koval@intel.com>
5876 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5877 to wide_int_bitmask.
5878 * wide-int-bitmask.h: New.
5880 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5883 * config/i386/i386.md: Replace Pmode with word_mode in
5884 builtin_setjmp_setup and builtin_longjmp to support x32.
5886 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5890 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5892 (rs6000_supported_cpu_names): New static variable.
5893 (linux_cpu_translation_table): Likewise.
5894 (elf_platform) <cpu>: Define new static variable and use it.
5895 Translate kernel AT_PLATFORM name to canonical name if needed.
5896 Error if platform name is unknown.
5898 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5901 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5903 2018-02-01 Jeff Law <law@redhat.com>
5906 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5907 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5908 the scratch if RELEASE_VIA_POP is false.
5909 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5910 If we have to save a temporary register, decrement SIZE appropriately.
5911 Pass new arguments to release_scratch_register_on_entry.
5912 (ix86_adjust_stack_and_probe): Likewise.
5913 (ix86_emit_probe_stack_range): Pass new arguments to
5914 release_scratch_register_on_entry.
5916 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5918 PR rtl-optimization/84157
5919 * combine.c (change_zero_ext): Use REG_P predicate in
5920 front of HARD_REGISTER_P predicate.
5922 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5924 * config/avr/avr.c (avr_option_override): Move disabling of
5925 -fdelete-null-pointer-checks to...
5926 * common/config/avr/avr-common.c (avr_option_optimization_table):
5929 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5931 PR tree-optimization/81635
5932 * tree-data-ref.c (split_constant_offset_1): For types that
5933 wrap on overflow, try to use range info to prove that wrapping
5936 2018-02-01 Renlin Li <renlin.li@arm.com>
5939 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5941 (aarch64_register_move_cost): Likewise.
5942 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5944 (REG_CLASS_NAMES): Likewise.
5945 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5946 TAILCALL_ADDR_REGS. Remove IP registers.
5947 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5949 2018-02-01 Richard Biener <rguenther@suse.de>
5951 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5952 for specifying RPO order and allow NULL for that.
5953 * domwalk.c (dom_walker::dom_walker): Likewise.
5954 (dom_walker::walk): Handle NULL RPO order.
5955 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5957 (rewrite_update_dom_walker): Likewise.
5958 (mark_def_dom_walker): Likewise.
5960 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5962 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5963 (aarch64_maybe_expand_sve_subreg_move): Declare.
5964 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5965 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5967 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5968 that are semantically a reverse operation.
5969 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5970 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5971 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5973 (aarch64_can_change_mode_class): For big-endian, forbid changes
5974 between two SVE modes if they have different element sizes.
5976 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5978 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5979 the TImode handling for big-endian targets.
5981 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5983 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5984 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5986 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5987 Remove BSWAP handing for big-endian targets and use the form of
5988 LD1RQ appropariate for the mode.
5990 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5992 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5993 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5996 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5999 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6000 check for operands that need to go through aarch64_sve_reload_be.
6002 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6004 PR tree-optimization/81661
6005 PR tree-optimization/84117
6006 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6007 * tree-eh.c: Include gimplify.h.
6008 (find_trapping_overflow, replace_trapping_overflow,
6009 rewrite_to_non_trapping_overflow): New functions.
6010 * tree-vect-loop.c: Include tree-eh.h.
6011 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6012 * tree-data-ref.c: Include tree-eh.h.
6013 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6015 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6017 PR rtl-optimization/84123
6018 * combine.c (change_zero_ext): Check if hard register satisfies
6019 can_change_dest_mode before calling gen_lowpart_SUBREG.
6021 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6024 * ira.c (ira_init_register_move_cost): Remove assert.
6026 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6028 PR rtl-optimization/84071
6029 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6030 * doc/tm.texi: Regenerate.
6032 2018-01-31 Richard Biener <rguenther@suse.de>
6034 PR tree-optimization/84132
6035 * tree-data-ref.c (analyze_miv_subscript): Properly
6036 check whether evolution_function_is_affine_multivariate_p
6037 before calling gcd_of_steps_may_divide_p.
6039 2018-01-31 Julia Koval <julia.koval@intel.com>
6042 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6043 * config/i386/i386.md (rdpid_rex64) New.
6044 (rdpid): Make 32bit only.
6046 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6049 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6050 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6052 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6055 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6057 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6059 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6061 PR rtl-optimization/84071
6062 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6063 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6065 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6067 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6068 (arc_attribute_table): Add 'aux' attribute.
6069 (arc_in_small_data_p): Consider aux like variables.
6070 (arc_is_aux_reg_p): New function.
6071 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6072 (arc_get_aux_arg): New function.
6073 (prepare_move_operands): Handle aux-register access.
6074 (arc_handle_aux_attribute): New function.
6075 * doc/extend.texi (ARC Variable attributes): Add subsection.
6077 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6079 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6080 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6081 (arc_attribute_table): Add 'uncached' attribute.
6082 (arc_print_operand): Print '.di' flag for uncached memory
6084 (arc_in_small_data_p): Do not consider for small data the uncached
6086 (arc_is_uncached_mem_p): New function.
6087 * config/arc/predicates.md (compact_store_memory_operand): Check
6088 for uncached memory accesses.
6089 (nonvol_nonimm_operand): Likewise.
6090 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6092 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6095 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6096 falign-loops=): Add Optimization flag.
6098 2018-01-30 Jeff Law <law@redhat.com>
6101 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6102 INT_REGISTERS_SAVED. Check it prior to calling
6103 get_scratch_register_on_entry.
6104 (ix86_adjust_stack_and_probe): Similarly.
6105 (ix86_emit_probe_stack_range): Similarly.
6106 (ix86_expand_prologue): Corresponding changes.
6108 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6111 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6112 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6114 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6117 * lra-constraints.c (curr_insn_transform): Process AND in the
6120 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6122 PR rtl-optimization/83986
6123 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6124 dependence against last_pending_memory_flush in addition to
6127 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6129 PR tree-optimization/81611
6130 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6133 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6136 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6139 2018-01-30 Richard Biener <rguenther@suse.de>
6140 Jakub Jelinek <jakub@redhat.com>
6142 PR tree-optimization/84111
6143 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6144 inner loops added during recursion, as they don't have up-to-date
6147 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6150 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6151 (can_inline_edge_by_limits_p): ... here.
6152 (can_early_inline_edge_p, check_callers,
6153 update_caller_keys, update_callee_keys, recursive_inlining,
6154 add_new_edges_to_heap, speculation_useful_p,
6155 inline_small_functions,
6156 inline_small_functions, flatten_function,
6157 inline_to_all_callers_1): Update.
6159 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6161 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6164 2018-01-30 Richard Biener <rguenther@suse.de>
6166 PR tree-optimization/83008
6167 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6168 invariant and constant vector uses in stmts when they need
6171 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6174 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6175 * configure: Regenerate.
6177 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6179 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6181 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6182 Use gen_rtx_REG rather than gen_lowpart.
6184 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6186 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6187 rather than 0 when creating partial subregs.
6189 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6191 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6194 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6197 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6198 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6199 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6200 flags. This restores the settings used before the 2017-07-24.
6201 Turning off pre increment/decrement/modify allows IVOPTS to
6202 optimize DF/SF loops where the index is an int.
6204 2018-01-29 Richard Biener <rguenther@suse.de>
6205 Kelvin Nilsen <kelvin@gcc.gnu.org>
6208 * tree-vect-stmts.c (vectorizable_call): Don't call
6209 targetm.vectorize_builtin_md_vectorized_function if callee is
6212 2018-01-22 Carl Love <cel@us.ibm.com>
6214 * doc/extend.tex: Fix typo in second arg in
6215 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6217 2018-01-29 Richard Biener <rguenther@suse.de>
6219 PR tree-optimization/84086
6220 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6221 (flush_ssaname_freelist): When SSA names were released reset
6222 the SCEV hash table.
6224 2018-01-29 Richard Biener <rguenther@suse.de>
6226 PR tree-optimization/84057
6227 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6228 removed paths when removing edges.
6230 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6232 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6233 -mfunction-return=@var{choice}.
6235 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6238 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6240 (layout::print_source_line): Likewise.
6241 (test_get_line_width_without_trailing_whitespace): Add test cases.
6243 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6246 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6249 2018-01-26 Jim Wilson <jimw@sifive.com>
6251 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6253 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6256 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6258 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6259 and CMP + SUB-immediate -> SUBS.
6261 2018-01-26 Martin Sebor <msebor@redhat.com>
6263 PR tree-optimization/83896
6264 * tree-ssa-strlen.c (get_string_len): Rename...
6265 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6266 Avoid assuming length is constant.
6267 (handle_char_store): Use HOST_WIDE_INT for string length.
6269 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6272 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6273 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6275 2018-01-26 Richard Biener <rguenther@suse.de>
6277 PR rtl-optimization/84003
6278 * dse.c (record_store): Only record redundant stores when
6279 the earlier store aliases at least all accesses the later one does.
6281 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6283 PR rtl-optimization/83985
6284 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6285 REG_CFA_RESTORE insns.
6286 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6287 REG_CFA_RESTORE insns here.
6290 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6291 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6293 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6295 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6296 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6297 (arc_init): Likewise.
6298 (arc_override_options): Likewise.
6299 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6301 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6303 * config/arc/arc.h (TARGET_DBNZ): Define.
6304 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6305 properly set the tune attribute.
6306 (dbnz): Use TARGET_DBNZ guard.
6307 * config/arc/arc.opt (mtune): Add core3 option.
6309 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6311 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6312 recognize new pic like addresses.
6313 (arc_delegitimize_address): Clean up.
6315 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6317 * config/arc/arc-arches.def: Option mrf16 valid for all
6319 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6320 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6321 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6322 * config/arc/arc-tables.opt: Regenerate.
6323 * config/arc/arc.c (arc_conditional_register_usage): Handle
6324 reduced register file case.
6325 (arc_file_start): Set must have build attributes.
6326 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6328 * config/arc/arc.opt (mrf16): Add new option.
6329 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6330 * config/arc/genmultilib.awk: Handle new mrf16 option.
6331 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6332 * config/arc/t-multilib: Regenerate.
6333 * doc/invoke.texi (ARC Options): Document mrf16 option.
6335 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6337 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6338 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6339 (arc_attribute_table): Add 'secure_call' attribute.
6340 (arc_print_operand): Print secure call operand.
6341 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6343 (arc_is_secure_call_p): New function. * config/arc/arc.md
6344 (call_i): Add support for sjli instruction.
6345 (call_value_i): Likewise.
6346 * config/arc/constraints.md (Csc): New constraint.
6348 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6349 John Eric Martin <John.Martin@emmicro-us.com>
6351 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6352 * config/arc/arc.c (_arc_jli_section): New struct.
6353 (arc_jli_section): New type.
6354 (rc_jli_sections): New static variable.
6355 (arc_handle_jli_attribute): New function.
6356 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6357 (arc_file_end): New function.
6358 (TARGET_ASM_FILE_END): Define.
6359 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6360 (arc_add_jli_section): New function.
6361 (jli_call_scan): Likewise.
6362 (arc_reorg): Call jli_call_scan.
6363 (arc_output_addsi): Remove 'S' from printing asm operand.
6364 (arc_is_jli_call_p): New function.
6365 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6367 (movhi_insn): Likewise.
6368 (movsi_insn): Likewise.
6369 (movsi_set_cc_insn): Likewise.
6370 (loadqi_update): Likewise.
6371 (load_zeroextendqisi_update): Likewise.
6372 (load_signextendqisi_update): Likewise.
6373 (loadhi_update): Likewise.
6374 (load_zeroextendhisi_update): Likewise.
6375 (load_signextendhisi_update): Likewise.
6376 (loadsi_update): Likewise.
6377 (loadsf_update): Likewise.
6378 (movsicc_insn): Likewise.
6379 (bset_insn): Likewise.
6380 (bxor_insn): Likewise.
6381 (bclr_insn): Likewise.
6382 (bmsk_insn): Likewise.
6383 (bicsi3_insn): Likewise.
6384 (cmpsi_cc_c_insn): Likewise.
6385 (movsi_ne): Likewise.
6386 (movsi_cond_exec): Likewise.
6387 (clrsbsi2): Likewise.
6398 (call_i): Remove 'S' asm letter, add jli instruction.
6399 (call_value_i): Likewise.
6400 * config/arc/arc.op (mjli-always): New option.
6401 * config/arc/constraints.md (Cji): New constraint.
6402 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6404 (subsf3_fpx): Likewise.
6405 (mulsf3_fpx): Likewise.
6406 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6408 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6410 * doc/invoke.texi (ARC): Document mjli-always option.
6412 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6414 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6415 avoid addition with 0 and use incw and decw where possible.
6417 2018-01-26 Richard Biener <rguenther@suse.de>
6419 PR tree-optimization/81082
6420 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6421 association if it requires casting to unsigned.
6422 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6423 from fold_plusminus_mult_expr to catch important cases late when
6424 range info is available.
6426 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6428 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6429 * configure.ac (hidden_linkonce): New test.
6430 * configure: Regenerate.
6431 * config.in: Regenerate.
6433 2018-01-26 Julia Koval <julia.koval@intel.com>
6435 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6436 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6437 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6438 _mm_mask_bitshuffle_epi64_mask): Fix type.
6439 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6440 USI_FTYPE_V4DI_V4DI_USI): Remove.
6441 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6442 __builtin_ia32_vpshufbitqmb256_mask,
6443 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6444 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6445 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6447 2018-01-26 Alan Modra <amodra@gmail.com>
6450 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6451 UNSPEC_VBPERMQ. Sort other unspecs.
6453 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6455 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6457 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6460 * predict.c (drop_profile): Do not push/pop cfun; update also
6462 (handle_missing_profiles): Fix logic looking for zero profiles.
6464 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6467 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6468 on functions with #pragma omp declare simd or functions with simd
6470 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6471 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6472 Remove trailing \n from warning_at calls.
6474 2018-01-25 Tom de Vries <tom@codesourcery.com>
6477 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6478 for neutered workers.
6480 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6483 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6484 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6486 2018-01-24 Jeff Law <law@redhat.com>
6489 * i386.c (get_probe_interval): Move to earlier point.
6490 (ix86_compute_frame_layout): If -fstack-clash-protection and
6491 the frame is larger than the probe interval, then use pushes
6492 to save registers rather than reg->mem moves.
6493 (ix86_expand_prologue): Remove conditional for int_registers_saved
6496 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6499 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6500 min/max for never referenced object.
6502 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6505 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6507 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6508 attributes from DECL_ATTRIBUTES (decl) without affecting
6509 DECL_ATTRIBUTES (current_function_decl).
6510 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6511 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6513 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6515 PR tree-optimization/83979
6516 * fold-const.c (fold_comparison): Use constant_boolean_node
6517 instead of boolean_{true,false}_node.
6519 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6521 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6524 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6526 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6527 Simplify the clause that sets the length attribute.
6528 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6529 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6530 clause that sets the length attribute.
6531 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6533 2018-01-24 Tom de Vries <tom@codesourcery.com>
6536 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6537 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6538 Add strict parameter.
6539 (prevent_branch_around_nothing): Insert dummy insn between branch to
6540 label and label with no ptx insn inbetween.
6541 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6543 2018-01-24 Tom de Vries <tom@codesourcery.com>
6546 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6547 for neutered threads in warp.
6548 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6550 2018-01-24 Richard Biener <rguenther@suse.de>
6552 PR tree-optimization/83176
6553 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6556 2018-01-24 Richard Biener <rguenther@suse.de>
6558 PR tree-optimization/82819
6559 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6560 code generating pluses that are no-ops in the target precision.
6562 2018-01-24 Richard Biener <rguenther@suse.de>
6565 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6567 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6569 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6570 to merge probabilities.
6571 * predict.c (probably_never_executed): Also mark as cold functions
6572 with global 0 profile and guessed local profile.
6573 * profile-count.c (profile_probability::combine_with_count): New
6575 * profile-count.h (profile_probability::operator*,
6576 profile_probability::operator*=, profile_probability::operator/,
6577 profile_probability::operator/=): Reduce precision to adjusted
6578 and set value to guessed on contradictory divisions.
6579 (profile_probability::combine_with_freq): Remove.
6580 (profile_probability::combine_wiht_count): Declare.
6581 (profile_count::force_nonzero):: Set to adjusted.
6582 (profile_count::probability_in):: Set quality to adjusted.
6583 * tree-ssa-tail-merge.c (replace_block_by): Use
6586 2018-01-23 Andrew Waterman <andrew@sifive.com>
6587 Jim Wilson <jimw@sifive.com>
6589 * config/riscv/riscv.c (riscv_stack_boundary): New.
6590 (riscv_option_override): Set riscv_stack_boundary. Handle
6591 riscv_preferred_stack_boundary_arg.
6592 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6593 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6594 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6595 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6596 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6597 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6599 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6602 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6603 of struct ix86_frame.
6604 (ix86_expand_epilogue): Likewise. Add a local variable for
6605 the reg_save_offset field in struct ix86_frame.
6607 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6609 PR tree-optimization/82604
6610 * tree-loop-distribution.c (enum partition_kind): New enum item
6611 PKIND_PARTIAL_MEMSET.
6612 (partition_builtin_p): Support above new enum item.
6613 (generate_code_for_partition): Ditto.
6614 (compute_access_range): Differentiate cases that equality can be
6615 proven at all loops, the innermost loops or no loops.
6616 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6617 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6618 (finalize_partitions, distribute_loop): Don't fuse partition of
6619 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6620 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6623 2018-01-23 Martin Liska <mliska@suse.cz>
6625 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6626 order to ignore the predictor.
6627 (PRED_POLYMORPHIC_CALL): Likewise.
6628 (PRED_RECURSIVE_CALL): Likewise.
6630 2018-01-23 Martin Liska <mliska@suse.cz>
6632 * tree-profile.c (tree_profiling): Print function header to
6633 aware reader which function we are working on.
6634 * value-prof.c (gimple_find_values_to_profile): Do not print
6635 not interesting value histograms.
6637 2018-01-23 Martin Liska <mliska@suse.cz>
6639 * profile-count.h (enum profile_quality): Add
6640 profile_uninitialized as the first value. Do not number values
6641 as they are zero based.
6642 (profile_count::verify): Update sanity check.
6643 (profile_probability::verify): Likewise.
6645 2018-01-23 Nathan Sidwell <nathan@acm.org>
6647 * doc/invoke.texi (ffor-scope): Deprecate.
6649 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6651 PR tree-optimization/83510
6652 * domwalk.c (set_all_edges_as_executable): New function.
6653 (dom_walker::dom_walker): Convert bool param
6654 "skip_unreachable_blocks" to enum reachability. Move setup of
6655 edge flags to set_all_edges_as_executable and only do it when
6656 reachability is REACHABLE_BLOCKS.
6657 * domwalk.h (enum dom_walker::reachability): New enum.
6658 (dom_walker::dom_walker): Convert bool param
6659 "skip_unreachable_blocks" to enum reachability.
6660 (set_all_edges_as_executable): New decl.
6661 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6662 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6664 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6665 but converting true to REACHABLE_BLOCKS.
6666 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6668 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6669 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6670 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6672 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6673 if check_all_array_refs will be called.
6675 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6677 * tree.c (selftest::test_location_wrappers): Add more test
6680 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6682 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6683 (selftest::test_bit_in_range): Likewise.
6685 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6688 * doc/sourcebuild.texi (vect_float): Say that the selector
6689 only describes the situation when -funsafe-math-optimizations is on.
6690 (vect_float_strict): Document.
6692 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6694 PR tree-optimization/83965
6695 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6696 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6697 instead of checking only for a reduction.
6698 (vect_recog_widen_sum_pattern): Likewise.
6700 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6702 * predict.c (probably_never_executed): Only use precise profile info.
6703 (compute_function_frequency): Skip after inlining hack since we now
6704 have quality checking.
6706 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6708 * profile-count.h (profile_probability::very_unlikely,
6709 profile_probability::unlikely, profile_probability::even): Set
6710 precision to guessed.
6712 2018-01-23 Richard Biener <rguenther@suse.de>
6714 PR tree-optimization/83963
6715 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6716 Properly terminate dominator walk when crossing the exit edge not
6717 when visiting its source block.
6719 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6722 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6723 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6725 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6727 PR tree-optimization/83957
6728 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6729 semicolon after for body surrounded by braces.
6731 PR tree-optimization/83081
6732 * profile-count.h (profile_probability::split): New method.
6733 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6734 Use profile_probability::split.
6735 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6736 when splitting a single conditional jump into 2.
6738 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6740 PR tree-optimization/69452
6741 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6744 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6746 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6747 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6748 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6750 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6752 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6753 * config/rl78/rl78.md (movdi): New define_expand.
6754 * config/rl78/rl78.c (rl78_split_movdi): New function.
6756 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6759 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6761 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6762 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6763 128-bit to produce an UNSPEC move to get the double word with the
6764 signbit and then a shift directly to do signbit.
6765 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6766 implementation with a new version that just does either a direct
6767 move or a regular move. Move memory interface to separate insns.
6768 Move insns so they are next to the expander.
6769 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6770 with signbit move. Split big and little endian case.
6771 (signbit<mode>2_dm_mem_le): Likewise.
6772 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6773 (signbit<mode>2_dm2): Likewise.
6775 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6777 * config/rl78/rl78.md (anddi3): New define_expand.
6779 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6781 * config/rl78/rl78.md (umindi3): New define_expand.
6783 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6785 * config/rl78/rl78.md (smindi3): New define_expand.
6787 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6789 * config/rl78/rl78.md (smaxdi3): New define_expand.
6791 2018-01-22 Carl Love <cel@us.ibm.com>
6793 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6794 LVX_V1TI): Add macro expansion.
6795 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6796 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6797 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6798 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6799 Change check to determine if the instruction is a byte reversing
6800 entry. Fix typo in comment.
6801 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6802 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6803 Add def_builtin calls for new builtins.
6804 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6805 Add define_insn expansion.
6807 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6809 * config/rl78/rl78.md (umaxdi3): New define_expand.
6811 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6813 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6814 for non-QImode registers.
6816 2018-01-22 Richard Biener <rguenther@suse.de>
6818 PR tree-optimization/83963
6819 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6820 including the loop exit block.
6821 (scop_detection::merge_sese): Likewise.
6822 (scop_detection::add_scop): Do it here instead.
6824 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6826 * doc/sourcebuild.texi (arm_softfloat): Document.
6828 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6831 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6832 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6833 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6835 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6836 David Edelsohn <dje.gcc@gmail.com>
6839 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6840 Change "crset eq" to "crset 2".
6841 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6842 (*call_indirect_aix<mode>_nospec): Likewise.
6843 (*call_value_indirect_aix<mode>_nospec): Likewise.
6844 (*call_indirect_elfv2<mode>_nospec): Likewise.
6845 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6846 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6847 change assembly output from . to $.
6848 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6849 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6850 (*tablejump<mode>_internal1_nospec): Likewise.
6852 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6855 * config/sh/sh_optimize_sett_clrt.cc:
6856 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6858 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6860 PR tree-optimization/83940
6861 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6862 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6863 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6864 use it to pass back the definition type.
6865 (vect_check_store_rhs): Likewise rhs_dt_out.
6866 (vect_build_gather_load_calls): Add a mask_dt argument and use
6867 it instead of a call to vect_is_simple_use.
6868 (vectorizable_store): Update calls to vect_check_load_store_mask
6869 and vect_check_store_rhs. Use the dt returned by the latter instead
6870 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6871 instead of calls to vect_is_simple_use. Pass the scalar rather
6872 than the vector operand to vect_is_simple_use when handling
6873 second and subsequent copies of an rhs value.
6874 (vectorizable_load): Update calls to vect_check_load_store_mask
6875 and vect_build_gather_load_calls. Use the cached mask_dt and
6876 gs_info.offset_dt instead of calls to vect_is_simple_use.
6878 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6881 * tree-emutls.c: Include gimplify.h.
6882 (lower_emutls_2): New function.
6883 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6884 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6885 it before further processing.
6888 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6889 UINTVAL (trueop1) instead of INTVAL (op1).
6891 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6895 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6896 INCOMING_FRAME_SP_OFFSET if not defined.
6897 (scan_trace): Add ENTRY argument. If true and
6898 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6899 emit a note to adjust the CFA offset.
6900 (create_cfi_notes): Adjust scan_trace callers.
6901 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6902 INCOMING_FRAME_SP_OFFSET in the CIE.
6903 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6904 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6906 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6907 * doc/tm.texi: Regenerated.
6909 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6911 PR rtl-optimization/83147
6912 * lra-constraints.c (remove_inheritance_pseudos): Use
6913 lra_substitute_pseudo_within_insn.
6915 2018-01-19 Tom de Vries <tom@codesourcery.com>
6916 Cesar Philippidis <cesar@codesourcery.com>
6919 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6921 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6924 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6925 spaces for function labels.
6927 2018-01-19 Martin Liska <mliska@suse.cz>
6929 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6930 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6931 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6932 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6933 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6934 (PRED_CONST_RETURN): Change from 69 to 65.
6935 (PRED_NULL_RETURN): Change from 91 to 71.
6936 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6937 (PRED_LOOP_GUARD): Change from 66 to 73.
6939 2018-01-19 Martin Liska <mliska@suse.cz>
6941 * predict.c (predict_insn_def): Add new assert.
6942 (struct branch_predictor): Change type to signed integer.
6943 (test_prediction_value_range): Amend test to cover
6945 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6946 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6947 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6948 (PRED_LOOP_IV_COMPARE): Likewise.
6949 * predict.h (PROB_UNINITIALIZED): Define new constant.
6951 2018-01-19 Martin Liska <mliska@suse.cz>
6953 * predict.c (dump_prediction): Add new format for
6954 analyze_brprob.py script which is enabled with -details
6956 * profile-count.h (precise_p): New function.
6958 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6960 PR tree-optimization/83922
6961 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6962 there are no statements that need masking.
6963 (vect_active_double_reduction_p): New function.
6964 (vect_analyze_loop_operations): Use it when handling phis that
6965 are not in the loop header.
6967 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6969 PR tree-optimization/83914
6970 * tree-vect-loop.c (vectorizable_induction): Don't convert
6971 init_expr or apply the peeling adjustment for inductions
6972 that are nested within the vectorized loop.
6974 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6976 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6979 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6983 * function.h (gimplify_parameters): Add gimple_seq * argument.
6984 * function.c: Include gimple.h and options.h.
6985 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6986 for the added local temporaries if needed.
6987 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6988 if there are any parameter cleanups, wrap whole body into a
6989 try/finally with the cleanups.
6991 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6994 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6995 Use GET_MODE_CLASS for scalar floating point.
6997 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7001 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7002 Fix call of call_cgraph_insertion_hooks.
7004 2018-01-18 Martin Sebor <msebor@redhat.com>
7006 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7008 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7011 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7014 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7017 * common.opt: (-ffile-prefix-map): New option.
7018 * opts.c (common_handle_option): Defer it.
7019 * opts-global.c (handle_common_deferred_options): Handle it.
7020 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7021 * file-prefix-map.h: New file.
7022 (remap_debug_filename, add_debug_prefix_map): ...here.
7023 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7024 * final.c (debug_prefix_map, add_debug_prefix_map
7025 remap_debug_filename): Move to...
7026 * file-prefix-map.c: New file.
7027 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7028 generalize, get rid of alloca(), use strrchr() instead of strchr().
7029 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7030 Implement in terms of add_prefix_map().
7031 (remap_macro_filename, remap_debug_filename): Implement in term of
7033 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7034 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7035 * dbxout.c: Include file-prefix-map.h.
7036 * varasm.c: Likewise.
7037 * vmsdbgout.c: Likewise.
7038 * xcoffout.c: Likewise.
7039 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7040 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7041 * doc/invoke.texi (-ffile-prefix-map): Document.
7042 (-fdebug-prefix-map): Update description.
7044 2018-01-18 Martin Liska <mliska@suse.cz>
7046 * config/i386/i386.c (indirect_thunk_name): Document that also
7048 (output_indirect_thunk): Document why both instructions
7049 (pause and lfence) are generated.
7051 2018-01-18 Richard Biener <rguenther@suse.de>
7053 PR tree-optimization/83887
7054 * graphite-scop-detection.c
7055 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7056 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7057 (scop_detection::merge_sese): Re-implement with a flood-fill
7058 algorithm that properly finds a SESE region if it exists.
7060 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7063 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7064 pointer_diff optimizations use view_convert instead of convert.
7066 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7068 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7069 Generate different code for -mno-speculate-indirect-jumps.
7070 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7071 (*call_indirect_aix<mode>): Disable for
7072 -mno-speculate-indirect-jumps.
7073 (*call_indirect_aix<mode>_nospec): New define_insn.
7074 (*call_value_indirect_aix<mode>): Disable for
7075 -mno-speculate-indirect-jumps.
7076 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7077 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7078 -mno-speculate-indirect-jumps.
7079 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7081 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7083 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7084 long double type, set the flags for noting the default long double
7085 type, even if we don't pass or return a long double type.
7087 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7090 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7093 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7095 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7096 support for merge[hl].
7097 (fold_mergehl_helper): New helper function.
7098 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7099 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7100 (altivec_vmrglw_direct): Add xxmrglw insn.
7102 2018-01-17 Andrew Waterman <andrew@sifive.com>
7104 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7105 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7107 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7110 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7111 call the lto_location_cache before reading the
7112 DECL_SOURCE_LOCATION of the types.
7114 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7115 Richard Sandiford <richard.sandiford@linaro.org>
7117 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7118 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7119 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7120 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7121 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7123 * config/aarch64/constraints.md (aarch64_movti_operand):
7125 * config/aarch64/predicates.md (Uti): Add new constraint.
7127 2018-01-17 Carl Love <cel@us.ibm.com>
7129 * config/rs6000/vsx.md (define_expand xl_len_r,
7130 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7131 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7133 (define_expand, define_insn): Move the shift left from the
7134 define_insn to the define_expand for lxvl and stxvl instructions.
7135 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7136 and XL_LEN_R definitions to PURE.
7138 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7140 * config/i386/i386.c (indirect_thunk_name): Declare regno
7141 as unsigned int. Compare regno with INVALID_REGNUM.
7142 (output_indirect_thunk): Ditto.
7143 (output_indirect_thunk_function): Ditto.
7144 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7145 in the call to output_indirect_thunk_function.
7147 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7150 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7151 rather than the size of inner_type to determine the stack slot size
7152 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7154 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7157 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7160 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7162 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7163 endian Linux systems to optionally enable multilibs for selecting
7164 the long double type if the user configured an explicit type.
7165 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7166 have no long double multilibs if not defined.
7167 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7168 warn if the user used -mabi={ieee,ibm}longdouble and we built
7169 multilibs for long double.
7170 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7171 appropriate multilib option.
7172 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7174 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7175 for building long double multilibs.
7176 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7178 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7180 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7183 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7185 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7188 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7191 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7194 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7196 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7197 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7199 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7201 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7202 different rtl trees depending on TARGET_64BIT.
7203 (rs6000_gen_lvx): Likewise.
7205 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7207 * config/visium/visium.md (nop): Tweak comment.
7208 (hazard_nop): Likewise.
7210 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7212 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7213 -mspeculate-indirect-jumps.
7214 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7215 for -mno-speculate-indirect-jumps.
7216 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7217 (*call_value_indirect_elfv2<mode>): Disable for
7218 -mno-speculate-indirect-jumps.
7219 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7220 (indirect_jump): Emit different RTL for
7221 -mno-speculate-indirect-jumps.
7222 (*indirect_jump<mode>): Disable for
7223 -mno-speculate-indirect-jumps.
7224 (*indirect_jump<mode>_nospec): New define_insn.
7225 (tablejump): Emit different RTL for
7226 -mno-speculate-indirect-jumps.
7227 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7228 (tablejumpsi_nospec): New define_expand.
7229 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7230 (tablejumpdi_nospec): New define_expand.
7231 (*tablejump<mode>_internal1): Disable for
7232 -mno-speculate-indirect-jumps.
7233 (*tablejump<mode>_internal1_nospec): New define_insn.
7234 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7237 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7239 * caller-save.c (insert_save): Drop unnecessary parameter. All
7242 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7243 Richard Biener <rguenth@suse.de>
7246 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7247 return early, inline manually is_gimple_sizepos. Make sure if we
7248 call gimplify_expr we don't end up with a gimple constant.
7249 * tree.c (variably_modified_type_p): Don't return true for
7250 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7251 * gimplify.h (is_gimple_sizepos): Remove.
7253 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7255 PR tree-optimization/83857
7256 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7257 vectorizable_live_operation for pure SLP statements.
7258 (vectorizable_live_operation): Handle PHIs.
7260 2018-01-16 Richard Biener <rguenther@suse.de>
7262 PR tree-optimization/83867
7263 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7264 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7266 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7269 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7270 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7271 If off is not INTEGER_CST, issue a may not be aligned warning
7272 rather than isn't aligned. Use isn%'t rather than isn't.
7273 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7275 <case MULT_EXPR>: Improve the case when bottom and one of the
7276 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7277 operand, in that case check if the other operand is multiple of
7278 bottom divided by the INTEGER_CST operand.
7280 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7283 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7284 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7285 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7286 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7287 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7288 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7289 (pa_function_arg_size): New function.
7291 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7293 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7294 in a separate statement.
7296 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7298 PR tree-optimization/83847
7299 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7300 group gathers and scatters.
7302 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7304 PR rtl-optimization/86620
7305 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7307 PR rtl-optimization/83213
7308 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7309 to last if both are JUMP_INSNs.
7311 PR tree-optimization/83843
7312 * gimple-ssa-store-merging.c
7313 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7314 store_immediate_info for bswap/nop orig_stores.
7316 2018-01-15 Andrew Waterman <andrew@sifive.com>
7318 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7320 <UDIV>: Increase cost if !TARGET_DIV.
7322 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7324 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7325 (define_attr "cr_logical_3op"): New.
7326 (cceq_ior_compare): Adjust.
7327 (cceq_ior_compare_complement): Adjust.
7328 (*cceq_rev_compare): Adjust.
7329 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7330 (is_cracked_insn): Adjust.
7331 (insn_must_be_first_in_group): Adjust.
7332 * config/rs6000/40x.md: Adjust.
7333 * config/rs6000/440.md: Adjust.
7334 * config/rs6000/476.md: Adjust.
7335 * config/rs6000/601.md: Adjust.
7336 * config/rs6000/603.md: Adjust.
7337 * config/rs6000/6xx.md: Adjust.
7338 * config/rs6000/7450.md: Adjust.
7339 * config/rs6000/7xx.md: Adjust.
7340 * config/rs6000/8540.md: Adjust.
7341 * config/rs6000/cell.md: Adjust.
7342 * config/rs6000/e300c2c3.md: Adjust.
7343 * config/rs6000/e500mc.md: Adjust.
7344 * config/rs6000/e500mc64.md: Adjust.
7345 * config/rs6000/e5500.md: Adjust.
7346 * config/rs6000/e6500.md: Adjust.
7347 * config/rs6000/mpc.md: Adjust.
7348 * config/rs6000/power4.md: Adjust.
7349 * config/rs6000/power5.md: Adjust.
7350 * config/rs6000/power6.md: Adjust.
7351 * config/rs6000/power7.md: Adjust.
7352 * config/rs6000/power8.md: Adjust.
7353 * config/rs6000/power9.md: Adjust.
7354 * config/rs6000/rs64.md: Adjust.
7355 * config/rs6000/titan.md: Adjust.
7357 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7359 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7360 ix86_indirect_branch_register logic.
7362 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7364 * config/i386/constraints.md (Bs): Update
7365 ix86_indirect_branch_register check. Don't check
7366 ix86_indirect_branch_register with GOT_memory_operand.
7368 * config/i386/predicates.md (GOT_memory_operand): Don't check
7369 ix86_indirect_branch_register here.
7370 (GOT32_symbol_operand): Likewise.
7372 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7374 * config/i386/predicates.md (constant_call_address_operand):
7375 Rewrite ix86_indirect_branch_register logic.
7376 (sibcall_insn_operand): Likewise.
7378 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7380 * config/i386/constraints.md (Bs): Replace
7381 ix86_indirect_branch_thunk_register with
7382 ix86_indirect_branch_register.
7384 * config/i386/i386.md (indirect_jump): Likewise.
7385 (tablejump): Likewise.
7386 (*sibcall_memory): Likewise.
7387 (*sibcall_value_memory): Likewise.
7388 Peepholes of indirect call and jump via memory: Likewise.
7389 * config/i386/i386.opt: Likewise.
7390 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7391 (GOT_memory_operand): Likewise.
7392 (call_insn_operand): Likewise.
7393 (sibcall_insn_operand): Likewise.
7394 (GOT32_symbol_operand): Likewise.
7396 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7399 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7400 type rather than type addr's type points to.
7401 (expand_omp_atomic_mutex): Likewise.
7402 (expand_omp_atomic): Likewise.
7404 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7407 * config/i386/i386.c (output_indirect_thunk_function): Use
7408 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7409 for __x86_return_thunk.
7411 2018-01-15 Richard Biener <rguenther@suse.de>
7414 * expmed.c (extract_bit_field_1): Fix typo.
7416 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7419 * config/arm/iterators.md (VF): New mode iterator.
7420 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7421 Remove integer-related logic from pattern.
7422 (neon_vabd<mode>_3): Likewise.
7424 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7427 * common.opt (fstrict-overflow): No longer an alias.
7428 (fwrapv-pointer): New option.
7429 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7430 also for pointer types based on flag_wrapv_pointer.
7431 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7432 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7433 opts->x_flag_wrapv got set.
7434 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7435 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7436 POINTER_TYPE_OVERFLOW_UNDEFINED.
7437 * match.pd: Likewise in address comparison pattern.
7438 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7440 2018-01-15 Richard Biener <rguenther@suse.de>
7443 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7444 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7445 Reset type names to their identifier if their TYPE_DECL doesn't
7446 have linkage (and thus is used for ODR and devirt).
7447 (save_debug_info_for_decl): Remove.
7448 (save_debug_info_for_type): Likewise.
7449 (add_tree_to_fld_list): Adjust.
7450 * tree-pretty-print.c (dump_generic_node): Make dumping of
7451 type names more robust.
7453 2018-01-15 Richard Biener <rguenther@suse.de>
7455 * BASE-VER: Bump to 8.0.1.
7457 2018-01-14 Martin Sebor <msebor@redhat.com>
7460 * builtins.c (check_access): Avoid warning when the no-warning bit
7463 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7465 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7466 * ira-color (allocno_hard_regs_compare): Likewise.
7468 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7471 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7472 Use .pushsection/.popsection.
7474 2018-01-14 Martin Sebor <msebor@redhat.com>
7477 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7479 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7481 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7482 entry from extra_headers.
7483 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7484 extra_headers, make the list bitwise identical to the i?86-*-* one.
7486 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7488 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7489 -mcmodel=large with -mindirect-branch=thunk,
7490 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7491 -mfunction-return=thunk-extern.
7492 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7493 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7494 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7496 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7498 * config/i386/i386.c (print_reg): Print the name of the full
7499 integer register without '%'.
7500 (ix86_print_operand): Handle 'V'.
7501 * doc/extend.texi: Document 'V' modifier.
7503 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7505 * config/i386/constraints.md (Bs): Disallow memory operand for
7506 -mindirect-branch-register.
7508 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7509 (GOT_memory_operand): Likewise.
7510 (call_insn_operand): Likewise.
7511 (sibcall_insn_operand): Likewise.
7512 (GOT32_symbol_operand): Likewise.
7513 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7514 for -mindirect-branch-register.
7515 (tablejump): Likewise.
7516 (*sibcall_memory): Likewise.
7517 (*sibcall_value_memory): Likewise.
7518 Disallow peepholes of indirect call and jump via memory for
7519 -mindirect-branch-register.
7520 (*call_pop): Replace m with Bw.
7521 (*call_value_pop): Likewise.
7522 (*sibcall_pop_memory): Replace m with Bs.
7523 * config/i386/i386.opt (mindirect-branch-register): New option.
7524 * doc/invoke.texi: Document -mindirect-branch-register option.
7526 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7528 * config/i386/i386-protos.h (ix86_output_function_return): New.
7529 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7530 set function_return_type.
7531 (indirect_thunk_name): Add ret_p to indicate thunk for function
7533 (output_indirect_thunk_function): Pass false to
7534 indirect_thunk_name.
7535 (ix86_output_indirect_branch_via_reg): Likewise.
7536 (ix86_output_indirect_branch_via_push): Likewise.
7537 (output_indirect_thunk_function): Create alias for function
7538 return thunk if regno < 0.
7539 (ix86_output_function_return): New function.
7540 (ix86_handle_fndecl_attribute): Handle function_return.
7541 (ix86_attribute_table): Add function_return.
7542 * config/i386/i386.h (machine_function): Add
7543 function_return_type.
7544 * config/i386/i386.md (simple_return_internal): Use
7545 ix86_output_function_return.
7546 (simple_return_internal_long): Likewise.
7547 * config/i386/i386.opt (mfunction-return=): New option.
7548 (indirect_branch): Mention -mfunction-return=.
7549 * doc/extend.texi: Document function_return function attribute.
7550 * doc/invoke.texi: Document -mfunction-return= option.
7552 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7554 * config/i386/i386-opts.h (indirect_branch): New.
7555 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7556 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7557 with local indirect jump when converting indirect call and jump.
7558 (ix86_set_indirect_branch_type): New.
7559 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7560 (indirectlabelno): New.
7561 (indirect_thunk_needed): Likewise.
7562 (indirect_thunk_bnd_needed): Likewise.
7563 (indirect_thunks_used): Likewise.
7564 (indirect_thunks_bnd_used): Likewise.
7565 (INDIRECT_LABEL): Likewise.
7566 (indirect_thunk_name): Likewise.
7567 (output_indirect_thunk): Likewise.
7568 (output_indirect_thunk_function): Likewise.
7569 (ix86_output_indirect_branch_via_reg): Likewise.
7570 (ix86_output_indirect_branch_via_push): Likewise.
7571 (ix86_output_indirect_branch): Likewise.
7572 (ix86_output_indirect_jmp): Likewise.
7573 (ix86_code_end): Call output_indirect_thunk_function if needed.
7574 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7576 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7577 (ix86_attribute_table): Add indirect_branch.
7578 * config/i386/i386.h (machine_function): Add indirect_branch_type
7579 and has_local_indirect_jump.
7580 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7582 (tablejump): Likewise.
7583 (*indirect_jump): Use ix86_output_indirect_jmp.
7584 (*tablejump_1): Likewise.
7585 (simple_return_indirect_internal): Likewise.
7586 * config/i386/i386.opt (mindirect-branch=): New option.
7587 (indirect_branch): New.
7590 (thunk-inline): Likewise.
7591 (thunk-extern): Likewise.
7592 * doc/extend.texi: Document indirect_branch function attribute.
7593 * doc/invoke.texi: Document -mindirect-branch= option.
7595 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7598 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7600 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7602 * ipa-inline.c (want_inline_small_function_p): Return false if
7603 inlining has already failed with CIF_FINAL_ERROR.
7604 (update_caller_keys): Call want_inline_small_function_p before
7606 (update_callee_keys): Likewise.
7608 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7610 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7612 (rs6000_quadword_masked_address_p): Likewise.
7613 (quad_aligned_load_p): Likewise.
7614 (quad_aligned_store_p): Likewise.
7615 (const_load_sequence_p): Add comment to describe the outer-most loop.
7616 (mimic_memory_attributes_and_flags): New function.
7617 (rs6000_gen_stvx): Likewise.
7618 (replace_swapped_aligned_store): Likewise.
7619 (rs6000_gen_lvx): Likewise.
7620 (replace_swapped_aligned_load): Likewise.
7621 (replace_swapped_load_constant): Capitalize argument name in
7622 comment describing this function.
7623 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7624 and stores that access quad-word aligned addresses and replace
7625 with stvx or lvx instructions when appropriate.
7626 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7627 New function prototype.
7628 (rs6000_quadword_masked_address_p): Likewise.
7629 (rs6000_gen_lvx): Likewise.
7630 (rs6000_gen_stvx): Likewise.
7631 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7632 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7633 when memory address is aligned.
7634 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7635 this split to select lvx instruction when memory address is aligned.
7636 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7637 instruction when memory address is aligned.
7638 (*vsx_le_perm_load_v16qi): Likewise.
7639 (four unnamed splitters): Modify to select the stvx instruction
7640 when memory is aligned.
7642 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7644 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7645 which appears in the queue multiple times.
7647 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7648 Alan Hayward <alan.hayward@arm.com>
7649 David Sherwood <david.sherwood@arm.com>
7651 * tree-vectorizer.h (vec_lower_bound): New structure.
7652 (_loop_vec_info): Add check_nonzero and lower_bounds.
7653 (LOOP_VINFO_CHECK_NONZERO): New macro.
7654 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7655 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7656 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7657 fields. Make seg_len the distance travelled, not including the
7659 (dr_direction_indicator): Declare.
7660 (dr_zero_step_indicator): Likewise.
7661 (dr_known_forward_stride_p): Likewise.
7662 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7664 (runtime_alias_check_p): Allow runtime alias checks with
7666 (operator ==): Compare access_size and align.
7667 (prune_runtime_alias_test_list): Rework for new distinction between
7668 the access_size and seg_len.
7669 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7671 (get_segment_min_max): New function.
7672 (create_intersect_range_checks): Use it.
7673 (dr_step_indicator): New function.
7674 (dr_direction_indicator): Likewise.
7675 (dr_zero_step_indicator): Likewise.
7676 (dr_known_forward_stride_p): Likewise.
7677 * tree-loop-distribution.c (data_ref_segment_size): Return
7678 DR_STEP * (niters - 1).
7679 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7681 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7682 (vect_preserves_scalar_order_p): New function, split out from...
7683 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7684 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7685 (vect_vfa_access_size): New function.
7686 (vect_vfa_align): Likewise.
7687 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7688 (dump_lower_bound): New function.
7689 (vect_check_lower_bound): Likewise.
7690 (vect_small_gap_p): Likewise.
7691 (vectorizable_with_step_bound_p): Likewise.
7692 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7693 depencies if the vectorization factor is 1. Convert the checks
7694 for nonzero steps into checks on the bounds of DR_STEP. Try using
7695 a bunds check for variable steps if the minimum required step is
7696 relatively small. Update calls to the dr_with_seg_len
7697 constructor and to vect_compile_time_alias.
7698 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7700 (vect_loop_versioning): Call it.
7701 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7703 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7705 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7706 Alan Hayward <alan.hayward@arm.com>
7707 David Sherwood <david.sherwood@arm.com>
7709 * doc/sourcebuild.texi (vect_scatter_store): Document.
7710 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7712 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7714 * genopinit.c (main): Add supports_vec_scatter_store and
7715 supports_vec_scatter_store_cached to target_optabs.
7716 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7717 IFN_MASK_SCATTER_STORE.
7718 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7720 * internal-fn.h (internal_store_fn_p): Declare.
7721 (internal_fn_stored_value_index): Likewise.
7722 * internal-fn.c (scatter_store_direct): New macro.
7723 (expand_scatter_store_optab_fn): New function.
7724 (direct_scatter_store_optab_supported_p): New macro.
7725 (internal_store_fn_p): New function.
7726 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7727 IFN_MASK_SCATTER_STORE.
7728 (internal_fn_mask_index): Likewise.
7729 (internal_fn_stored_value_index): New function.
7730 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7732 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7733 * optabs-query.c (supports_vec_scatter_store_p): New function.
7734 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7735 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7736 true for scatter stores.
7737 (vect_gather_scatter_fn_p): Handle scatter stores too.
7738 (vect_check_gather_scatter): Consider using scatter stores if
7739 supports_vec_scatter_store_p.
7740 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7742 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7743 internal_fn_stored_value_index.
7744 (check_load_store_masking): Handle scatter stores too.
7745 (vect_get_store_rhs): Make public.
7746 (vectorizable_call): Use internal_store_fn_p.
7747 (vectorizable_store): Handle scatter store internal functions.
7748 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7749 when deciding whether the end of the group has been reached.
7750 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7751 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7752 (mask_scatter_store<mode>): New insns.
7754 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7755 Alan Hayward <alan.hayward@arm.com>
7756 David Sherwood <david.sherwood@arm.com>
7758 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7759 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7760 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7762 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7763 Use vect_truncate_gather_scatter_offset if we can't treat the
7764 operation as a normal gather load or scatter store.
7765 (get_group_load_store_type): Take the gather_scatter_info
7766 as argument. Try using a gather load or scatter store for
7767 single-element groups.
7768 (get_load_store_type): Update calls to get_group_load_store_type
7769 and vect_use_strided_gather_scatters_p.
7771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7772 Alan Hayward <alan.hayward@arm.com>
7773 David Sherwood <david.sherwood@arm.com>
7775 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7776 optional tree argument.
7777 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7779 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7780 but continue to use the current value as a fallback.
7781 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7782 to compare the updates.
7783 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7784 (get_load_store_type): Use it when handling a strided access.
7785 (vect_get_strided_load_store_ops): New function.
7786 (vect_get_data_ptr_increment): Likewise.
7787 (vectorizable_load): Handle strided gather loads. Always pass
7788 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7790 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7791 Alan Hayward <alan.hayward@arm.com>
7792 David Sherwood <david.sherwood@arm.com>
7794 * doc/md.texi (gather_load@var{m}): Document.
7795 (mask_gather_load@var{m}): Likewise.
7796 * genopinit.c (main): Add supports_vec_gather_load and
7797 supports_vec_gather_load_cached to target_optabs.
7798 * optabs-tree.c (init_tree_optimization_optabs): Use
7799 ggc_cleared_alloc to allocate target_optabs.
7800 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7801 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7803 * internal-fn.h (internal_load_fn_p): Declare.
7804 (internal_gather_scatter_fn_p): Likewise.
7805 (internal_fn_mask_index): Likewise.
7806 (internal_gather_scatter_fn_supported_p): Likewise.
7807 * internal-fn.c (gather_load_direct): New macro.
7808 (expand_gather_load_optab_fn): New function.
7809 (direct_gather_load_optab_supported_p): New macro.
7810 (direct_internal_fn_optab): New function.
7811 (internal_load_fn_p): Likewise.
7812 (internal_gather_scatter_fn_p): Likewise.
7813 (internal_fn_mask_index): Likewise.
7814 (internal_gather_scatter_fn_supported_p): Likewise.
7815 * optabs-query.c (supports_at_least_one_mode_p): New function.
7816 (supports_vec_gather_load_p): Likewise.
7817 * optabs-query.h (supports_vec_gather_load_p): Declare.
7818 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7819 and memory_type field.
7820 (NUM_PATTERNS): Bump to 15.
7821 * tree-vect-data-refs.c: Include internal-fn.h.
7822 (vect_gather_scatter_fn_p): New function.
7823 (vect_describe_gather_scatter_call): Likewise.
7824 (vect_check_gather_scatter): Try using internal functions for
7825 gather loads. Recognize existing calls to a gather load function.
7826 (vect_analyze_data_refs): Consider using gather loads if
7827 supports_vec_gather_load_p.
7828 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7829 (vect_get_gather_scatter_offset_type): Likewise.
7830 (vect_convert_mask_for_vectype): Likewise.
7831 (vect_add_conversion_to_patterm): Likewise.
7832 (vect_try_gather_scatter_pattern): Likewise.
7833 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7834 (vect_vect_recog_func_ptrs): Add it.
7835 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7836 internal_fn_mask_index and internal_gather_scatter_fn_p.
7837 (check_load_store_masking): Take the gather_scatter_info as an
7838 argument and handle gather loads.
7839 (vect_get_gather_scatter_ops): New function.
7840 (vectorizable_call): Check internal_load_fn_p.
7841 (vectorizable_load): Likewise. Handle gather load internal
7843 (vectorizable_store): Update call to check_load_store_masking.
7844 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7845 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7846 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7847 (aarch64_gather_scale_operand_d): New predicates.
7848 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7849 (mask_gather_load<mode>): New insns.
7851 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7852 Alan Hayward <alan.hayward@arm.com>
7853 David Sherwood <david.sherwood@arm.com>
7855 * optabs.def (fold_left_plus_optab): New optab.
7856 * doc/md.texi (fold_left_plus_@var{m}): Document.
7857 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7858 * internal-fn.c (fold_left_direct): Define.
7859 (expand_fold_left_optab_fn): Likewise.
7860 (direct_fold_left_optab_supported_p): Likewise.
7861 * fold-const-call.c (fold_const_fold_left): New function.
7862 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7863 * tree-parloops.c (valid_reduction_p): New function.
7864 (gather_scalar_reductions): Use it.
7865 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7866 (vect_finish_replace_stmt): Declare.
7867 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7868 (needs_fold_left_reduction_p): New function, split out from...
7869 (vect_is_simple_reduction): ...here. Accept reductions that
7870 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7871 (vect_force_simple_reduction): Also store the reduction type in
7872 the assignment's STMT_VINFO_REDUC_TYPE.
7873 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7874 (merge_with_identity): New function.
7875 (vect_expand_fold_left): Likewise.
7876 (vectorize_fold_left_reduction): Likewise.
7877 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7878 scalar phi in place for it. Check for target support and reject
7879 cases that would reassociate the operation. Defer the transform
7880 phase to vectorize_fold_left_reduction.
7881 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7882 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7883 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7885 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7887 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7888 call to ifc_temp_var.
7890 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7891 Alan Hayward <alan.hayward@arm.com>
7892 David Sherwood <david.sherwood@arm.com>
7894 * target.def (legitimize_address_displacement): Take the original
7895 offset as a poly_int.
7896 * targhooks.h (default_legitimize_address_displacement): Update
7898 * targhooks.c (default_legitimize_address_displacement): Likewise.
7899 * doc/tm.texi: Regenerate.
7900 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7901 as an argument, moving assert of ad->disp == ad->disp_term to...
7902 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7903 Try calling targetm.legitimize_address_displacement before expanding
7904 the address rather than afterwards, and adjust for the new interface.
7905 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7906 Match the new hook interface. Handle SVE addresses.
7907 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7910 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7912 * Makefile.in (OBJS): Add early-remat.o.
7913 * target.def (select_early_remat_modes): New hook.
7914 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7915 * doc/tm.texi: Regenerate.
7916 * targhooks.h (default_select_early_remat_modes): Declare.
7917 * targhooks.c (default_select_early_remat_modes): New function.
7918 * timevar.def (TV_EARLY_REMAT): New timevar.
7919 * passes.def (pass_early_remat): New pass.
7920 * tree-pass.h (make_pass_early_remat): Declare.
7921 * early-remat.c: New file.
7922 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7924 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7926 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7927 Alan Hayward <alan.hayward@arm.com>
7928 David Sherwood <david.sherwood@arm.com>
7930 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7931 vfm1 with a bound_epilog parameter.
7932 (vect_do_peeling): Update calls accordingly, and move the prologue
7933 call earlier in the function. Treat the base bound_epilog as 0 for
7934 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7935 this base when peeling for gaps.
7936 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7937 with fully-masked loops.
7938 (vect_estimate_min_profitable_iters): Handle the single peeled
7939 iteration in that case.
7941 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7942 Alan Hayward <alan.hayward@arm.com>
7943 David Sherwood <david.sherwood@arm.com>
7945 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7946 single-element interleaving even if the size is not a power of 2.
7947 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7948 accesses for single-element interleaving if the group size is
7951 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7952 Alan Hayward <alan.hayward@arm.com>
7953 David Sherwood <david.sherwood@arm.com>
7955 * doc/md.texi (fold_extract_last_@var{m}): Document.
7956 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7957 * optabs.def (fold_extract_last_optab): New optab.
7958 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7959 * internal-fn.c (fold_extract_direct): New macro.
7960 (expand_fold_extract_optab_fn): Likewise.
7961 (direct_fold_extract_optab_supported_p): Likewise.
7962 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7963 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7964 EXTRACT_LAST_REDUCTION.
7965 (get_initial_def_for_reduction): Do not create an initial vector
7966 for EXTRACT_LAST_REDUCTION reductions.
7967 (vectorizable_reduction): Leave the scalar phi in place for
7968 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7969 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7970 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7971 transform phase to vectorizable_condition.
7972 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7974 (vect_finish_stmt_generation): ...here.
7975 (vect_finish_replace_stmt): New function.
7976 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7977 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7979 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7981 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7982 Alan Hayward <alan.hayward@arm.com>
7983 David Sherwood <david.sherwood@arm.com>
7985 * doc/md.texi (extract_last_@var{m}): Document.
7986 * optabs.def (extract_last_optab): New optab.
7987 * internal-fn.def (EXTRACT_LAST): New internal function.
7988 * internal-fn.c (cond_unary_direct): New macro.
7989 (expand_cond_unary_optab_fn): Likewise.
7990 (direct_cond_unary_optab_supported_p): Likewise.
7991 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7992 loops using EXTRACT_LAST.
7993 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7994 (extract_last_<mode>): ...this optab.
7995 (vec_extract<mode><Vel>): Update accordingly.
7997 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7998 Alan Hayward <alan.hayward@arm.com>
7999 David Sherwood <david.sherwood@arm.com>
8001 * target.def (empty_mask_is_expensive): New hook.
8002 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8003 * doc/tm.texi: Regenerate.
8004 * targhooks.h (default_empty_mask_is_expensive): Declare.
8005 * targhooks.c (default_empty_mask_is_expensive): New function.
8006 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8007 if the target says that empty masks are expensive.
8008 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8010 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8012 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8013 Alan Hayward <alan.hayward@arm.com>
8014 David Sherwood <david.sherwood@arm.com>
8016 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8017 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8018 (vect_use_loop_mask_for_alignment_p): New function.
8019 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8020 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8021 niters_skip argument. Make sure that the first niters_skip elements
8022 of the first iteration are inactive.
8023 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8024 Update call to vect_set_loop_masks_directly.
8025 (get_misalign_in_elems): New function, split out from...
8026 (vect_gen_prolog_loop_niters): ...here.
8027 (vect_update_init_of_dr): Take a code argument that specifies whether
8028 the adjustment should be added or subtracted.
8029 (vect_update_init_of_drs): Likewise.
8030 (vect_prepare_for_masked_peels): New function.
8031 (vect_do_peeling): Skip prologue peeling if we're using a mask
8032 instead. Update call to vect_update_inits_of_drs.
8033 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8035 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8036 alignment. Do not include the number of peeled iterations in
8037 the minimum threshold in that case.
8038 (vectorizable_induction): Adjust the start value down by
8039 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8040 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8041 Take the number of skipped iterations into account when calculating
8043 * tree-vect-stmts.c (vect_gen_while_not): New function.
8045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8046 Alan Hayward <alan.hayward@arm.com>
8047 David Sherwood <david.sherwood@arm.com>
8049 * doc/sourcebuild.texi (vect_fully_masked): Document.
8050 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8052 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8054 (vect_analyze_loop_2): ...here. Don't check the vectorization
8055 factor against the number of loop iterations if the loop is
8058 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8059 Alan Hayward <alan.hayward@arm.com>
8060 David Sherwood <david.sherwood@arm.com>
8062 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8063 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8064 (dump_groups): Update accordingly.
8065 (iv_use::mem_type): New member variable.
8066 (address_p): New function.
8067 (record_use): Add a mem_type argument and initialize the new
8069 (record_group_use): Add a mem_type argument. Use address_p.
8070 Remove obsolete null checks of base_object. Update call to record_use.
8071 (find_interesting_uses_op): Update call to record_group_use.
8072 (find_interesting_uses_cond): Likewise.
8073 (find_interesting_uses_address): Likewise.
8074 (get_mem_type_for_internal_fn): New function.
8075 (find_address_like_use): Likewise.
8076 (find_interesting_uses_stmt): Try find_address_like_use before
8077 calling find_interesting_uses_op.
8078 (addr_offset_valid_p): Use the iv mem_type field as the type
8079 of the addressed memory.
8080 (add_autoinc_candidates): Likewise.
8081 (get_address_cost): Likewise.
8082 (split_small_address_groups_p): Use address_p.
8083 (split_address_groups): Likewise.
8084 (add_iv_candidate_for_use): Likewise.
8085 (autoinc_possible_for_pair): Likewise.
8086 (rewrite_groups): Likewise.
8087 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8088 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8089 (get_alias_ptr_type_for_ptr_address): New function.
8090 (rewrite_use_address): Rewrite address uses in calls that were
8091 identified by find_address_like_use.
8093 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8094 Alan Hayward <alan.hayward@arm.com>
8095 David Sherwood <david.sherwood@arm.com>
8097 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8099 * gimple-expr.h (is_gimple_addressable: Likewise.
8100 * gimple-expr.c (is_gimple_address): Likewise.
8101 * internal-fn.c (expand_call_mem_ref): New function.
8102 (expand_mask_load_optab_fn): Use it.
8103 (expand_mask_store_optab_fn): Likewise.
8105 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8106 Alan Hayward <alan.hayward@arm.com>
8107 David Sherwood <david.sherwood@arm.com>
8109 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8110 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8111 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8112 (cond_umax@var{mode}): Document.
8113 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8114 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8115 (cond_umin_optab, cond_umax_optab): New optabs.
8116 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8117 (COND_IOR, COND_XOR): New internal functions.
8118 * internal-fn.h (get_conditional_internal_fn): Declare.
8119 * internal-fn.c (cond_binary_direct): New macro.
8120 (expand_cond_binary_optab_fn): Likewise.
8121 (direct_cond_binary_optab_supported_p): Likewise.
8122 (get_conditional_internal_fn): New function.
8123 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8124 Cope with reduction statements that are vectorized as calls rather
8126 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8127 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8128 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8129 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8130 (UNSPEC_COND_EOR): New unspecs.
8131 (optab): Add mappings for them.
8132 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8133 (sve_int_op, sve_fp_op): New int attributes.
8135 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8136 Alan Hayward <alan.hayward@arm.com>
8137 David Sherwood <david.sherwood@arm.com>
8139 * optabs.def (while_ult_optab): New optab.
8140 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8141 * internal-fn.def (WHILE_ULT): New internal function.
8142 * internal-fn.h (direct_internal_fn_supported_p): New override
8143 that takes two types as argument.
8144 * internal-fn.c (while_direct): New macro.
8145 (expand_while_optab_fn): New function.
8146 (convert_optab_supported_p): Likewise.
8147 (direct_while_optab_supported_p): New macro.
8148 * wide-int.h (wi::udiv_ceil): New function.
8149 * tree-vectorizer.h (rgroup_masks): New structure.
8150 (vec_loop_masks): New typedef.
8151 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8153 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8154 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8155 (vect_max_vf): New function.
8156 (slpeel_make_loop_iterate_ntimes): Delete.
8157 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8158 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8159 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8160 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8161 internal-fn.h, stor-layout.h and optabs-query.h.
8162 (vect_set_loop_mask): New function.
8163 (add_preheader_seq): Likewise.
8164 (add_header_seq): Likewise.
8165 (interleave_supported_p): Likewise.
8166 (vect_maybe_permute_loop_masks): Likewise.
8167 (vect_set_loop_masks_directly): Likewise.
8168 (vect_set_loop_condition_masked): Likewise.
8169 (vect_set_loop_condition_unmasked): New function, split out from
8170 slpeel_make_loop_iterate_ntimes.
8171 (slpeel_make_loop_iterate_ntimes): Rename to..
8172 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8173 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8174 (vect_do_peeling): Update call accordingly.
8175 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8177 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8178 mask_compare_type, can_fully_mask_p and fully_masked_p.
8179 (release_vec_loop_masks): New function.
8180 (_loop_vec_info): Use it to free the loop masks.
8181 (can_produce_all_loop_masks_p): New function.
8182 (vect_get_max_nscalars_per_iter): Likewise.
8183 (vect_verify_full_masking): Likewise.
8184 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8185 retries, and free the mask rgroups before retrying. Check loop-wide
8186 reasons for disallowing fully-masked loops. Make the final decision
8187 about whether use a fully-masked loop or not.
8188 (vect_estimate_min_profitable_iters): Do not assume that peeling
8189 for the number of iterations will be needed for fully-masked loops.
8190 (vectorizable_reduction): Disable fully-masked loops.
8191 (vectorizable_live_operation): Likewise.
8192 (vect_halve_mask_nunits): New function.
8193 (vect_double_mask_nunits): Likewise.
8194 (vect_record_loop_mask): Likewise.
8195 (vect_get_loop_mask): Likewise.
8196 (vect_transform_loop): Handle the case in which the final loop
8197 iteration might handle a partial vector. Call vect_set_loop_condition
8198 instead of slpeel_make_loop_iterate_ntimes.
8199 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8200 (check_load_store_masking): New function.
8201 (prepare_load_store_mask): Likewise.
8202 (vectorizable_store): Handle fully-masked loops.
8203 (vectorizable_load): Likewise.
8204 (supportable_widening_operation): Use vect_halve_mask_nunits for
8206 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8207 (vect_gen_while): New function.
8208 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8209 (aarch64_uqdec<mode>): New insn.
8211 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8212 Alan Hayward <alan.hayward@arm.com>
8213 David Sherwood <david.sherwood@arm.com>
8215 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8216 (reduc_xor_scal_optab): New optabs.
8217 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8218 (reduc_xor_scal_@var{m}): Document.
8219 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8220 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8222 * fold-const-call.c (fold_const_call): Handle them.
8223 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8224 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8225 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8226 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8227 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8228 (UNSPEC_XORV): New unspecs.
8229 (optab): Add entries for them.
8230 (BITWISEV): New int iterator.
8231 (bit_reduc_op): New int attributes.
8233 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8234 Alan Hayward <alan.hayward@arm.com>
8235 David Sherwood <david.sherwood@arm.com>
8237 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8238 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8239 * optabs.def (vec_shl_insert_optab): New optab.
8240 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8241 (duplicate_and_interleave): Likewise.
8242 * tree-vect-loop.c: Include internal-fn.h.
8243 (neutral_op_for_slp_reduction): New function, split out from
8244 get_initial_defs_for_reduction.
8245 (get_initial_def_for_reduction): Handle option 2 for variable-length
8246 vectors by loading the neutral value into a vector and then shifting
8247 the initial value into element 0.
8248 (get_initial_defs_for_reduction): Replace the code argument with
8249 the neutral value calculated by neutral_op_for_slp_reduction.
8250 Use gimple_build_vector for constant-length vectors.
8251 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8252 but the first group_size elements have a neutral value.
8253 Use duplicate_and_interleave otherwise.
8254 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8255 Update call to get_initial_defs_for_reduction. Handle SLP
8256 reductions for variable-length vectors by creating one vector
8257 result for each scalar result, with the elements associated
8258 with other scalar results stubbed out with the neutral value.
8259 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8260 Require IFN_VEC_SHL_INSERT for double reductions on
8261 variable-length vectors, or SLP reductions that have
8262 a neutral value. Require can_duplicate_and_interleave_p
8263 support for variable-length unchained SLP reductions if there
8264 is no neutral value, such as for MIN/MAX reductions. Also require
8265 the number of vector elements to be a multiple of the number of
8266 SLP statements when doing variable-length unchained SLP reductions.
8267 Update call to vect_create_epilog_for_reduction.
8268 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8269 and remove initial values.
8270 (duplicate_and_interleave): Make public.
8271 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8272 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8274 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8275 Alan Hayward <alan.hayward@arm.com>
8276 David Sherwood <david.sherwood@arm.com>
8278 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8279 (can_duplicate_and_interleave_p): New function.
8280 (vect_get_and_check_slp_defs): Take the vector of statements
8281 rather than just the current one. Remove excess parentheses.
8282 Restriction rejectinon of vect_constant_def and vect_external_def
8283 for variable-length vectors to boolean types, or types for which
8284 can_duplicate_and_interleave_p is false.
8285 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8286 (duplicate_and_interleave): New function.
8287 (vect_get_constant_vectors): Use gimple_build_vector for
8288 constant-length vectors and suitable variable-length constant
8289 vectors. Use duplicate_and_interleave for other variable-length
8290 vectors. Don't defer the update when inserting new statements.
8292 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8293 Alan Hayward <alan.hayward@arm.com>
8294 David Sherwood <david.sherwood@arm.com>
8296 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8297 min_profitable_iters doesn't go negative.
8299 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8300 Alan Hayward <alan.hayward@arm.com>
8301 David Sherwood <david.sherwood@arm.com>
8303 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8304 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8305 * optabs.def (vec_mask_load_lanes_optab): New optab.
8306 (vec_mask_store_lanes_optab): Likewise.
8307 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8308 (MASK_STORE_LANES): Likewise.
8309 * internal-fn.c (mask_load_lanes_direct): New macro.
8310 (mask_store_lanes_direct): Likewise.
8311 (expand_mask_load_optab_fn): Handle masked operations.
8312 (expand_mask_load_lanes_optab_fn): New macro.
8313 (expand_mask_store_optab_fn): Handle masked operations.
8314 (expand_mask_store_lanes_optab_fn): New macro.
8315 (direct_mask_load_lanes_optab_supported_p): Likewise.
8316 (direct_mask_store_lanes_optab_supported_p): Likewise.
8317 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8319 (vect_load_lanes_supported): Likewise.
8320 * tree-vect-data-refs.c (strip_conversion): New function.
8321 (can_group_stmts_p): Likewise.
8322 (vect_analyze_data_ref_accesses): Use it instead of checking
8323 for a pair of assignments.
8324 (vect_store_lanes_supported): Take a masked_p parameter.
8325 (vect_load_lanes_supported): Likewise.
8326 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8327 vect_store_lanes_supported and vect_load_lanes_supported.
8328 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8329 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8330 parameter. Don't allow gaps for masked accesses.
8331 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8332 and vect_load_lanes_supported.
8333 (get_load_store_type): Take a masked_p parameter and update
8334 call to get_group_load_store_type.
8335 (vectorizable_store): Update call to get_load_store_type.
8336 Handle IFN_MASK_STORE_LANES.
8337 (vectorizable_load): Update call to get_load_store_type.
8338 Handle IFN_MASK_LOAD_LANES.
8340 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8341 Alan Hayward <alan.hayward@arm.com>
8342 David Sherwood <david.sherwood@arm.com>
8344 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8346 * config/aarch64/aarch64-protos.h
8347 (aarch64_sve_struct_memory_operand_p): Declare.
8348 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8349 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8350 (VPRED, vpred): Handle SVE structure modes.
8351 * config/aarch64/constraints.md (Utx): New constraint.
8352 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8353 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8354 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8355 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8356 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8357 structure modes. Split into pieces after RA.
8358 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8359 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8361 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8362 SVE structure modes.
8363 (aarch64_classify_address): Likewise.
8364 (sizetochar): Move earlier in file.
8365 (aarch64_print_operand): Handle SVE register lists.
8366 (aarch64_array_mode): New function.
8367 (aarch64_sve_struct_memory_operand_p): Likewise.
8368 (TARGET_ARRAY_MODE): Redefine.
8370 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8371 Alan Hayward <alan.hayward@arm.com>
8372 David Sherwood <david.sherwood@arm.com>
8374 * target.def (array_mode): New target hook.
8375 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8376 * doc/tm.texi: Regenerate.
8377 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8378 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8379 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8381 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8385 Alan Hayward <alan.hayward@arm.com>
8386 David Sherwood <david.sherwood@arm.com>
8388 * fold-const.c (fold_binary_loc): Check the argument types
8389 rather than the result type when testing for a vector operation.
8391 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8393 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8394 * doc/tm.texi: Regenerate.
8396 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8397 Alan Hayward <alan.hayward@arm.com>
8398 David Sherwood <david.sherwood@arm.com>
8400 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8401 (sve): Document new AArch64 extension.
8402 * doc/md.texi (w): Extend the description of the AArch64
8403 constraint to include SVE vectors.
8404 (Upl, Upa): Document new AArch64 predicate constraints.
8405 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8407 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8408 (msve-vector-bits=): New option.
8409 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8410 SVE when these are disabled.
8411 (sve): New extension.
8412 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8413 modes. Adjust their number of units based on aarch64_sve_vg.
8414 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8415 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8416 aarch64_addr_query_type.
8417 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8418 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8419 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8420 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8421 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8422 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8423 (aarch64_simd_imm_zero_p): Delete.
8424 (aarch64_check_zero_based_sve_index_immediate): Declare.
8425 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8426 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8427 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8428 (aarch64_sve_float_mul_immediate_p): Likewise.
8429 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8431 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8432 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8433 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8434 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8435 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8436 (aarch64_regmode_natural_size): Likewise.
8437 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8438 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8440 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8441 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8442 for VG and the SVE predicate registers.
8443 (V_ALIASES): Add a "z"-prefixed alias.
8444 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8445 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8446 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8447 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8448 (REG_CLASS_NAMES): Add entries for them.
8449 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8450 and the predicate registers.
8451 (aarch64_sve_vg): Declare.
8452 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8453 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8454 (REGMODE_NATURAL_SIZE): Define.
8455 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8457 * config/aarch64/aarch64.c: Include cfgrtl.h.
8458 (simd_immediate_info): Add a constructor for series vectors,
8459 and an associated step field.
8460 (aarch64_sve_vg): New variable.
8461 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8462 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8463 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8464 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8465 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8466 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8467 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8468 (aarch64_get_mask_mode): New functions.
8469 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8470 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8471 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8472 predicate modes and predicate registers. Explicitly restrict
8473 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8474 to store a vector mode if it is recognized by
8475 aarch64_classify_vector_mode.
8476 (aarch64_regmode_natural_size): New function.
8477 (aarch64_hard_regno_caller_save_mode): Return the original mode
8479 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8480 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8481 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8482 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8484 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8485 does not overlap dest if the function is frame-related. Handle
8487 (aarch64_split_add_offset): New function.
8488 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8489 them aarch64_add_offset.
8490 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8491 and update call to aarch64_sub_sp.
8492 (aarch64_add_cfa_expression): New function.
8493 (aarch64_expand_prologue): Pass extra temporary registers to the
8494 functions above. Handle the case in which we need to emit new
8495 DW_CFA_expressions for registers that were originally saved
8496 relative to the stack pointer, but now have to be expressed
8497 relative to the frame pointer.
8498 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8500 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8501 IP0 and IP1 values for SVE frames.
8502 (aarch64_expand_vec_series): New function.
8503 (aarch64_expand_sve_widened_duplicate): Likewise.
8504 (aarch64_expand_sve_const_vector): Likewise.
8505 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8506 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8507 into the register, rather than emitting a SET directly.
8508 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8509 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8510 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8511 (offset_9bit_signed_scaled_p): New functions.
8512 (aarch64_replicate_bitmask_imm): New function.
8513 (aarch64_bitmask_imm): Use it.
8514 (aarch64_cannot_force_const_mem): Reject expressions involving
8515 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8516 (aarch64_classify_index): Handle SVE indices, by requiring
8517 a plain register index with a scale that matches the element size.
8518 (aarch64_classify_address): Handle SVE addresses. Assert that
8519 the mode of the address is VOIDmode or an integer mode.
8520 Update call to aarch64_classify_symbol.
8521 (aarch64_classify_symbolic_expression): Update call to
8522 aarch64_classify_symbol.
8523 (aarch64_const_vec_all_in_range_p): New function.
8524 (aarch64_print_vector_float_operand): Likewise.
8525 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8526 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8527 and the FP immediates 1.0 and 0.5.
8528 (aarch64_print_address_internal): Handle SVE addresses.
8529 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8530 (aarch64_regno_regclass): Handle predicate registers.
8531 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8533 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8534 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8535 (aarch64_convert_sve_vector_bits): New function.
8536 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8537 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8539 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8540 Handle SVE vector and predicate modes. Accept VL-based constants
8541 that need only one temporary register, and VL offsets that require
8542 no temporary registers.
8543 (aarch64_conditional_register_usage): Mark the predicate registers
8544 as fixed if SVE isn't available.
8545 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8546 Return true for SVE vector and predicate modes.
8547 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8548 rather than an unsigned int. Handle SVE modes.
8549 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8551 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8553 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8554 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8555 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8556 (aarch64_sve_float_mul_immediate_p): New functions.
8557 (aarch64_sve_valid_immediate): New function.
8558 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8559 Explicitly reject structure modes. Check for INDEX constants.
8560 Handle PTRUE and PFALSE constants.
8561 (aarch64_check_zero_based_sve_index_immediate): New function.
8562 (aarch64_simd_imm_zero_p): Delete.
8563 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8564 vector modes. Accept constants in the range of CNT[BHWD].
8565 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8566 ask for an Advanced SIMD mode.
8567 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8568 (aarch64_simd_vector_alignment): Handle SVE predicates.
8569 (aarch64_vectorize_preferred_vector_alignment): New function.
8570 (aarch64_simd_vector_alignment_reachable): Use it instead of
8572 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8573 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8575 (MAX_VECT_LEN): Delete.
8576 (expand_vec_perm_d): Add a vec_flags field.
8577 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8578 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8579 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8581 (aarch64_evpc_rev): Rename to...
8582 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8583 (aarch64_evpc_rev_global): New function.
8584 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8585 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8587 (aarch64_evpc_sve_tbl): New function.
8588 (aarch64_expand_vec_perm_const_1): Update after rename of
8589 aarch64_evpc_rev. Handle SVE permutes too, trying
8590 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8591 than aarch64_evpc_tbl.
8592 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8593 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8594 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8595 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8596 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8597 (aarch64_expand_sve_vcond): New functions.
8598 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8599 of aarch64_vector_mode_p.
8600 (aarch64_dwarf_poly_indeterminate_value): New function.
8601 (aarch64_compute_pressure_classes): Likewise.
8602 (aarch64_can_change_mode_class): Likewise.
8603 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8604 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8605 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8606 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8607 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8608 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8609 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8610 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8612 (Dn, Dl, Dr): Accept const as well as const_vector.
8613 (Dz): Likewise. Compare against CONST0_RTX.
8614 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8615 of "vector" where appropriate.
8616 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8617 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8618 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8619 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8620 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8621 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8622 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8623 (v_int_equiv): Extend to SVE modes.
8624 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8626 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8627 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8628 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8629 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8630 (SVE_COND_FP_CMP): New int iterators.
8631 (perm_hilo): Handle the new unpack unspecs.
8632 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8634 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8635 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8636 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8637 (aarch64_equality_operator, aarch64_constant_vector_operand)
8638 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8639 (aarch64_sve_nonimmediate_operand): Likewise.
8640 (aarch64_sve_general_operand): Likewise.
8641 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8642 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8643 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8644 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8645 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8646 (aarch64_sve_float_arith_immediate): Likewise.
8647 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8648 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8649 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8650 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8651 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8652 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8653 (aarch64_sve_float_arith_operand): Likewise.
8654 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8655 (aarch64_sve_float_mul_operand): Likewise.
8656 (aarch64_sve_vec_perm_operand): Likewise.
8657 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8658 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8659 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8660 as well as const_vector.
8661 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8662 in file. Use CONST0_RTX and CONSTM1_RTX.
8663 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8664 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8665 Use aarch64_simd_imm_zero.
8666 * config/aarch64/aarch64-sve.md: New file.
8667 * config/aarch64/aarch64.md: Include it.
8668 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8669 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8670 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8671 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8672 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8673 (sve): New attribute.
8674 (enabled): Disable instructions with the sve attribute unless
8676 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8677 aarch64_expand_mov_immediate.
8678 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8679 CNT[BHSD] immediates.
8680 (movti): Split CONST_POLY_INT moves into two halves.
8681 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8682 Split additions that need a temporary here if the destination
8683 is the stack pointer.
8684 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8685 (*add<mode>3_poly_1): New instruction.
8686 (set_clobber_cc): New expander.
8688 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8690 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8691 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8692 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8693 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8694 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8695 Change innermode from fixed_mode_size to machine_mode.
8696 (simplify_subreg): Update call accordingly. Handle a constant-sized
8697 subreg of a variable-length CONST_VECTOR.
8699 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8700 Alan Hayward <alan.hayward@arm.com>
8701 David Sherwood <david.sherwood@arm.com>
8703 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8704 (add_offset_to_base): New function, split out from...
8705 (create_mem_ref): ...here. When handling a scale other than 1,
8706 check first whether the address is valid without the offset.
8707 Add it into the base if so, leaving the index and scale as-is.
8709 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8712 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8713 fold_for_warn before checking if arg2 is INTEGER_CST.
8715 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8717 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8718 (store_multiple_operation): Delete.
8719 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8720 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8721 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8722 guarded by TARGET_STRING.
8723 (rs6000_output_load_multiple): Delete.
8724 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8725 OPTION_MASK_STRING / TARGET_STRING handling.
8726 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8727 (const rs6000_opt_masks) <"string">: Change mask to 0.
8728 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8729 (MASK_STRING): Delete.
8730 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8732 (load_multiple): Delete.
8739 (store_multiple): Delete.
8746 (movmemsi_8reg): Delete.
8747 (corresponding unnamed define_insn): Delete.
8748 (movmemsi_6reg): Delete.
8749 (corresponding unnamed define_insn): Delete.
8750 (movmemsi_4reg): Delete.
8751 (corresponding unnamed define_insn): Delete.
8752 (movmemsi_2reg): Delete.
8753 (corresponding unnamed define_insn): Delete.
8754 (movmemsi_1reg): Delete.
8755 (corresponding unnamed define_insn): Delete.
8756 * config/rs6000/rs6000.opt (mno-string): New.
8757 (mstring): Replace by deprecation warning stub.
8758 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8760 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8762 * regrename.c (regrename_do_replace): If replacing the same
8763 reg multiple times, try to reuse last created gen_raw_REG.
8766 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8767 main to workaround a bug in GDB.
8769 2018-01-12 Tom de Vries <tom@codesourcery.com>
8772 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8774 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8776 PR rtl-optimization/80481
8777 * ira-color.c (get_cap_member): New function.
8778 (allocnos_conflict_by_live_ranges_p): Use it.
8779 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8780 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8782 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8785 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8786 (*saddl_se_1): Ditto.
8788 (*ssubl_se_1): Ditto.
8790 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8792 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8793 rather than wi::to_widest for DR_INITs.
8794 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8795 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8796 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8798 (vect_analyze_group_access_1): Note that here.
8800 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8802 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8803 polynomial type sizes.
8805 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8807 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8808 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8809 (gimple_add_tmp_var): Likewise.
8811 2018-01-12 Martin Liska <mliska@suse.cz>
8813 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8814 (gimple_alloc_sizes): Likewise.
8815 (dump_gimple_statistics): Use PRIu64 in printf format.
8816 * gimple.h: Change uint64_t to int.
8818 2018-01-12 Martin Liska <mliska@suse.cz>
8820 * tree-core.h: Use uint64_t instead of int.
8821 * tree.c (tree_node_counts): Likewise.
8822 (tree_node_sizes): Likewise.
8823 (dump_tree_statistics): Use PRIu64 in printf format.
8825 2018-01-12 Martin Liska <mliska@suse.cz>
8827 * Makefile.in: As qsort_chk is implemented in vec.c, add
8828 vec.o to linkage of gencfn-macros.
8829 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8830 passing the info to record_node_allocation_statistics.
8831 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8833 * ggc-common.c (struct ggc_usage): Add operator== and use
8834 it in operator< and compare function.
8835 * mem-stats.h (struct mem_usage): Likewise.
8836 * vec.c (struct vec_usage): Remove operator< and compare
8837 function. Can be simply inherited.
8839 2018-01-12 Martin Jambor <mjambor@suse.cz>
8842 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8843 * tree-ssa-math-opts.c: Include domwalk.h.
8844 (convert_mult_to_fma_1): New function.
8845 (fma_transformation_info): New type.
8846 (fma_deferring_state): Likewise.
8847 (cancel_fma_deferring): New function.
8848 (result_of_phi): Likewise.
8849 (last_fma_candidate_feeds_initial_phi): Likewise.
8850 (convert_mult_to_fma): Added deferring logic, split actual
8851 transformation to convert_mult_to_fma_1.
8852 (math_opts_dom_walker): New type.
8853 (math_opts_dom_walker::after_dom_children): New method, body moved
8854 here from pass_optimize_widening_mul::execute, added deferring logic
8856 (pass_optimize_widening_mul::execute): Moved most of code to
8857 math_opts_dom_walker::after_dom_children.
8858 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8859 * config/i386/i386.c (ix86_option_override_internal): Added
8860 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8862 2018-01-12 Richard Biener <rguenther@suse.de>
8865 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8866 inline instance vars.
8868 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8871 * config/rx/rx.c (rx_is_restricted_memory_address):
8874 2018-01-12 Richard Biener <rguenther@suse.de>
8876 PR tree-optimization/80846
8877 * target.def (split_reduction): New target hook.
8878 * targhooks.c (default_split_reduction): New function.
8879 * targhooks.h (default_split_reduction): Declare.
8880 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8881 target requests first reduce vectors by combining low and high
8883 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8884 (get_vectype_for_scalar_type_and_size): Export.
8885 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8886 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8887 * doc/tm.texi: Regenerate.
8888 * config/i386/i386.c (ix86_split_reduction): Implement
8889 TARGET_VECTORIZE_SPLIT_REDUCTION.
8891 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8894 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8895 in PIC mode except for TARGET_VXWORKS_RTP.
8896 * config/sparc/sparc.c: Include cfgrtl.h.
8897 (TARGET_INIT_PIC_REG): Define.
8898 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8899 (sparc_pic_register_p): New predicate.
8900 (sparc_legitimate_address_p): Use it.
8901 (sparc_legitimize_pic_address): Likewise.
8902 (sparc_delegitimize_address): Likewise.
8903 (sparc_mode_dependent_address_p): Likewise.
8904 (gen_load_pcrel_sym): Remove 4th parameter.
8905 (load_got_register): Adjust call to above. Remove obsolete stuff.
8906 (sparc_expand_prologue): Do not call load_got_register here.
8907 (sparc_flat_expand_prologue): Likewise.
8908 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8909 (sparc_use_pseudo_pic_reg): New function.
8910 (sparc_init_pic_reg): Likewise.
8911 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8912 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8914 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8916 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8917 Add item for branch_cost.
8919 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8921 PR rtl-optimization/83565
8922 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8923 not extend the result to a larger mode for rotate operations.
8924 (num_sign_bit_copies1): Likewise.
8926 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8929 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8931 Use values-Xc.o for -pedantic.
8932 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8934 2018-01-12 Martin Liska <mliska@suse.cz>
8937 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8939 (possible_polymorphic_call_targets): Use it.
8940 (ipa_devirt): Likewise.
8942 2018-01-12 Martin Liska <mliska@suse.cz>
8944 * profile-count.h (enum profile_quality): Use 0 as invalid
8945 enum value of profile_quality.
8947 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8949 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8950 -mext-string options.
8952 2018-01-12 Richard Biener <rguenther@suse.de>
8954 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8955 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8956 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8958 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8960 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8962 * configure.ac (--with-long-double-format): Add support for the
8963 configuration option to change the default long double format on
8965 * config.gcc (powerpc*-linux*-*): Likewise.
8966 * configure: Regenerate.
8967 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8968 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8969 used without modification.
8971 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8973 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8974 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8975 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8976 MISC_BUILTIN_SPEC_BARRIER.
8977 (rs6000_init_builtins): Likewise.
8978 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8980 (speculation_barrier): New define_insn.
8981 * doc/extend.texi: Document __builtin_speculation_barrier.
8983 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8986 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8987 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8988 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8990 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8991 integral modes instead of "ss" and "sd".
8992 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8993 vectors with 32-bit and 64-bit elements.
8994 (vecdupssescalarmodesuffix): New mode attribute.
8995 (vec_dup<mode>): Use it.
8997 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9000 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9001 frame if argument is passed on stack.
9003 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9006 * ree.c (combine_reaching_defs): Optimize also
9007 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9008 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9010 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9013 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9015 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9018 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9019 after they are computed.
9021 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9023 PR tree-optimization/83695
9024 * gimple-loop-linterchange.cc
9025 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9026 reset cached scev information after interchange.
9027 (pass_linterchange::execute): Remove call to scev_reset_htab.
9029 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9031 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9032 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9033 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9034 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9035 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9036 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9037 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9038 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9039 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9040 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9041 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9042 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9043 (V_lane_reg): Likewise.
9044 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9046 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9047 (vfmal_lane_low<mode>_intrinsic,
9048 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9049 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9050 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9051 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9052 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9053 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9055 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9057 * config/arm/arm-cpus.in (fp16fml): New feature.
9058 (ALL_SIMD): Add fp16fml.
9059 (armv8.2-a): Add fp16fml as an option.
9060 (armv8.3-a): Likewise.
9061 (armv8.4-a): Add fp16fml as part of fp16.
9062 * config/arm/arm.h (TARGET_FP16FML): Define.
9063 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9065 * config/arm/arm-modes.def (V2HF): Define.
9066 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9067 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9068 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9069 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9070 vfmsl_low, vfmsl_high): New set of builtins.
9071 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9072 (vfml_op): New code attribute.
9073 (VFMLHALVES): New int iterator.
9074 (VFML, VFMLSEL): New mode attributes.
9075 (V_reg): Define mapping for V2HF.
9076 (V_hi, V_lo): New mode attributes.
9077 (VF_constraint): Likewise.
9078 (vfml_half, vfml_half_selector): New int attributes.
9079 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9081 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9082 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9084 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9085 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9086 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9087 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9089 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9090 Document new effective target and option set.
9092 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9094 * config/arm/arm-cpus.in (armv8_4): New feature.
9095 (ARMv8_4a): New fgroup.
9096 (armv8.4-a): New arch.
9097 * config/arm/arm-tables.opt: Regenerate.
9098 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9099 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9100 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9101 Add matching rules for -march=armv8.4-a and extensions.
9102 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9104 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9107 * config/rx/rx.md (BW): New mode attribute.
9108 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9110 2018-01-11 Richard Biener <rguenther@suse.de>
9112 PR tree-optimization/83435
9113 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9114 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9115 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9117 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9118 Alan Hayward <alan.hayward@arm.com>
9119 David Sherwood <david.sherwood@arm.com>
9121 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9123 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9124 (aarch64_print_address_internal): Use it to check for a zero offset.
9126 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9127 Alan Hayward <alan.hayward@arm.com>
9128 David Sherwood <david.sherwood@arm.com>
9130 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9131 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9132 Return a poly_int64 rather than a HOST_WIDE_INT.
9133 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9134 rather than a HOST_WIDE_INT.
9135 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9136 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9137 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9138 final_offset from HOST_WIDE_INT to poly_int64.
9139 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9140 to_constant when getting the number of units in an Advanced SIMD
9142 (aarch64_builtin_vectorized_function): Check for a constant number
9144 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9146 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9147 attribute instead of GET_MODE_NUNITS.
9148 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9149 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9150 GET_MODE_SIZE for fixed-size registers.
9151 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9152 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9153 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9154 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9155 (aarch64_print_operand, aarch64_print_address_internal)
9156 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9157 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9158 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9159 Handle polynomial GET_MODE_SIZE.
9160 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9161 wider than SImode without modification.
9162 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9163 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9164 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9165 passing and returning SVE modes.
9166 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9167 rather than GEN_INT.
9168 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9169 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9170 (aarch64_allocate_and_probe_stack_space): Likewise.
9171 (aarch64_layout_frame): Cope with polynomial offsets.
9172 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9173 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9175 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9176 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9177 poly_int64 rather than a HOST_WIDE_INT.
9178 (aarch64_get_separate_components, aarch64_process_components)
9179 (aarch64_expand_prologue, aarch64_expand_epilogue)
9180 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9181 (aarch64_anchor_offset): New function, split out from...
9182 (aarch64_legitimize_address): ...here.
9183 (aarch64_builtin_vectorization_cost): Handle polynomial
9184 TYPE_VECTOR_SUBPARTS.
9185 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9187 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9188 number of elements from the PARALLEL rather than the mode.
9189 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9190 rather than GET_MODE_BITSIZE.
9191 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9192 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9193 (aarch64_expand_vec_perm_const_1): Handle polynomial
9194 d->perm.length () and d->perm elements.
9195 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9196 Apply to_constant to d->perm elements.
9197 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9198 polynomial CONST_VECTOR_NUNITS.
9199 (aarch64_move_pointer): Take amount as a poly_int64 rather
9201 (aarch64_progress_pointer): Avoid temporary variable.
9202 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9203 the mode attribute instead of GET_MODE.
9205 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9206 Alan Hayward <alan.hayward@arm.com>
9207 David Sherwood <david.sherwood@arm.com>
9209 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9210 x exists before using it.
9211 (aarch64_add_constant_internal): Rename to...
9212 (aarch64_add_offset_1): ...this. Replace regnum with separate
9213 src and dest rtxes. Handle the case in which they're different,
9214 including when the offset is zero. Replace scratchreg with an rtx.
9215 Use 2 additions if there is no spare register into which we can
9216 move a 16-bit constant.
9217 (aarch64_add_constant): Delete.
9218 (aarch64_add_offset): Replace reg with separate src and dest
9219 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9220 Use aarch64_add_offset_1.
9221 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9222 an rtx rather than an int. Take the delta as a poly_int64
9223 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9224 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9225 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9226 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9227 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9229 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9230 aarch64_add_constant.
9232 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9234 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9235 Use scalar_float_mode.
9237 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9239 * config/aarch64/aarch64-simd.md
9240 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9241 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9242 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9243 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9244 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9245 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9246 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9247 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9248 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9249 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9251 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9254 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9255 targ_options->x_arm_arch_string is non NULL.
9257 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9259 * config/aarch64/aarch64.h
9260 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9262 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9265 * expmed.c (emit_store_flag_force): Swap if const op0
9266 and change VOIDmode to mode of op0.
9268 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9270 PR rtl-optimization/83761
9271 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9272 than bytes to mode_for_size.
9274 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9277 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9278 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9281 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9284 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9285 when in layout mode.
9286 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9287 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9290 2018-01-10 Michael Collison <michael.collison@arm.com>
9292 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9293 * config/aarch64/aarch64-option-extension.def: Add
9294 AARCH64_OPT_EXTENSION of 'fp16fml'.
9295 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9296 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9297 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9298 * config/aarch64/constraints.md (Ui7): New constraint.
9299 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9300 (VFMLA_SEL_W): Ditto.
9303 (VFMLA16_LOW): New int iterator.
9304 (VFMLA16_HIGH): Ditto.
9305 (UNSPEC_FMLAL): New unspec.
9306 (UNSPEC_FMLSL): Ditto.
9307 (UNSPEC_FMLAL2): Ditto.
9308 (UNSPEC_FMLSL2): Ditto.
9309 (f16mac): New code attribute.
9310 * config/aarch64/aarch64-simd-builtins.def
9311 (aarch64_fmlal_lowv2sf): Ditto.
9312 (aarch64_fmlsl_lowv2sf): Ditto.
9313 (aarch64_fmlalq_lowv4sf): Ditto.
9314 (aarch64_fmlslq_lowv4sf): Ditto.
9315 (aarch64_fmlal_highv2sf): Ditto.
9316 (aarch64_fmlsl_highv2sf): Ditto.
9317 (aarch64_fmlalq_highv4sf): Ditto.
9318 (aarch64_fmlslq_highv4sf): Ditto.
9319 (aarch64_fmlal_lane_lowv2sf): Ditto.
9320 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9321 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9322 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9323 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9324 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9325 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9326 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9327 (aarch64_fmlal_lane_highv2sf): Ditto.
9328 (aarch64_fmlsl_lane_highv2sf): Ditto.
9329 (aarch64_fmlal_laneq_highv2sf): Ditto.
9330 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9331 (aarch64_fmlalq_lane_highv4sf): Ditto.
9332 (aarch64_fmlsl_lane_highv4sf): Ditto.
9333 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9334 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9335 * config/aarch64/aarch64-simd.md:
9336 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9337 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9338 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9339 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9340 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9341 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9342 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9343 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9344 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9345 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9346 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9347 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9348 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9349 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9350 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9351 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9352 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9353 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9354 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9355 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9356 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9357 (vfmlsl_low_u32): Ditto.
9358 (vfmlalq_low_u32): Ditto.
9359 (vfmlslq_low_u32): Ditto.
9360 (vfmlal_high_u32): Ditto.
9361 (vfmlsl_high_u32): Ditto.
9362 (vfmlalq_high_u32): Ditto.
9363 (vfmlslq_high_u32): Ditto.
9364 (vfmlal_lane_low_u32): Ditto.
9365 (vfmlsl_lane_low_u32): Ditto.
9366 (vfmlal_laneq_low_u32): Ditto.
9367 (vfmlsl_laneq_low_u32): Ditto.
9368 (vfmlalq_lane_low_u32): Ditto.
9369 (vfmlslq_lane_low_u32): Ditto.
9370 (vfmlalq_laneq_low_u32): Ditto.
9371 (vfmlslq_laneq_low_u32): Ditto.
9372 (vfmlal_lane_high_u32): Ditto.
9373 (vfmlsl_lane_high_u32): Ditto.
9374 (vfmlal_laneq_high_u32): Ditto.
9375 (vfmlsl_laneq_high_u32): Ditto.
9376 (vfmlalq_lane_high_u32): Ditto.
9377 (vfmlslq_lane_high_u32): Ditto.
9378 (vfmlalq_laneq_high_u32): Ditto.
9379 (vfmlslq_laneq_high_u32): Ditto.
9380 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9381 (AARCH64_FL_FOR_ARCH8_4): New.
9382 (AARCH64_ISA_F16FML): New ISA flag.
9383 (TARGET_F16FML): New feature flag for fp16fml.
9384 (doc/invoke.texi): Document new fp16fml option.
9386 2018-01-10 Michael Collison <michael.collison@arm.com>
9388 * config/aarch64/aarch64-builtins.c:
9389 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9390 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9391 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9392 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9393 (AARCH64_ISA_SHA3): New ISA flag.
9394 (TARGET_SHA3): New feature flag for sha3.
9395 * config/aarch64/iterators.md (sha512_op): New int attribute.
9396 (CRYPTO_SHA512): New int iterator.
9397 (UNSPEC_SHA512H): New unspec.
9398 (UNSPEC_SHA512H2): Ditto.
9399 (UNSPEC_SHA512SU0): Ditto.
9400 (UNSPEC_SHA512SU1): Ditto.
9401 * config/aarch64/aarch64-simd-builtins.def
9402 (aarch64_crypto_sha512hqv2di): New builtin.
9403 (aarch64_crypto_sha512h2qv2di): Ditto.
9404 (aarch64_crypto_sha512su0qv2di): Ditto.
9405 (aarch64_crypto_sha512su1qv2di): Ditto.
9406 (aarch64_eor3qv8hi): Ditto.
9407 (aarch64_rax1qv2di): Ditto.
9408 (aarch64_xarqv2di): Ditto.
9409 (aarch64_bcaxqv8hi): Ditto.
9410 * config/aarch64/aarch64-simd.md:
9411 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9412 (aarch64_crypto_sha512su0qv2di): Ditto.
9413 (aarch64_crypto_sha512su1qv2di): Ditto.
9414 (aarch64_eor3qv8hi): Ditto.
9415 (aarch64_rax1qv2di): Ditto.
9416 (aarch64_xarqv2di): Ditto.
9417 (aarch64_bcaxqv8hi): Ditto.
9418 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9419 (vsha512h2q_u64): Ditto.
9420 (vsha512su0q_u64): Ditto.
9421 (vsha512su1q_u64): Ditto.
9422 (veor3q_u16): Ditto.
9423 (vrax1q_u64): Ditto.
9425 (vbcaxq_u16): Ditto.
9426 * config/arm/types.md (crypto_sha512): New type attribute.
9427 (crypto_sha3): Ditto.
9428 (doc/invoke.texi): Document new sha3 option.
9430 2018-01-10 Michael Collison <michael.collison@arm.com>
9432 * config/aarch64/aarch64-builtins.c:
9433 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9434 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9435 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9436 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9437 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9438 (AARCH64_ISA_SM4): New ISA flag.
9439 (TARGET_SM4): New feature flag for sm4.
9440 * config/aarch64/aarch64-simd-builtins.def
9441 (aarch64_sm3ss1qv4si): Ditto.
9442 (aarch64_sm3tt1aq4si): Ditto.
9443 (aarch64_sm3tt1bq4si): Ditto.
9444 (aarch64_sm3tt2aq4si): Ditto.
9445 (aarch64_sm3tt2bq4si): Ditto.
9446 (aarch64_sm3partw1qv4si): Ditto.
9447 (aarch64_sm3partw2qv4si): Ditto.
9448 (aarch64_sm4eqv4si): Ditto.
9449 (aarch64_sm4ekeyqv4si): Ditto.
9450 * config/aarch64/aarch64-simd.md:
9451 (aarch64_sm3ss1qv4si): Ditto.
9452 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9453 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9454 (aarch64_sm4eqv4si): Ditto.
9455 (aarch64_sm4ekeyqv4si): Ditto.
9456 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9457 (sm3part_op): Ditto.
9458 (CRYPTO_SM3TT): Ditto.
9459 (CRYPTO_SM3PART): Ditto.
9460 (UNSPEC_SM3SS1): New unspec.
9461 (UNSPEC_SM3TT1A): Ditto.
9462 (UNSPEC_SM3TT1B): Ditto.
9463 (UNSPEC_SM3TT2A): Ditto.
9464 (UNSPEC_SM3TT2B): Ditto.
9465 (UNSPEC_SM3PARTW1): Ditto.
9466 (UNSPEC_SM3PARTW2): Ditto.
9467 (UNSPEC_SM4E): Ditto.
9468 (UNSPEC_SM4EKEY): Ditto.
9469 * config/aarch64/constraints.md (Ui2): New constraint.
9470 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9471 * config/arm/types.md (crypto_sm3): New type attribute.
9472 (crypto_sm4): Ditto.
9473 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9474 (vsm3tt1aq_u32): Ditto.
9475 (vsm3tt1bq_u32): Ditto.
9476 (vsm3tt2aq_u32): Ditto.
9477 (vsm3tt2bq_u32): Ditto.
9478 (vsm3partw1q_u32): Ditto.
9479 (vsm3partw2q_u32): Ditto.
9480 (vsm4eq_u32): Ditto.
9481 (vsm4ekeyq_u32): Ditto.
9482 (doc/invoke.texi): Document new sm4 option.
9484 2018-01-10 Michael Collison <michael.collison@arm.com>
9486 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9487 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9488 (AARCH64_FL_FOR_ARCH8_4): New.
9489 (AARCH64_FL_V8_4): New flag.
9490 (doc/invoke.texi): Document new armv8.4-a option.
9492 2018-01-10 Michael Collison <michael.collison@arm.com>
9494 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9495 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9496 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9497 * config/aarch64/aarch64-option-extension.def: Add
9498 AARCH64_OPT_EXTENSION of 'sha2'.
9499 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9500 (crypto): Disable sha2 and aes if crypto disabled.
9501 (crypto): Enable aes and sha2 if enabled.
9502 (simd): Disable sha2 and aes if simd disabled.
9503 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9505 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9506 (TARGET_SHA2): New feature flag for sha2.
9507 (TARGET_AES): New feature flag for aes.
9508 * config/aarch64/aarch64-simd.md:
9509 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9510 conditional on TARGET_AES.
9511 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9512 (aarch64_crypto_sha1hsi): Make pattern conditional
9514 (aarch64_crypto_sha1hv4si): Ditto.
9515 (aarch64_be_crypto_sha1hv4si): Ditto.
9516 (aarch64_crypto_sha1su1v4si): Ditto.
9517 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9518 (aarch64_crypto_sha1su0v4si): Ditto.
9519 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9520 (aarch64_crypto_sha256su0v4si): Ditto.
9521 (aarch64_crypto_sha256su1v4si): Ditto.
9522 (doc/invoke.texi): Document new aes and sha2 options.
9524 2018-01-10 Martin Sebor <msebor@redhat.com>
9526 PR tree-optimization/83781
9527 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9530 2018-01-11 Martin Sebor <msebor@gmail.com>
9531 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9533 PR tree-optimization/83501
9534 PR tree-optimization/81703
9536 * tree-ssa-strlen.c (get_string_cst): Rename...
9537 (get_string_len): ...to this. Handle global constants.
9538 (handle_char_store): Adjust.
9540 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9541 Jim Wilson <jimw@sifive.com>
9543 * config/riscv/riscv-protos.h (riscv_output_return): New.
9544 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9545 (riscv_attribute_table, riscv_output_return),
9546 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9547 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9548 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9550 (riscv_expand_prologue): Add early return for naked function.
9551 (riscv_expand_epilogue): Likewise.
9552 (riscv_function_ok_for_sibcall): Return false for naked function.
9553 (riscv_set_current_function): New.
9554 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9555 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9556 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9557 * doc/extend.texi (RISC-V Function Attributes): New.
9559 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9561 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9562 check for 128-bit long double before checking TCmode.
9563 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9564 128-bit long doubles before checking TFmode or TCmode.
9565 (FLOAT128_IBM_P): Likewise.
9567 2018-01-10 Martin Sebor <msebor@redhat.com>
9569 PR tree-optimization/83671
9570 * builtins.c (c_strlen): Unconditionally return zero for the empty
9572 Use -Warray-bounds for warnings.
9573 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9574 for non-constant array indices with COMPONENT_REF, arrays of
9575 arrays, and pointers to arrays.
9576 (gimple_fold_builtin_strlen): Determine and set length range for
9577 non-constant character arrays.
9579 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9582 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9585 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9587 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9589 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9592 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9593 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9594 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9595 indexed_or_indirect_operand predicate.
9596 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9597 (*vsx_le_perm_load_v8hi): Likewise.
9598 (*vsx_le_perm_load_v16qi): Likewise.
9599 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9600 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9601 (*vsx_le_perm_store_v8hi): Likewise.
9602 (*vsx_le_perm_store_v16qi): Likewise.
9603 (eight unnamed splitters): Likewise.
9605 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9607 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9608 * config/rs6000/emmintrin.h: Likewise.
9609 * config/rs6000/mmintrin.h: Likewise.
9610 * config/rs6000/xmmintrin.h: Likewise.
9612 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9615 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9617 * tree.c (tree_nop_conversion): Return true for location wrapper
9619 (maybe_wrap_with_location): New function.
9620 (selftest::check_strip_nops): New function.
9621 (selftest::test_location_wrappers): New function.
9622 (selftest::tree_c_tests): Call it.
9623 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9624 (maybe_wrap_with_location): New decl.
9625 (EXPR_LOCATION_WRAPPER_P): New macro.
9626 (location_wrapper_p): New inline function.
9627 (tree_strip_any_location_wrapper): New inline function.
9629 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9632 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9633 stack_realign_offset for the largest alignment of stack slot
9635 (ix86_find_max_used_stack_alignment): New function.
9636 (ix86_finalize_stack_frame_flags): Use it. Set
9637 max_used_stack_alignment if we don't realign stack.
9638 * config/i386/i386.h (machine_function): Add
9639 max_used_stack_alignment.
9641 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9643 * config/arm/arm.opt (-mbranch-cost): New option.
9644 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9647 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9650 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9651 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9653 2018-01-10 Richard Biener <rguenther@suse.de>
9656 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9657 early out so it also covers the case where we have a non-NULL
9660 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9662 PR tree-optimization/83753
9663 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9664 for non-strided grouped accesses if the number of elements is 1.
9666 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9669 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9670 * i386.h (TARGET_USE_GATHER): Define.
9671 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9673 2018-01-10 Martin Liska <mliska@suse.cz>
9676 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9677 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9679 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9680 CLEANUP_NO_PARTITIONING is not set.
9682 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9684 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9685 for vectors, as a partial revert of r254296.
9686 * rtl.h (const_vec_p): Delete.
9687 (const_vec_duplicate_p): Don't test for vector CONSTs.
9688 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9689 * expmed.c (make_tree): Likewise.
9692 * common.md (E, F): Use CONSTANT_P instead of checking for
9694 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9695 checking for CONST_VECTOR.
9697 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9700 * predict.c (force_edge_cold): Handle in more sane way edges
9703 2018-01-09 Carl Love <cel@us.ibm.com>
9705 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9707 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9708 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9709 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9710 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9711 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9712 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9713 * config/rs6000/rs6000-protos.h: Add extern defition for
9714 rs6000_generate_float2_double_code.
9715 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9717 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9718 (float2_v2df): Add define_expand.
9720 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9723 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9724 op_mode in the force_to_mode call.
9726 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9728 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9729 instead of checking each element individually.
9730 (aarch64_evpc_uzp): Likewise.
9731 (aarch64_evpc_zip): Likewise.
9732 (aarch64_evpc_ext): Likewise.
9733 (aarch64_evpc_rev): Likewise.
9734 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9735 instead of checking each element individually. Return true without
9737 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9738 whether all selected elements come from the same input, instead of
9739 checking each element individually. Remove calls to gen_rtx_REG,
9740 start_sequence and end_sequence and instead assert that no rtl is
9743 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9745 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9746 order of HIGH and CONST checks.
9748 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9750 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9751 if the destination isn't an SSA_NAME.
9753 2018-01-09 Richard Biener <rguenther@suse.de>
9755 PR tree-optimization/83668
9756 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9758 (canonicalize_loop_form): ... here, renamed from ...
9759 (canonicalize_loop_closed_ssa_form): ... this and amended to
9760 swap successor edges for loop exit blocks to make us use
9761 the RPO order we need for initial schedule generation.
9763 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9765 PR tree-optimization/64811
9766 * match.pd: When optimizing comparisons with Inf, avoid
9767 introducing or losing exceptions from comparisons with NaN.
9769 2018-01-09 Martin Liska <mliska@suse.cz>
9772 * asan.c (shadow_mem_size): Add gcc_assert.
9774 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9776 Don't save registers in main().
9779 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9780 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9781 * config/avr/avr.c (avr_set_current_function): Don't error if
9782 naked, OS_task or OS_main are specified at the same time.
9783 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9785 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9787 * common/config/avr/avr-common.c (avr_option_optimization_table):
9788 Switch on -mmain-is-OS_task for optimizing compilations.
9790 2018-01-09 Richard Biener <rguenther@suse.de>
9792 PR tree-optimization/83572
9793 * graphite.c: Include cfganal.h.
9794 (graphite_transform_loops): Connect infinite loops to exit
9795 and remove fake edges at the end.
9797 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9799 * ipa-inline.c (edge_badness): Revert accidental checkin.
9801 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9804 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9805 symbols; not inline clones.
9807 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9810 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9811 hard registers. Formatting fixes.
9813 PR preprocessor/83722
9814 * gcc.c (try_generate_repro): Pass
9815 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9816 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9819 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9820 Kito Cheng <kito.cheng@gmail.com>
9822 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9823 (riscv_leaf_function_p): Delete.
9824 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9826 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9828 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9830 (do_ifelse): New function.
9831 (do_isel): New function.
9832 (do_sub3): New function.
9833 (do_add3): New function.
9834 (do_load_mask_compare): New function.
9835 (do_overlap_load_compare): New function.
9836 (expand_compare_loop): New function.
9837 (expand_block_compare): Call expand_compare_loop() when appropriate.
9838 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9840 (-mblock-compare-inline-loop-limit): New option.
9842 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9845 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9846 Reverse order of second and third operands in first alternative.
9847 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9848 of first and second elements in UNSPEC_VPERMR vector.
9849 (altivec_expand_vec_perm_le): Likewise.
9851 2018-01-08 Jeff Law <law@redhat.com>
9853 PR rtl-optimizatin/81308
9854 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9855 (process_switch): If group_case_labels makes a change, then set
9857 (pass_convert_switch::execute): If a switch is converted, then
9858 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9860 PR rtl-optimization/81308
9861 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9864 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9866 PR target/83663 - Revert r255946
9867 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9868 generation for cases where splatting a value is not useful.
9869 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9870 across a vec_duplicate and a paradoxical subreg forming a vector
9871 mode to a vec_concat.
9873 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9875 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9876 -march=armv8.3-a variants.
9877 * config/arm/t-multilib: Likewise.
9878 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9880 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9882 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9884 (cceq_ior_compare_complement): Give it a name so I can use it, and
9885 change boolean_or_operator predicate to boolean_operator so it can
9886 be used to generate a crand.
9887 (eqne): New code iterator.
9888 (bd/bd_neg): New code_attrs.
9889 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9890 a single define_insn.
9891 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9892 decrement (bdnzt/bdnzf/bdzt/bdzf).
9893 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9894 with the new names of the branch decrement patterns, and added the
9895 names of the branch decrement conditional patterns.
9897 2018-01-08 Richard Biener <rguenther@suse.de>
9899 PR tree-optimization/83563
9900 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9903 2018-01-08 Richard Biener <rguenther@suse.de>
9906 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9908 2018-01-08 Richard Biener <rguenther@suse.de>
9910 PR tree-optimization/83685
9911 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9912 references to abnormals.
9914 2018-01-08 Richard Biener <rguenther@suse.de>
9917 * dwarf2out.c (output_indirect_strings): Handle empty
9918 skeleton_debug_str_hash.
9919 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9921 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9923 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9924 (emit_store_direct): Likewise.
9925 (arc_trampoline_adjust_address): Likewise.
9926 (arc_asm_trampoline_template): New function.
9927 (arc_initialize_trampoline): Use asm_trampoline_template.
9928 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9929 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9930 * config/arc/arc.md (flush_icache): Delete pattern.
9932 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9934 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9935 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9938 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9941 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9942 by not USED_FOR_TARGET.
9943 (make_pass_resolve_sw_modes): Likewise.
9945 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9947 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9950 2018-01-08 Richard Biener <rguenther@suse.de>
9953 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9955 2018-01-08 Richard Biener <rguenther@suse.de>
9958 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9960 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9963 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9964 basic blocks with a small number of successors.
9965 (convert_control_dep_chain_into_preds): Improve handling of
9967 (dump_predicates): Split apart into...
9968 (dump_pred_chain): ...here...
9969 (dump_pred_info): ...and here.
9970 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9971 (can_chain_union_be_invalidated_p): Improve check for invalidation
9973 (uninit_uses_cannot_happen): Avoid unnecessary if
9974 convert_control_dep_chain_into_preds yielded nothing.
9976 2018-01-06 Martin Sebor <msebor@redhat.com>
9978 PR tree-optimization/83640
9979 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9980 subtracting negative offset from size.
9981 (builtin_access::overlap): Adjust offset bounds of the access to fall
9982 within the size of the object if possible.
9984 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9986 PR rtl-optimization/83699
9987 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9988 extract_bit_field_as_subreg to cases in which the extracted
9989 value is also a vector.
9991 * lra-constraints.c (process_alt_operands): Test for the equivalence
9992 substitutions when detecting a possible reload cycle.
9994 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9997 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9998 by default if flag_selective_schedling{,2}. Formatting fixes.
10000 PR rtl-optimization/83682
10001 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10002 if it has non-VECTOR_MODE element mode.
10003 (vec_duplicate_p): Likewise.
10005 PR middle-end/83694
10006 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10007 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10009 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10012 * config/i386/i386-builtin.def
10013 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10014 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10015 Require also OPTION_MASK_ISA_AVX512F in addition to
10016 OPTION_MASK_ISA_GFNI.
10017 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10018 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10019 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10020 to OPTION_MASK_ISA_GFNI.
10021 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10022 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10023 OPTION_MASK_ISA_AVX512BW.
10024 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10025 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10026 addition to OPTION_MASK_ISA_GFNI.
10027 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10028 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10029 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10030 to OPTION_MASK_ISA_GFNI.
10031 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10032 a requirement for all ISAs rather than any of them with a few
10034 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10036 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10037 bitmasks to be enabled with 3 exceptions, instead of requiring any
10038 enabled ISA with lots of exceptions.
10039 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10040 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10041 Change avx512bw in isa attribute to avx512f.
10042 * config/i386/sgxintrin.h: Add license boilerplate.
10043 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10044 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10045 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10046 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10048 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10049 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10050 temporarily sse2 rather than sse if not enabled already.
10053 * config/i386/sse.md (VI248_VLBW): Rename to ...
10054 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10055 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10056 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10057 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10058 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10059 mode iterator instead of VI248_VLBW.
10061 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10063 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10064 (record_modified): Skip clobbers; add debug output.
10065 (param_change_prob): Use sreal frequencies.
10067 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10069 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10070 punt for user-aligned variables.
10072 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10074 * tree-chrec.c (chrec_contains_symbols): Return true for
10077 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10080 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10081 of (x|y) == x for BICS pattern.
10083 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10085 PR tree-optimization/83605
10086 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10087 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10090 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10092 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10093 * config/epiphany/rtems.h: New file.
10095 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10096 Uros Bizjak <ubizjak@gmail.com>
10099 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10100 QIreg_operand instead of register_operand predicate.
10101 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10102 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10103 comments instead of -fmitigate[-_]rop.
10105 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10108 * cgraphunit.c (symbol_table::compile): Switch to text_section
10109 before calling assembly_start debug hook.
10110 * run-rtl-passes.c (run_rtl_passes): Likewise.
10113 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10115 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10116 range_int_cst_p rather than !symbolic_range_p before calling
10117 extract_range_from_multiplicative_op_1.
10119 2018-01-04 Jeff Law <law@redhat.com>
10121 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10122 redundant test in assertion.
10124 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10126 * doc/rtl.texi: Document machine_mode wrapper classes.
10128 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10130 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10131 using tree_to_uhwi.
10133 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10135 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10136 the VEC_PERM_EXPR fold to fail.
10138 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10141 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10142 to switched_sections.
10144 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10147 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10148 test for d.testing.
10150 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10153 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10154 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10156 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10159 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10160 is BLKmode and bitpos not zero or mode change is needed.
10162 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10165 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10168 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10171 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10172 instead of MULT rtx. Update all corresponding splitters.
10173 (*saddl_se): Ditto.
10174 (*ssub<modesuffix>): Ditto.
10175 (*ssubl_se): Ditto.
10176 (*cmp_sadd_di): Update split patterns.
10177 (*cmp_sadd_si): Ditto.
10178 (*cmp_sadd_sidi): Ditto.
10179 (*cmp_ssub_di): Ditto.
10180 (*cmp_ssub_si): Ditto.
10181 (*cmp_ssub_sidi): Ditto.
10182 * config/alpha/predicates.md (const23_operand): New predicate.
10183 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10184 Look for ASHIFT, not MULT inner operand.
10185 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10187 2018-01-04 Martin Liska <mliska@suse.cz>
10189 PR gcov-profile/83669
10190 * gcov.c (output_intermediate_file): Add version to intermediate
10192 * doc/gcov.texi: Document new field 'version' in intermediate
10193 file format. Fix location of '-k' option of gcov command.
10195 2018-01-04 Martin Liska <mliska@suse.cz>
10198 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10200 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10202 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10204 2018-01-03 Martin Sebor <msebor@redhat.com>
10206 PR tree-optimization/83655
10207 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10208 checking calls with invalid arguments.
10210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10212 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10213 (vectorizable_mask_load_store): Delete.
10214 (vectorizable_call): Return false for masked loads and stores.
10215 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10216 instead of gimple_assign_rhs1.
10217 (vectorizable_load): Handle IFN_MASK_LOAD.
10218 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10222 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10224 (vectorizable_mask_load_store): ...here.
10225 (vectorizable_load): ...and here.
10227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10229 * tree-vect-stmts.c (vect_build_all_ones_mask)
10230 (vect_build_zero_merge_argument): New functions, split out from...
10231 (vectorizable_load): ...here.
10233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10235 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10237 (vectorizable_mask_load_store): ...here.
10238 (vectorizable_store): ...and here.
10240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10242 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10244 (vectorizable_mask_load_store): ...here.
10246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10248 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10249 (vect_model_store_cost): Take a vec_load_store_type instead of a
10251 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10252 (vect_model_store_cost): Take a vec_load_store_type instead of a
10254 (vectorizable_mask_load_store): Update accordingly.
10255 (vectorizable_store): Likewise.
10256 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10258 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10260 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10261 IFN_MASK_LOAD calls here rather than...
10262 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10265 Alan Hayward <alan.hayward@arm.com>
10266 David Sherwood <david.sherwood@arm.com>
10268 * expmed.c (extract_bit_field_1): For vector extracts,
10269 fall back to extract_bit_field_as_subreg if vec_extract
10272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10273 Alan Hayward <alan.hayward@arm.com>
10274 David Sherwood <david.sherwood@arm.com>
10276 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10277 they are variable or constant sized.
10278 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10279 slots for constant-sized data.
10281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10282 Alan Hayward <alan.hayward@arm.com>
10283 David Sherwood <david.sherwood@arm.com>
10285 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10286 handling COND_EXPRs with boolean comparisons, try to find a better
10287 basis for the mask type than the boolean itself.
10289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10291 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10292 is calculated and how it can be overridden.
10293 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10294 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10296 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10300 Alan Hayward <alan.hayward@arm.com>
10301 David Sherwood <david.sherwood@arm.com>
10303 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10304 Remove the mode argument.
10305 (aarch64_simd_valid_immediate): Remove the mode and inverse
10307 * config/aarch64/iterators.md (bitsize): New iterator.
10308 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10309 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10310 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10311 aarch64_simd_valid_immediate.
10312 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10313 (aarch64_reg_or_bic_imm): Likewise.
10314 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10315 with an insn_type enum and msl with a modifier_type enum.
10316 Replace element_width with a scalar_mode. Change the shift
10317 to unsigned int. Add constructors for scalar_float_mode and
10318 scalar_int_mode elements.
10319 (aarch64_vect_float_const_representable_p): Delete.
10320 (aarch64_can_const_movi_rtx_p)
10321 (aarch64_simd_scalar_immediate_valid_for_move)
10322 (aarch64_simd_make_constant): Update call to
10323 aarch64_simd_valid_immediate.
10324 (aarch64_advsimd_valid_immediate_hs): New function.
10325 (aarch64_advsimd_valid_immediate): Likewise.
10326 (aarch64_simd_valid_immediate): Remove mode and inverse
10327 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10328 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10329 and aarch64_float_const_representable_p on the result.
10330 (aarch64_output_simd_mov_immediate): Remove mode argument.
10331 Update call to aarch64_simd_valid_immediate and use of
10332 simd_immediate_info.
10333 (aarch64_output_scalar_simd_mov_immediate): Update call
10336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10337 Alan Hayward <alan.hayward@arm.com>
10338 David Sherwood <david.sherwood@arm.com>
10340 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10341 (mode_nunits): Likewise CONST_MODE_NUNITS.
10342 * machmode.def (ADJUST_NUNITS): Document.
10343 * genmodes.c (mode_data::need_nunits_adj): New field.
10344 (blank_mode): Update accordingly.
10345 (adj_nunits): New variable.
10346 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10348 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10349 listed in adj_nunits.
10350 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10351 listed in adj_nunits. Don't emit case statements for such modes.
10352 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10353 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10354 nothing if adj_nunits is nonnull.
10355 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10356 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10357 (emit_mode_fbit): Update use of print_maybe_const_decl.
10358 (emit_move_size): Likewise. Treat the array as non-const
10360 (emit_mode_adjustments): Handle adj_nunits.
10362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10364 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10365 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10366 (VECTOR_MODES): Use it.
10367 (make_vector_modes): Take the prefix as an argument.
10369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10370 Alan Hayward <alan.hayward@arm.com>
10371 David Sherwood <david.sherwood@arm.com>
10373 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10374 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10375 for MODE_VECTOR_BOOL.
10376 * machmode.def (VECTOR_BOOL_MODE): Document.
10377 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10378 (make_vector_bool_mode): New function.
10379 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10381 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10382 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10384 * stor-layout.c (int_mode_for_mode): Likewise.
10385 * tree.c (build_vector_type_for_mode): Likewise.
10386 * varasm.c (output_constant_pool_2): Likewise.
10387 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10388 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10389 for MODE_VECTOR_BOOL.
10390 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10391 of mode class checks.
10392 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10393 instead of a list of mode class checks.
10394 (expand_vector_scalar_condition): Likewise.
10395 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10398 Alan Hayward <alan.hayward@arm.com>
10399 David Sherwood <david.sherwood@arm.com>
10401 * machmode.h (mode_size): Change from unsigned short to
10403 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10404 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10405 or if measurement_type is not polynomial.
10406 (fixed_size_mode::includes_p): Check for constant-sized modes.
10407 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10408 return a poly_uint16 rather than an unsigned short.
10409 (emit_mode_size): Change the type of mode_size from unsigned short
10410 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10411 (emit_mode_adjustments): Cope with polynomial vector sizes.
10412 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10414 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10416 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10417 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10418 * caller-save.c (setup_save_areas): Likewise.
10419 (replace_reg_with_saved_mem): Likewise.
10420 * calls.c (emit_library_call_value_1): Likewise.
10421 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10422 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10423 (gen_lowpart_for_combine): Likewise.
10424 * convert.c (convert_to_integer_1): Likewise.
10425 * cse.c (equiv_constant, cse_insn): Likewise.
10426 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10427 (cselib_subst_to_values): Likewise.
10428 * dce.c (word_dce_process_block): Likewise.
10429 * df-problems.c (df_word_lr_mark_ref): Likewise.
10430 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10431 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10432 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10433 (rtl_for_decl_location): Likewise.
10434 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10435 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10436 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10437 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10438 (expand_expr_real_1): Likewise.
10439 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10440 (pad_below): Likewise.
10441 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10442 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10443 * ira.c (get_subreg_tracking_sizes): Likewise.
10444 * ira-build.c (ira_create_allocno_objects): Likewise.
10445 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10446 (ira_sort_regnos_for_alter_reg): Likewise.
10447 * ira-costs.c (record_operand_costs): Likewise.
10448 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10449 (resolve_simple_move): Likewise.
10450 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10451 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10452 (lra_constraints): Likewise.
10453 (CONST_POOL_OK_P): Reject variable-sized modes.
10454 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10455 (add_pseudo_to_slot, lra_spill): Likewise.
10456 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10457 * optabs-query.c (get_best_extraction_insn): Likewise.
10458 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10459 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10460 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10461 * recog.c (offsettable_address_addr_space_p): Likewise.
10462 * regcprop.c (maybe_mode_change): Likewise.
10463 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10464 * regrename.c (build_def_use): Likewise.
10465 * regstat.c (dump_reg_info): Likewise.
10466 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10467 (find_reloads, find_reloads_subreg_address): Likewise.
10468 * reload1.c (eliminate_regs_1): Likewise.
10469 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10470 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10471 (simplify_binary_operation_1, simplify_subreg): Likewise.
10472 * targhooks.c (default_function_arg_padding): Likewise.
10473 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10474 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10475 (verify_gimple_assign_ternary): Likewise.
10476 * tree-inline.c (estimate_move_cost): Likewise.
10477 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10478 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10479 (get_address_cost_ainc): Likewise.
10480 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10481 (vect_supportable_dr_alignment): Likewise.
10482 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10483 (vectorizable_reduction): Likewise.
10484 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10485 (vectorizable_operation, vectorizable_load): Likewise.
10486 * tree.c (build_same_sized_truth_vector_type): Likewise.
10487 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10488 * var-tracking.c (emit_note_insn_var_location): Likewise.
10489 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10490 (ADDR_VEC_ALIGN): Likewise.
10492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10493 Alan Hayward <alan.hayward@arm.com>
10494 David Sherwood <david.sherwood@arm.com>
10496 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10498 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10499 or if measurement_type is polynomial.
10500 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10501 * combine.c (make_extraction): Likewise.
10502 * dse.c (find_shift_sequence): Likewise.
10503 * dwarf2out.c (mem_loc_descriptor): Likewise.
10504 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10505 (extract_bit_field, extract_low_bits): Likewise.
10506 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10507 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10508 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10509 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10510 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10511 * reload.c (find_reloads): Likewise.
10512 * reload1.c (alter_reg): Likewise.
10513 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10514 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10515 * tree-if-conv.c (predicate_mem_writes): Likewise.
10516 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10517 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10518 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10519 * valtrack.c (dead_debug_insert_temp): Likewise.
10520 * varasm.c (mergeable_constant_section): Likewise.
10521 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10523 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10524 Alan Hayward <alan.hayward@arm.com>
10525 David Sherwood <david.sherwood@arm.com>
10527 * expr.c (expand_assignment): Cope with polynomial mode sizes
10528 when assigning to a CONCAT.
10530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10531 Alan Hayward <alan.hayward@arm.com>
10532 David Sherwood <david.sherwood@arm.com>
10534 * machmode.h (mode_precision): Change from unsigned short to
10536 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10538 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10539 or if measurement_type is not polynomial.
10540 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10541 in which the mode is already known to be a scalar_int_mode.
10542 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10543 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10545 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10546 for GET_MODE_PRECISION.
10547 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10548 for GET_MODE_PRECISION.
10549 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10551 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10552 (expand_field_assignment, make_extraction): Likewise.
10553 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10554 (get_last_value): Likewise.
10555 * convert.c (convert_to_integer_1): Likewise.
10556 * cse.c (cse_insn): Likewise.
10557 * expr.c (expand_expr_real_1): Likewise.
10558 * lra-constraints.c (simplify_operand_subreg): Likewise.
10559 * optabs-query.c (can_atomic_load_p): Likewise.
10560 * optabs.c (expand_atomic_load): Likewise.
10561 (expand_atomic_store): Likewise.
10562 * ree.c (combine_reaching_defs): Likewise.
10563 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10564 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10565 * tree.h (type_has_mode_precision_p): Likewise.
10566 * ubsan.c (instrument_si_overflow): Likewise.
10568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10569 Alan Hayward <alan.hayward@arm.com>
10570 David Sherwood <david.sherwood@arm.com>
10572 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10573 polynomial numbers of units.
10574 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10575 (valid_vector_subparts_p): New function.
10576 (build_vector_type): Remove temporary shim and take the number
10577 of units as a poly_uint64 rather than an int.
10578 (build_opaque_vector_type): Take the number of units as a
10579 poly_uint64 rather than an int.
10580 * tree.c (build_vector_from_ctor): Handle polynomial
10581 TYPE_VECTOR_SUBPARTS.
10582 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10583 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10584 (build_vector_from_val): If the number of units is variable,
10585 use build_vec_duplicate_cst for constant operands and
10586 VEC_DUPLICATE_EXPR otherwise.
10587 (make_vector_type): Remove temporary is_constant ().
10588 (build_vector_type, build_opaque_vector_type): Take the number of
10589 units as a poly_uint64 rather than an int.
10590 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10592 * cfgexpand.c (expand_debug_expr): Likewise.
10593 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10594 (store_constructor, expand_expr_real_1): Likewise.
10595 (const_scalar_mask_from_tree): Likewise.
10596 * fold-const-call.c (fold_const_reduction): Likewise.
10597 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10598 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10599 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10600 (fold_relational_const): Likewise.
10601 (native_interpret_vector): Likewise. Change the size from an
10602 int to an unsigned int.
10603 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10604 TYPE_VECTOR_SUBPARTS.
10605 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10606 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10607 duplicating a non-constant operand into a variable-length vector.
10608 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10609 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10610 * ipa-icf.c (sem_variable::equals): Likewise.
10611 * match.pd: Likewise.
10612 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10613 * print-tree.c (print_node): Likewise.
10614 * stor-layout.c (layout_type): Likewise.
10615 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10616 * tree-cfg.c (verify_gimple_comparison): Likewise.
10617 (verify_gimple_assign_binary): Likewise.
10618 (verify_gimple_assign_ternary): Likewise.
10619 (verify_gimple_assign_single): Likewise.
10620 * tree-pretty-print.c (dump_generic_node): Likewise.
10621 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10622 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10623 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10624 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10625 (vect_shift_permute_load_chain): Likewise.
10626 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10627 (expand_vector_condition, optimize_vector_constructor): Likewise.
10628 (lower_vec_perm, get_compute_type): Likewise.
10629 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10630 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10631 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10632 (vect_recog_mask_conversion_pattern): Likewise.
10633 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10634 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10635 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10636 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10637 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10638 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10639 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10640 (supportable_widening_operation): Likewise.
10641 (supportable_narrowing_operation): Likewise.
10642 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10644 * varasm.c (output_constant): Likewise.
10646 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10647 Alan Hayward <alan.hayward@arm.com>
10648 David Sherwood <david.sherwood@arm.com>
10650 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10651 so that both the length == 3 and length != 3 cases set up their
10652 own permute vectors. Add comments explaining why we know the
10653 number of elements is constant.
10654 (vect_permute_load_chain): Likewise.
10656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10657 Alan Hayward <alan.hayward@arm.com>
10658 David Sherwood <david.sherwood@arm.com>
10660 * machmode.h (mode_nunits): Change from unsigned char to
10662 (ONLY_FIXED_SIZE_MODES): New macro.
10663 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10664 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10665 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10667 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10668 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10669 or if measurement_type is not polynomial.
10670 * genmodes.c (ZERO_COEFFS): New macro.
10671 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10673 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10674 Use ZERO_COEFFS when emitting initializers.
10675 * data-streamer.h (bp_pack_poly_value): New function.
10676 (bp_unpack_poly_value): Likewise.
10677 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10678 for GET_MODE_NUNITS.
10679 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10680 for GET_MODE_NUNITS.
10681 * tree.c (make_vector_type): Remove temporary shim and make
10682 the real function take the number of units as a poly_uint64
10683 rather than an int.
10684 (build_vector_type_for_mode): Handle polynomial nunits.
10685 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10686 * emit-rtl.c (const_vec_series_p_1): Likewise.
10687 (gen_rtx_CONST_VECTOR): Likewise.
10688 * fold-const.c (test_vec_duplicate_folding): Likewise.
10689 * genrecog.c (validate_pattern): Likewise.
10690 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10691 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10692 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10693 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10694 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10695 * rtlanal.c (subreg_get_info): Likewise.
10696 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10697 (vect_grouped_load_supported): Likewise.
10698 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10699 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10700 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10701 (simplify_const_unary_operation, simplify_binary_operation_1)
10702 (simplify_const_binary_operation, simplify_ternary_operation)
10703 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10704 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10705 instead of CONST_VECTOR_NUNITS.
10706 * varasm.c (output_constant_pool_2): Likewise.
10707 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10708 explicit-encoded elements in the XVEC for variable-length vectors.
10710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10712 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10715 Alan Hayward <alan.hayward@arm.com>
10716 David Sherwood <david.sherwood@arm.com>
10718 * coretypes.h (fixed_size_mode): Declare.
10719 (fixed_size_mode_pod): New typedef.
10720 * builtins.h (target_builtins::x_apply_args_mode)
10721 (target_builtins::x_apply_result_mode): Change type to
10722 fixed_size_mode_pod.
10723 * builtins.c (apply_args_size, apply_result_size, result_vector)
10724 (expand_builtin_apply_args_1, expand_builtin_apply)
10725 (expand_builtin_return): Update accordingly.
10727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10729 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10730 * cselib.c (cselib_hash_rtx): Likewise.
10731 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10732 CONST_VECTOR encoding.
10734 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10735 Jeff Law <law@redhat.com>
10738 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10739 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10740 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10741 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10744 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10745 explicitly probe *sp in a noreturn function if there were any callee
10746 register saves or frame pointer is needed.
10748 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10751 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10752 BLKmode for ternary, binary or unary expressions.
10755 * var-tracking.c (delete_vta_debug_insn): New inline function.
10756 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10757 insns from get_insns () to NULL instead of each bb separately.
10758 Use delete_vta_debug_insn. No longer static.
10759 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10760 delete_vta_debug_insns callers.
10761 * rtl.h (delete_vta_debug_insns): Declare.
10762 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10763 instead of variable_tracking_main.
10765 2018-01-03 Martin Sebor <msebor@redhat.com>
10767 PR tree-optimization/83603
10768 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10769 arguments past the endof the argument list in functions declared
10770 without a prototype.
10771 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10772 Avoid checking when arguments are null.
10774 2018-01-03 Martin Sebor <msebor@redhat.com>
10777 * doc/extend.texi (attribute const): Fix a typo.
10778 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10779 issuing -Wsuggest-attribute for void functions.
10781 2018-01-03 Martin Sebor <msebor@redhat.com>
10783 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10784 offset_int::from instead of wide_int::to_shwi.
10785 (maybe_diag_overlap): Remove assertion.
10786 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10787 * gimple-ssa-sprintf.c (format_directive): Same.
10788 (parse_directive): Same.
10789 (sprintf_dom_walker::compute_format_length): Same.
10790 (try_substitute_return_value): Same.
10792 2018-01-03 Jeff Law <law@redhat.com>
10794 PR middle-end/83654
10795 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10796 non-constant residual for zero at runtime and avoid probing in
10797 that case. Reorganize code for trailing problem to mirror handling
10800 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10802 PR tree-optimization/83501
10803 * tree-ssa-strlen.c (get_string_cst): New.
10804 (handle_char_store): Call get_string_cst.
10806 2018-01-03 Martin Liska <mliska@suse.cz>
10808 PR tree-optimization/83593
10809 * tree-ssa-strlen.c: Include tree-cfg.h.
10810 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10811 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10812 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10814 (strlen_dom_walker::before_dom_children): Call
10815 gimple_purge_dead_eh_edges. Dump tranformation with details
10817 (strlen_dom_walker::before_dom_children): Update call by adding
10818 new argument cleanup_eh.
10819 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10821 2018-01-03 Martin Liska <mliska@suse.cz>
10824 * cif-code.def (VARIADIC_THUNK): New enum value.
10825 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10828 2018-01-03 Jan Beulich <jbeulich@suse.com>
10830 * sse.md (mov<mode>_internal): Tighten condition for when to use
10831 vmovdqu<ssescalarsize> for TI and OI modes.
10833 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10835 Update copyright years.
10837 2018-01-03 Martin Liska <mliska@suse.cz>
10840 * ipa-visibility.c (function_and_variable_visibility): Skip
10841 functions with noipa attribure.
10843 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10845 * gcc.c (process_command): Update copyright notice dates.
10846 * gcov-dump.c (print_version): Ditto.
10847 * gcov.c (print_version): Ditto.
10848 * gcov-tool.c (print_version): Ditto.
10849 * gengtype.c (create_file): Ditto.
10850 * doc/cpp.texi: Bump @copying's copyright year.
10851 * doc/cppinternals.texi: Ditto.
10852 * doc/gcc.texi: Ditto.
10853 * doc/gccint.texi: Ditto.
10854 * doc/gcov.texi: Ditto.
10855 * doc/install.texi: Ditto.
10856 * doc/invoke.texi: Ditto.
10858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10860 * vector-builder.h (vector_builder::m_full_nelts): Change from
10861 unsigned int to poly_uint64.
10862 (vector_builder::full_nelts): Update prototype accordingly.
10863 (vector_builder::new_vector): Likewise.
10864 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10865 (vector_builder::operator ==): Likewise.
10866 (vector_builder::finalize): Likewise.
10867 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10868 Take the number of elements as a poly_uint64 rather than an
10870 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10871 from unsigned int to poly_uint64.
10872 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10873 (vec_perm_indices::new_vector): Likewise.
10874 (vec_perm_indices::length): Likewise.
10875 (vec_perm_indices::nelts_per_input): Likewise.
10876 (vec_perm_indices::input_nelts): Likewise.
10877 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10878 number of elements per input as a poly_uint64 rather than an
10879 unsigned int. Use the original encoding for variable-length
10880 vectors, rather than clamping each individual element.
10881 For the second and subsequent elements in each pattern,
10882 clamp the step and base before clamping their sum.
10883 (vec_perm_indices::series_p): Handle polynomial element counts.
10884 (vec_perm_indices::all_in_range_p): Likewise.
10885 (vec_perm_indices_to_tree): Likewise.
10886 (vec_perm_indices_to_rtx): Likewise.
10887 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10888 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10889 (tree_vector_builder::new_binary_operation): Handle polynomial
10890 element counts. Return false if we need to know the number
10891 of elements at compile time.
10892 * fold-const.c (fold_vec_perm): Punt if the number of elements
10893 isn't known at compile time.
10895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10897 * vec-perm-indices.h (vec_perm_builder): Change element type
10898 from HOST_WIDE_INT to poly_int64.
10899 (vec_perm_indices::element_type): Update accordingly.
10900 (vec_perm_indices::clamp): Handle polynomial element_types.
10901 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10902 (vec_perm_indices::all_in_range_p): Likewise.
10903 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10905 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10906 polynomial vec_perm_indices element types.
10907 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10908 * fold-const.c (fold_vec_perm): Likewise.
10909 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10910 * tree-vect-generic.c (lower_vec_perm): Likewise.
10911 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10912 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10913 element type to HOST_WIDE_INT.
10915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10916 Alan Hayward <alan.hayward@arm.com>
10917 David Sherwood <david.sherwood@arm.com>
10919 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10920 rather than an int. Use plus_constant.
10921 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10922 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10925 Alan Hayward <alan.hayward@arm.com>
10926 David Sherwood <david.sherwood@arm.com>
10928 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10929 a HOST_WIDE_INT to a poly_int64.
10931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10932 Alan Hayward <alan.hayward@arm.com>
10933 David Sherwood <david.sherwood@arm.com>
10935 * calls.c (load_register_parameters): Cope with polynomial
10936 mode sizes. Require a constant size for BLKmode parameters
10937 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10938 forces a parameter to be padded at the lsb end in order to
10939 fill a complete number of words, require the parameter size
10940 to be ordered wrt UNITS_PER_WORD.
10942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10943 Alan Hayward <alan.hayward@arm.com>
10944 David Sherwood <david.sherwood@arm.com>
10946 * reload1.c (spill_stack_slot_width): Change element type
10947 from unsigned int to poly_uint64_pod.
10948 (alter_reg): Treat mode sizes as polynomial.
10950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10951 Alan Hayward <alan.hayward@arm.com>
10952 David Sherwood <david.sherwood@arm.com>
10954 * reload.c (complex_word_subreg_p): New function.
10955 (reload_inner_reg_of_subreg, push_reload): Use it.
10957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10958 Alan Hayward <alan.hayward@arm.com>
10959 David Sherwood <david.sherwood@arm.com>
10961 * lra-constraints.c (process_alt_operands): Reject matched
10962 operands whose sizes aren't ordered.
10963 (match_reload): Refer to this check here.
10965 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10966 Alan Hayward <alan.hayward@arm.com>
10967 David Sherwood <david.sherwood@arm.com>
10969 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10970 that the mode size is in the set {1, 2, 4, 8, 16}.
10972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10973 Alan Hayward <alan.hayward@arm.com>
10974 David Sherwood <david.sherwood@arm.com>
10976 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10977 Use plus_constant instead of gen_rtx_PLUS.
10979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10980 Alan Hayward <alan.hayward@arm.com>
10981 David Sherwood <david.sherwood@arm.com>
10983 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10984 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10985 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10986 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10987 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10988 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10989 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10990 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10991 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10992 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10994 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10995 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10996 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10997 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10998 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10999 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11000 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11001 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11002 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11003 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11005 * expr.c (emit_move_resolve_push): Treat the input and result
11006 of PUSH_ROUNDING as a poly_int64.
11007 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11008 (emit_push_insn): Likewise.
11009 * lra-eliminations.c (mark_not_eliminable): Likewise.
11010 * recog.c (push_operand): Likewise.
11011 * reload1.c (elimination_effects): Likewise.
11012 * rtlanal.c (nonzero_bits1): Likewise.
11013 * calls.c (store_one_arg): Likewise. Require the padding to be
11014 known at compile time.
11016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11017 Alan Hayward <alan.hayward@arm.com>
11018 David Sherwood <david.sherwood@arm.com>
11020 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11021 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11028 rather than an int.
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 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11035 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11036 via stack temporaries. Treat the mode size as polynomial too.
11038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11039 Alan Hayward <alan.hayward@arm.com>
11040 David Sherwood <david.sherwood@arm.com>
11042 * expr.c (expand_expr_real_2): When handling conversions involving
11043 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11044 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11045 as a poly_uint64 too.
11047 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11048 Alan Hayward <alan.hayward@arm.com>
11049 David Sherwood <david.sherwood@arm.com>
11051 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11054 Alan Hayward <alan.hayward@arm.com>
11055 David Sherwood <david.sherwood@arm.com>
11057 * combine.c (can_change_dest_mode): Handle polynomial
11058 REGMODE_NATURAL_SIZE.
11059 * expmed.c (store_bit_field_1): Likewise.
11060 * expr.c (store_constructor): Likewise.
11061 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11062 and polynomial REGMODE_NATURAL_SIZE.
11063 (gen_lowpart_common): Likewise.
11064 * reginfo.c (record_subregs_of_mode): Likewise.
11065 * rtlanal.c (read_modify_subreg_p): Likewise.
11067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11068 Alan Hayward <alan.hayward@arm.com>
11069 David Sherwood <david.sherwood@arm.com>
11071 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11072 numbers of elements.
11074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11075 Alan Hayward <alan.hayward@arm.com>
11076 David Sherwood <david.sherwood@arm.com>
11078 * match.pd: Cope with polynomial numbers of vector elements.
11080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11081 Alan Hayward <alan.hayward@arm.com>
11082 David Sherwood <david.sherwood@arm.com>
11084 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11085 in a POINTER_PLUS_EXPR.
11087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11088 Alan Hayward <alan.hayward@arm.com>
11089 David Sherwood <david.sherwood@arm.com>
11091 * omp-simd-clone.c (simd_clone_subparts): New function.
11092 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11093 (ipa_simd_modify_function_body): Likewise.
11095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11096 Alan Hayward <alan.hayward@arm.com>
11097 David Sherwood <david.sherwood@arm.com>
11099 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11100 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11101 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11102 (expand_vector_condition, vector_element): Likewise.
11103 (subparts_gt): New function.
11104 (get_compute_type): Use subparts_gt.
11105 (count_type_subparts): Delete.
11106 (expand_vector_operations_1): Use subparts_gt instead of
11107 count_type_subparts.
11109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11110 Alan Hayward <alan.hayward@arm.com>
11111 David Sherwood <david.sherwood@arm.com>
11113 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11114 (vect_compile_time_alias): ...this new function. Do the calculation
11115 on poly_ints rather than trees.
11116 (vect_prune_runtime_alias_test_list): Update call accordingly.
11118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11119 Alan Hayward <alan.hayward@arm.com>
11120 David Sherwood <david.sherwood@arm.com>
11122 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11124 (vect_schedule_slp_instance): Likewise.
11126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11127 Alan Hayward <alan.hayward@arm.com>
11128 David Sherwood <david.sherwood@arm.com>
11130 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11131 constant and extern definitions for variable-length vectors.
11132 (vect_get_constant_vectors): Note that the number of units
11133 is known to be constant.
11135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11136 Alan Hayward <alan.hayward@arm.com>
11137 David Sherwood <david.sherwood@arm.com>
11139 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11140 of units as polynomial. Choose between WIDE and NARROW based
11143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11144 Alan Hayward <alan.hayward@arm.com>
11145 David Sherwood <david.sherwood@arm.com>
11147 * tree-vect-stmts.c (simd_clone_subparts): New function.
11148 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11151 Alan Hayward <alan.hayward@arm.com>
11152 David Sherwood <david.sherwood@arm.com>
11154 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11155 vectors as polynomial. Use build_index_vector for
11156 IFN_GOMP_SIMD_LANE.
11158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11159 Alan Hayward <alan.hayward@arm.com>
11160 David Sherwood <david.sherwood@arm.com>
11162 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11163 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11164 for variable-length vectors.
11165 (vectorizable_mask_load_store): Treat the number of units as
11166 polynomial, asserting that it is constant if the condition has
11167 already been enforced.
11168 (vectorizable_store, vectorizable_load): Likewise.
11170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11171 Alan Hayward <alan.hayward@arm.com>
11172 David Sherwood <david.sherwood@arm.com>
11174 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11175 of units as polynomial. Punt if we can't tell at compile time
11176 which vector contains the final result.
11178 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11179 Alan Hayward <alan.hayward@arm.com>
11180 David Sherwood <david.sherwood@arm.com>
11182 * tree-vect-loop.c (vectorizable_induction): Treat the number
11183 of units as polynomial. Punt on SLP inductions. Use an integer
11184 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11185 cast of such a series for variable-length floating-point
11188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11189 Alan Hayward <alan.hayward@arm.com>
11190 David Sherwood <david.sherwood@arm.com>
11192 * tree.h (build_index_vector): Declare.
11193 * tree.c (build_index_vector): New function.
11194 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11195 of units as polynomial, forcibly converting it to a constant if
11196 vectorizable_reduction has already enforced the condition.
11197 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11198 to create a {1,2,3,...} vector.
11199 (vectorizable_reduction): Treat the number of units as polynomial.
11200 Choose vectype_in based on the largest scalar element size rather
11201 than the smallest number of units. Enforce the restrictions
11204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11205 Alan Hayward <alan.hayward@arm.com>
11206 David Sherwood <david.sherwood@arm.com>
11208 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11209 number of units as polynomial.
11211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11212 Alan Hayward <alan.hayward@arm.com>
11213 David Sherwood <david.sherwood@arm.com>
11215 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11216 * target.def (autovectorize_vector_sizes): Return the vector sizes
11217 by pointer, using vector_sizes rather than a bitmask.
11218 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11219 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11220 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11222 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11223 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11224 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11225 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11226 * omp-general.c (omp_max_vf): Likewise.
11227 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11228 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11229 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11230 * tree-vect-slp.c (vect_slp_bb): Likewise.
11231 * doc/tm.texi: Regenerate.
11232 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11234 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11235 the vector size as a poly_uint64 rather than an unsigned int.
11236 (current_vector_size): Change from an unsigned int to a poly_uint64.
11237 (get_vectype_for_scalar_type): Update accordingly.
11238 * tree.h (build_truth_vector_type): Take the size and number of
11239 units as a poly_uint64 rather than an unsigned int.
11240 (build_vector_type): Add a temporary overload that takes
11241 the number of units as a poly_uint64 rather than an unsigned int.
11242 * tree.c (make_vector_type): Likewise.
11243 (build_truth_vector_type): Take the number of units as a poly_uint64
11244 rather than an unsigned int.
11246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11247 Alan Hayward <alan.hayward@arm.com>
11248 David Sherwood <david.sherwood@arm.com>
11250 * target.def (get_mask_mode): Take the number of units and length
11251 as poly_uint64s rather than unsigned ints.
11252 * targhooks.h (default_get_mask_mode): Update accordingly.
11253 * targhooks.c (default_get_mask_mode): Likewise.
11254 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11255 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11262 * omp-general.c (omp_max_vf): Likewise.
11263 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11264 (expand_omp_simd): Handle polynomial safelen.
11265 * omp-low.c (omplow_simd_context): Add a default constructor.
11266 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11267 (lower_rec_simd_input_clauses): Update accordingly.
11268 (lower_rec_input_clauses): Likewise.
11270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11271 Alan Hayward <alan.hayward@arm.com>
11272 David Sherwood <david.sherwood@arm.com>
11274 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11275 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11276 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11277 (vect_analyze_slp_cost): Likewise.
11278 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11279 (vect_model_load_cost): Likewise.
11281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11282 Alan Hayward <alan.hayward@arm.com>
11283 David Sherwood <david.sherwood@arm.com>
11285 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11286 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11287 from an unsigned int * to a poly_uint64_pod *.
11288 (calculate_unrolling_factor): New function.
11289 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11292 Alan Hayward <alan.hayward@arm.com>
11293 David Sherwood <david.sherwood@arm.com>
11295 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11296 from an unsigned int to a poly_uint64.
11297 (_loop_vec_info::slp_unrolling_factor): Likewise.
11298 (_loop_vec_info::vectorization_factor): Change from an int
11300 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11301 (vect_get_num_vectors): New function.
11302 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11303 (vect_get_num_copies): Use vect_get_num_vectors.
11304 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11305 to an unsigned int *.
11306 (vect_analyze_data_refs): Change min_vf from an int * to a
11308 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11309 than an unsigned HOST_WIDE_INT.
11310 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11311 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11312 to an unsigned int *.
11313 (vect_analyze_data_ref_dependences): Likewise.
11314 (vect_compute_data_ref_alignment): Handle polynomial vf.
11315 (vect_enhance_data_refs_alignment): Likewise.
11316 (vect_prune_runtime_alias_test_list): Likewise.
11317 (vect_shift_permute_load_chain): Likewise.
11318 (vect_supportable_dr_alignment): Likewise.
11319 (dependence_distance_ge_vf): Take the vectorization factor as a
11320 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11321 (vect_analyze_data_refs): Change min_vf from an int * to a
11323 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11324 vfm1 as a poly_uint64 rather than an int. Make the same change
11325 for the returned bound_scalar.
11326 (vect_gen_vector_loop_niters): Handle polynomial vf.
11327 (vect_do_peeling): Likewise. Update call to
11328 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11329 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11331 * tree-vect-loop.c (vect_determine_vectorization_factor)
11332 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11333 (vect_get_known_peeling_cost): Likewise.
11334 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11335 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11336 (vect_transform_loop): Likewise. Use the lowest possible VF when
11337 updating the upper bounds of the loop.
11338 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11339 rather than an int.
11340 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11341 polynomial unroll factors.
11342 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11343 (vect_make_slp_decision): Likewise.
11344 (vect_supported_load_permutation_p): Likewise, and polynomial
11346 (vect_analyze_slp_cost): Handle polynomial vf.
11347 (vect_slp_analyze_node_operations): Likewise.
11348 (vect_slp_analyze_bb_1): Likewise.
11349 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11350 than an unsigned HOST_WIDE_INT.
11351 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11352 (vectorizable_load): Handle polynomial vf.
11353 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11355 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11358 Alan Hayward <alan.hayward@arm.com>
11359 David Sherwood <david.sherwood@arm.com>
11361 * match.pd: Handle bit operations involving three constants
11362 and try to fold one pair.
11364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11366 * tree-vect-loop-manip.c: Include gimple-fold.h.
11367 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11368 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11369 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11370 Add a path that uses a step of VF instead of 1, but disable it
11372 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11373 and niters_no_overflow parameters. Update calls to
11374 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11375 Create a new SSA name if the latter choses to use a ste other
11376 than zero, and return it via niters_vector_mult_vf_var.
11377 * tree-vect-loop.c (vect_transform_loop): Update calls to
11378 vect_do_peeling, vect_gen_vector_loop_niters and
11379 slpeel_make_loop_iterate_ntimes.
11380 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11381 (vect_gen_vector_loop_niters): Update declarations after above changes.
11383 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11385 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11386 128-bit round to integer instructions.
11387 (ceil<mode>2): Likewise.
11388 (btrunc<mode>2): Likewise.
11389 (round<mode>2): Likewise.
11391 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11393 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11394 unaligned VSX load/store on P8/P9.
11395 (expand_block_clear): Allow the use of unaligned VSX
11396 load/store on P8/P9.
11398 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11400 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11402 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11403 swap associated with both a load and a store.
11405 2018-01-02 Andrew Waterman <andrew@sifive.com>
11407 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11408 * config/riscv/riscv.md (clear_cache): Use it.
11410 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11412 * web.c: Remove out-of-date comment.
11414 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11416 * expr.c (fixup_args_size_notes): Check that any existing
11417 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11418 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11419 (emit_single_push_insn): ...here.
11421 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11423 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11424 (const_vector_encoded_nelts): New function.
11425 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11426 (const_vector_int_elt, const_vector_elt): Declare.
11427 * emit-rtl.c (const_vector_int_elt_1): New function.
11428 (const_vector_elt): Likewise.
11429 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11430 of CONST_VECTOR_ELT.
11432 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11434 * expr.c: Include rtx-vector-builder.h.
11435 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11436 directly on the tree encoding.
11437 (const_vector_from_tree): Likewise.
11438 * optabs.c: Include rtx-vector-builder.h.
11439 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11440 sequence of "u" values.
11441 * vec-perm-indices.c: Include rtx-vector-builder.h.
11442 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11443 directly on the vec_perm_indices encoding.
11445 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11447 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11448 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11449 * rtx-vector-builder.h: New file.
11450 * rtx-vector-builder.c: Likewise.
11451 * rtl.h (rtx_def::u2): Add a const_vector field.
11452 (CONST_VECTOR_NPATTERNS): New macro.
11453 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11454 (CONST_VECTOR_DUPLICATE_P): Likewise.
11455 (CONST_VECTOR_STEPPED_P): Likewise.
11456 (CONST_VECTOR_ENCODED_ELT): Likewise.
11457 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11458 (unwrap_const_vec_duplicate): Likewise.
11459 (const_vec_series_p): Check for a non-duplicated vector encoding.
11460 Say that the function only returns true for integer vectors.
11461 * emit-rtl.c: Include rtx-vector-builder.h.
11462 (gen_const_vec_duplicate_1): Delete.
11463 (gen_const_vector): Call gen_const_vec_duplicate instead of
11464 gen_const_vec_duplicate_1.
11465 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11466 (gen_const_vec_duplicate): Use rtx_vector_builder.
11467 (gen_const_vec_series): Likewise.
11468 (gen_rtx_CONST_VECTOR): Likewise.
11469 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11470 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11471 Build a new vector rather than modifying a CONST_VECTOR in-place.
11472 (handle_special_swappables): Update call accordingly.
11473 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11474 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11475 Build a new vector rather than modifying a CONST_VECTOR in-place.
11476 (handle_special_swappables): Update call accordingly.
11478 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11480 * simplify-rtx.c (simplify_const_binary_operation): Use
11481 CONST_VECTOR_ELT instead of XVECEXP.
11483 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11485 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11486 the selector elements to be different from the data elements
11487 if the selector is a VECTOR_CST.
11488 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11489 ssizetype for the selector.
11491 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11493 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11494 before testing each element individually.
11495 * tree-vect-generic.c (lower_vec_perm): Likewise.
11497 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11499 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11500 * selftest-run-tests.c (selftest::run_tests): Call it.
11501 * vector-builder.h (vector_builder::operator ==): New function.
11502 (vector_builder::operator !=): Likewise.
11503 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11504 (vec_perm_indices::all_from_input_p): New function.
11505 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11506 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11507 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11508 instead of reading the VECTOR_CST directly. Detect whether both
11509 vector inputs are the same before constructing the vec_perm_indices,
11510 and update the number of inputs argument accordingly. Use the
11511 utility functions added above. Only construct sel2 if we need to.
11513 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11515 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11516 the broadcast of the low byte.
11517 (expand_mult_highpart): Use an explicit encoding for the permutes.
11518 * optabs-query.c (can_mult_highpart_p): Likewise.
11519 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11520 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11521 (vectorizable_bswap): Likewise.
11522 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11523 explicit encoding for the power-of-2 permutes.
11524 (vect_permute_store_chain): Likewise.
11525 (vect_grouped_load_supported): Likewise.
11526 (vect_permute_load_chain): Likewise.
11528 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11530 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11531 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11532 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11533 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11534 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11535 (vect_gen_perm_mask_any): Likewise.
11537 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11539 * int-vector-builder.h: New file.
11540 * vec-perm-indices.h: Include int-vector-builder.h.
11541 (vec_perm_indices): Redefine as an int_vector_builder.
11542 (auto_vec_perm_indices): Delete.
11543 (vec_perm_builder): Redefine as a stand-alone class.
11544 (vec_perm_indices::vec_perm_indices): New function.
11545 (vec_perm_indices::clamp): Likewise.
11546 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11547 (vec_perm_indices::new_vector): New function.
11548 (vec_perm_indices::new_expanded_vector): Update for new
11549 vec_perm_indices class.
11550 (vec_perm_indices::rotate_inputs): New function.
11551 (vec_perm_indices::all_in_range_p): Operate directly on the
11552 encoded form, without computing elided elements.
11553 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11554 encoding. Update for new vec_perm_indices class.
11555 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11556 the given vec_perm_builder.
11557 (expand_vec_perm_var): Update vec_perm_builder constructor.
11558 (expand_mult_highpart): Use vec_perm_builder instead of
11559 auto_vec_perm_indices.
11560 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11561 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11562 or double series encoding as appropriate.
11563 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11564 vec_perm_indices instead of auto_vec_perm_indices.
11565 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11566 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11567 (vect_permute_store_chain): Likewise.
11568 (vect_grouped_load_supported): Likewise.
11569 (vect_permute_load_chain): Likewise.
11570 (vect_shift_permute_load_chain): Likewise.
11571 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11572 (vect_transform_slp_perm_load): Likewise.
11573 (vect_schedule_slp_instance): Likewise.
11574 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11575 (vectorizable_mask_load_store): Likewise.
11576 (vectorizable_bswap): Likewise.
11577 (vectorizable_store): Likewise.
11578 (vectorizable_load): Likewise.
11579 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11580 vec_perm_indices instead of auto_vec_perm_indices. Use
11581 tree_to_vec_perm_builder to read the vector from a tree.
11582 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11583 vec_perm_builder instead of a vec_perm_indices.
11584 (have_whole_vector_shift): Use vec_perm_builder and
11585 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11586 truncation to calc_vec_perm_mask_for_shift.
11587 (vect_create_epilog_for_reduction): Likewise.
11588 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11589 from auto_vec_perm_indices to vec_perm_indices.
11590 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11591 instead of changing individual elements.
11592 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11593 the vector in d.perm.
11594 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11595 from auto_vec_perm_indices to vec_perm_indices.
11596 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11597 instead of changing individual elements.
11598 (arm_vectorize_vec_perm_const): Use new_vector to install
11599 the vector in d.perm.
11600 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11601 Update vec_perm_builder constructor.
11602 (rs6000_expand_interleave): Likewise.
11603 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11604 (rs6000_expand_interleave): Likewise.
11606 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11608 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11609 to qimode could truncate the indices.
11610 * optabs.c (expand_vec_perm_var): Likewise.
11612 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11614 * Makefile.in (OBJS): Add vec-perm-indices.o.
11615 * vec-perm-indices.h: New file.
11616 * vec-perm-indices.c: Likewise.
11617 * target.h (vec_perm_indices): Replace with a forward class
11619 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11620 * optabs.h: Include vec-perm-indices.h.
11621 (expand_vec_perm): Delete.
11622 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11623 (expand_vec_perm_const): Declare.
11624 * target.def (vec_perm_const_ok): Replace with...
11625 (vec_perm_const): ...this new hook.
11626 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11627 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11628 * doc/tm.texi: Regenerate.
11629 * optabs.def (vec_perm_const): Delete.
11630 * doc/md.texi (vec_perm_const): Likewise.
11631 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11632 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11633 expand_vec_perm for constant permutation vectors. Assert that
11634 the mode of variable permutation vectors is the integer equivalent
11635 of the mode that is being permuted.
11636 * optabs-query.h (selector_fits_mode_p): Declare.
11637 * optabs-query.c: Include vec-perm-indices.h.
11638 (selector_fits_mode_p): New function.
11639 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11640 is defined, instead of checking whether the vec_perm_const_optab
11641 exists. Use targetm.vectorize.vec_perm_const instead of
11642 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11643 fit in the vector mode before using a variable permute.
11644 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11645 vec_perm_indices instead of an rtx.
11646 (expand_vec_perm): Replace with...
11647 (expand_vec_perm_const): ...this new function. Take the selector
11648 as a vec_perm_indices rather than an rtx. Also take the mode of
11649 the selector. Update call to shift_amt_for_vec_perm_mask.
11650 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11651 Use vec_perm_indices::new_expanded_vector to expand the original
11652 selector into bytes. Check whether the indices fit in the vector
11653 mode before using a variable permute.
11654 (expand_vec_perm_var): Make global.
11655 (expand_mult_highpart): Use expand_vec_perm_const.
11656 * fold-const.c: Includes vec-perm-indices.h.
11657 * tree-ssa-forwprop.c: Likewise.
11658 * tree-vect-data-refs.c: Likewise.
11659 * tree-vect-generic.c: Likewise.
11660 * tree-vect-loop.c: Likewise.
11661 * tree-vect-slp.c: Likewise.
11662 * tree-vect-stmts.c: Likewise.
11663 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11665 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11666 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11667 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11668 (aarch64_vectorize_vec_perm_const): ...this new function.
11669 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11670 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11671 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11672 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11673 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11674 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11675 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11677 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11678 check for NEON modes.
11679 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11680 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11681 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11682 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11684 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11685 the old VEC_PERM_CONST conditions.
11686 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11687 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11688 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11689 (ia64_vectorize_vec_perm_const_ok): Merge into...
11690 (ia64_vectorize_vec_perm_const): ...this new function.
11691 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11692 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11693 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11694 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11695 * config/mips/mips.c (mips_expand_vec_perm_const)
11696 (mips_vectorize_vec_perm_const_ok): Merge into...
11697 (mips_vectorize_vec_perm_const): ...this new function.
11698 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11699 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11700 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11701 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11702 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11703 (rs6000_expand_vec_perm_const): Delete.
11704 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11706 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11707 (altivec_expand_vec_perm_const_le): Take each operand individually.
11708 Operate on constant selectors rather than rtxes.
11709 (altivec_expand_vec_perm_const): Likewise. Update call to
11710 altivec_expand_vec_perm_const_le.
11711 (rs6000_expand_vec_perm_const): Delete.
11712 (rs6000_vectorize_vec_perm_const_ok): Delete.
11713 (rs6000_vectorize_vec_perm_const): New function.
11714 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11715 an element count and rtx array.
11716 (rs6000_expand_extract_even): Update call accordingly.
11717 (rs6000_expand_interleave): Likewise.
11718 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11719 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11720 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11721 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11722 (rs6000_expand_vec_perm_const): Delete.
11723 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11724 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11725 (altivec_expand_vec_perm_const_le): Take each operand individually.
11726 Operate on constant selectors rather than rtxes.
11727 (altivec_expand_vec_perm_const): Likewise. Update call to
11728 altivec_expand_vec_perm_const_le.
11729 (rs6000_expand_vec_perm_const): Delete.
11730 (rs6000_vectorize_vec_perm_const_ok): Delete.
11731 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11732 reference to the SPE evmerge intructions.
11733 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11734 an element count and rtx array.
11735 (rs6000_expand_extract_even): Update call accordingly.
11736 (rs6000_expand_interleave): Likewise.
11737 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11738 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11740 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11742 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11744 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11745 vector mode and that that mode matches the mode of the data
11747 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11748 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11749 directly using expand_vec_perm_1 when forcing selectors into
11751 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11753 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11755 * optabs-query.h (can_vec_perm_p): Delete.
11756 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11757 * optabs-query.c (can_vec_perm_p): Split into...
11758 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11759 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11760 particular selector is valid.
11761 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11762 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11763 (vect_grouped_load_supported): Likewise.
11764 (vect_shift_permute_load_chain): Likewise.
11765 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11766 (vect_transform_slp_perm_load): Likewise.
11767 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11768 (vectorizable_bswap): Likewise.
11769 (vect_gen_perm_mask_checked): Likewise.
11770 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11771 implementations of variable permutation vectors into account
11772 when deciding which selector to use.
11773 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11774 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11775 with a false third argument.
11776 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11777 to test whether the constant selector is valid and can_vec_perm_var_p
11778 to test whether a variable selector is valid.
11780 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11782 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11783 * optabs-query.c (can_vec_perm_p): Likewise.
11784 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11785 instead of vec_perm_indices.
11786 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11787 (vect_gen_perm_mask_checked): Likewise,
11788 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11789 (vect_gen_perm_mask_checked): Likewise,
11791 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11793 * optabs-query.h (qimode_for_vec_perm): Declare.
11794 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11795 (qimode_for_vec_perm): ...this new function.
11796 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11798 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11800 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11801 does not have a conditional at the top.
11803 2018-01-02 Richard Biener <rguenther@suse.de>
11805 * ipa-inline.c (big_speedup_p): Fix expression.
11807 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11810 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11813 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11817 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11818 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11819 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11820 cond_taken_branch_cost 3->4.
11822 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11824 PR tree-optimization/83581
11825 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11826 TODO_cleanup_cfg if any changes have been made.
11828 PR middle-end/83608
11829 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11830 convert_modes if target mode has the right side, but different mode
11833 PR middle-end/83609
11834 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11835 last argument when extracting from CONCAT. If either from_real or
11836 from_imag is NULL, use expansion through memory. If result is not
11837 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11838 the parts directly to inner mode, if even that fails, use expansion
11841 PR middle-end/83623
11842 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11843 check for bswap in mode rather than HImode and use that in expand_unop
11846 Copyright (C) 2018 Free Software Foundation, Inc.
11848 Copying and distribution of this file, with or without modification,
11849 are permitted in any medium without royalty provided the copyright
11850 notice and this notice are preserved.