1 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
4 --push-state --as-needed and --pop-state instead of --as-needed and
5 --no-as-needed if ld supports it.
6 * configure: Regenerated.
9 * doc/install.texi2html: Replace _002d with - and _002a with * in
10 generated html files using sed.
12 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15 * gcc-rich-location.c (blank_line_before_p): New function.
16 (use_new_line): New function.
17 (gcc_rich_location::add_fixit_insert_formatted): New function.
19 (gcc_rich_location::add_fixit_insert_formatted): New function.
21 2018-04-30 David Malcolm <dmalcolm@redhat.com>
23 * selftest.c (assert_streq): Rename "expected" and "actual" to
24 "val1" and "val2". Extend NULL-handling to cover both inputs
25 symmetrically, while still requiring both to be non-NULL for a pass.
26 * selftest.h (assert_streq): Rename "expected" and "actual" to
28 (ASSERT_EQ): Likewise.
29 (ASSERT_EQ_AT): Likewise.
30 (ASSERT_KNOWN_EQ): Likewise.
31 (ASSERT_KNOWN_EQ_AT): Likewise.
32 (ASSERT_NE): Likewise.
33 (ASSERT_MAYBE_NE): Likewise.
34 (ASSERT_MAYBE_NE_AT): Likewise.
35 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
36 the assertion to pass.
37 (ASSERT_STREQ_AT): Likewise.
39 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
41 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
42 interaction with -pie.
44 2018-04-30 David Malcolm <dmalcolm@redhat.com>
46 * selftest.h: Fix alphabetization of per-source-file selftest
49 2018-04-30 Jason Merrill <jason@redhat.com>
51 PR c++/61982 - dead stores to destroyed objects.
52 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
55 2018-04-30 Jason Merrill <jason@redhat.com>
57 * tree.c (build_clobber): New.
59 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
61 2018-04-30 David Malcolm <dmalcolm@redhat.com>
63 * diagnostic-show-locus.c (layout::layout): Update for
64 location_get_source_line returning a char_span.
65 (struct char_span): Move to input.h.
66 (struct correction): Update for fields in char_span becoming
68 (struct source_line): Update for location_get_source_line
69 returning a char_span.
70 (layout::print_line): Likewise.
71 * edit-context.c (edited_file::print_content): Likewise.
72 (edited_file::print_diff_hunk): Likewise.
73 (edited_file::print_run_of_changed_lines): Likewise.
74 (edited_file::get_num_lines): Likewise.
75 (edited_line::edited_line): Likewise.
76 * final.c (asm_show_source): Likewise.
77 * input.c (location_get_source_line): Convert return type
78 from const char * to char_span, losing the final "line_len"
80 (dump_location_info): Update for the above.
81 (get_substring_ranges_for_loc): Likewise. Use a char_span
82 when handling the literal within the line.
83 (test_reading_source_line): Update for location_get_source_line
84 returning a char_span.
85 * input.h (class char_span): Move here from
86 diagnostic-show-locus.c, converting from a struct to a class.
87 Make data members private.
88 (char_span::operator bool): New.
89 (char_span::length): New.
90 (char_span::get_buffer): New.
91 (char_span::operator[]): New.
92 (char_span::subspan): Make const.
93 (char_span::xstrdup): New.
94 (location_get_source_line): Convert return type from const char *
95 to char_span, losing the final "line_size" param.
97 2018-04-30 Jan Hubicka <jh@suse.cz>
99 * lto-wrapper.c (ltrans_priorities): New static var.
101 (run_gcc): Read priorities and if doing parallel build order
102 the Makefile by them.
104 2018-04-30 David Malcolm <dmalcolm@redhat.com>
106 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
108 2018-04-30 Richard Biener <rguenther@suse.de>
110 * tree-cfg.c (verify_address): Remove base argument, add
111 flag whether to check TREE_ADDRESSABLE and do that.
112 (verify_expr): Remove.
113 (verify_types_in_gimple_reference): Add pieces from verify_expr.
114 (verify_gimple_assign_single): Likewise.
115 (verify_gimple_switch): Likewise.
116 (verify_expr_location_1): Dereference tp once. Add (disabled)
117 piece from verify_expr.
118 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
120 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
122 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
124 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
126 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
127 (small_data_pattern): Likewise.
128 (arc_rewrite_small_data): Likewise.
129 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
130 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
131 (get_symbol_alignment): New function.
132 (legitimate_small_data_address_p): Likewise.
133 (legitimate_scaled_address): Update, call
134 legitimate_small_data_address_p.
135 (output_sdata): New static variable.
136 (arc_print_operand): Update how we handle small data operands.
137 (arc_print_operand_address): Likewise.
138 (arc_legitimate_address_p): Update, use
139 legitimate_small_data_address_p.
140 (arc_rewrite_small_data_p): Remove.
141 (arc_rewrite_small_data_1): Likewise.
142 (arc_rewrite_small_data): Likewise.
143 (small_data_pattern): Likewise.
144 (compact_sda_memory_operand): Update to use
145 legitimate_small_data_address_p and get_symbol_alignment.
146 (prepare_move_operands): Don't rewite sdata pattern.
147 (prepare_extend_operands): Remove.
148 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
150 (zero_extendqisi2): Likewise.
151 (zero_extendhisi2): Likewise.
152 (extendqihi2): Likewise.
153 (extendqisi2): Likewise.
154 (extendhisi2): Likewise.
158 * config/arc/constraints.md (Usd): Change it to memory constraint.
160 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
162 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
163 as source of std instructions.
164 * config/arc/arc.md (movsi_insn): Update pattern predicate to
165 allow 6-bit constants as source for store instructions.
166 (movdi_insn): Update instruction pattern to allow 6-bit constants
167 as source for store instructions.
169 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
171 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
173 2018-04-30 Nathan Sidwell <nathan@acm.org>
174 Sandra Loosemore <sandra@codesourcery.com>
176 * dumpfile.c (dump_open): Allow '-' for stdout.
177 * doc/invoke.texi (Developer Options): Document dump filename
178 determination early. Document stdin/stdout selection.
180 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
182 Microblaze Target: PIC data text relative
184 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
185 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
187 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
188 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
189 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
190 New addressing mode for data-text relative position indepenedent code.
191 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
192 'ADDRESS_SYMBOLIC_TXT_REL'.
193 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
194 (microblaze_legitimate_pic_operand): Exclude function calls from
195 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
196 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
198 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
199 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
200 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
201 for 'address + offset'.
202 (microblaze_expand_prologue): Add new function prologue call for
204 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
205 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
206 table in case of TARGET_PIC_DATA_TEXT_REL.
207 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
208 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
209 Add new macros 'UNSPEC_TEXT',
210 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
211 + exclude function calls from 'UNSPEC_PLT' in case of data text
213 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
214 new target hook for generating address diff vector tables in case of
216 * doc/tm.texi : Regenerate.
217 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
218 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
219 of addr diff vector generation.
220 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
221 target hook definition.
222 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
223 Add default function for generate_pic_addr_diff_vec -> flag_pic.
224 * doc/invoke.texi (Add new pic option): Add new microblaze pic
225 option for data text relative.
227 2018-04-30 Richard Biener <rguenther@suse.de>
229 * tree-chrec.h (evolution_function_is_constant_p): Remove
231 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
233 2018-04-30 Richard Biener <rguenther@suse.de>
236 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
238 2018-04-30 Richard Biener <rguenther@suse.de>
240 PR tree-optimization/28364
241 PR tree-optimization/85275
242 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
243 copying first exit test.
245 2018-04-28 Mark Wielaard <mark@klomp.org>
247 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
249 (dwarf_AT): Handle DW_AT_addr_base.
250 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
252 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
255 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
256 (*ashl<dwi>3_doubleword_mask_1): Ditto.
257 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
258 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
260 2018-04-28 Richard Biener <rguenther@suse.de>
262 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
263 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
264 to reflect use. Only add interesting stmts.
266 2018-04-27 Martin Jambor <mjambor@suse.cz>
269 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
270 the jump function allows for passing through aggregate values.
272 2018-04-27 David Malcolm <dmalcolm@redhat.com>
274 * input.h (in_system_header_at): Convert from macro to inline
276 (from_macro_expansion_at): Likewise.
277 (from_macro_definition_at): Likewise.
279 2018-04-27 Jeff Law <law@redhat.com>
281 * config.gcc: Mark tile* targets as deprecated/obsolete.
283 2018-04-27 Richard Biener <rguenther@suse.de>
285 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
288 2018-04-27 Richard Biener <rguenther@suse.de>
290 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
292 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
294 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
295 with Yd constraint. Set "preferred_for_speed" attribute from
296 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
298 (*movdi_internal): Ditto.
299 (movti_interunit splitters): Remove
300 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
301 (movdi_interunit splitters): Ditto.
302 * config/i386/constraints.md (Ye): Remove.
303 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
305 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
308 * config/aarch64/constraints.md (Usg): Limit to 31.
311 2018-04-27 Jakub Jelinek <jakub@redhat.com>
313 PR tree-optimization/85529
314 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
315 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
316 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
317 zero extension or masking of the MSB bit.
318 (optimize_range_tests): Add FIRST_BB argument, pass it through
319 to optimize_range_tests_var_bound.
320 (maybe_optimize_range_tests, reassociate_bb): Adjust
321 optimize_range_tests callers.
323 2018-04-26 Richard Biener <rguenther@suse.de>
324 Jakub Jelinek <jakub@redhat.com>
326 * cgraph.h (symbol_table): Just declare debug method here.
327 * symtab.c (symbol_table::debug): Define.
329 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
331 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
333 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
335 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
336 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
337 (*movdi_internal): Substitute Yi and Yj constraint with x
338 and Ym and Yn constraint with y constraint. Update "isa"
339 attribute and set "preferred_for_speed" attribute from
340 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
341 (*movsi_internal): Ditto.
342 (*movdf_internal): Ditto.
343 (*movsf_internal): Ditto.
344 (*zero_extendsidi2): Ditto.
345 * config/i386/sse.md (vec_set<mode>_0): Ditto.
346 (sse2_loadld): Ditto.
347 (*vec_extract<ssevecmodelower>_0): Ditto.
348 (*vec_extractv4si_0_zext_sse4): Ditto.
349 (vec_concatv2di): Ditto.
350 (*vec_dup<mode>): Ditto.
351 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
352 * config/i386/constraints.md (Yi): Remove.
357 2018-04-26 Nathan Sidwell <nathan@acm.org>
359 * dumpfile.c (dump_open): New.
360 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
361 (dump_finish): Detect stdio/stderr by value not name.
363 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
365 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
367 2018-04-26 Tom de Vries <tom@codesourcery.com>
370 * config/nvptx/nvptx.c (verify_neutering_jumps)
371 (verify_neutering_labels): New function
372 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
374 2018-04-26 Tom de Vries <tom@codesourcery.com>
377 * config/nvptx/nvptx.c (needs_neutering_p): New function.
378 (nvptx_single): Use needs_neutering_p to skip over insns that do not
381 2018-04-26 Richard Biener <rguenther@suse.de>
382 Tom de Vries <tom@codesourcery.com>
385 * lto-streamer-out.c (output_function): Fixup loops if required to match
386 discovery done in the reader.
388 2018-04-26 Richard Biener <rguenther@suse.de>
390 PR tree-optimization/85116
391 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
392 have a loop exit from the single latch predecessor. Remove
393 case of header with just condition.
394 (ch_base::copy_headers): Exclude infinite loops from any
396 (pass_ch::execute): Record exits.
398 2018-04-26 Richard Biener <rguenther@suse.de>
400 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
401 prologue cost vector and pass it to vect_get_load_cost.
402 (vect_get_peeling_costs_all_drs): Likewise.
403 (vect_peeling_hash_get_lowest_cost): Likewise.
404 (vect_enhance_data_refs_alignment): Likewise.
406 2018-04-26 Richard Biener <rguenther@suse.de>
409 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
410 checking of integer<->pointer conversions.
411 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
412 sign-/zero-extending pointer types.
413 (expand_omp_for_static_chunk): Likewise.
415 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
416 Jean Lee <xiaoyur347@gmail.com>
418 * config/mips/mips.c (mips_asan_shadow_offset): New function.
419 (TARGET_ASAN_SHADOW_OFFSET): Define.
420 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
421 true for -fsanitize=address.
423 2018-04-25 Mark Wielaard <mark@klomp.org>
425 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
428 2018-04-25 Jakub Jelinek <jakub@redhat.com>
430 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
431 than "alu", remove explicit "memory" and "imm_disp" attributes.
432 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
435 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
436 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
439 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
442 * config/i386/i386.c (ix86_expand_builtin): Change memory
443 operand to XI, extend p0 to Pmode.
444 * config/i386/i386.md: Change unspec volatile and operand
445 1 mode to XI, change operand 0 mode to P.
447 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
449 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
450 GET_MODE_MASK before any checking.
451 (nds32_can_use_bset_p): Likewise.
452 (nds32_can_use_btgl_p): Likewise.
454 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
456 * config/nds32/nds32-doubleword.md: New define_split pattern for
457 illegal register number.
459 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
461 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
463 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
465 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
467 2018-04-25 Richard Biener <rguenther@suse.de>
469 * lto-streamer.h (LTO_major_version): Bump to 8.
471 2018-04-25 Jakub Jelinek <jakub@redhat.com>
473 * BASE-VER: Set to 9.0.0.
475 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
477 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
478 in __abskf2 and __powikf2.
480 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
483 * config/aarch64/constraints.md (Usg, Usj): New constraints.
484 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
485 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
486 Use the above on operand 2. Reindent.
487 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
489 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
492 * common/config/i386/i386-common.c (ix86_handle_option): Don't
494 * config/i386/i386.opt (mcet): Removed.
495 * doc/install.texi: Remove -mcet documentation.
496 * doc/invoke.texi: Likewise.
498 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
501 * doc/install.texi: Remove -mcet from bootstrap-cet.
503 2018-04-24 Jakub Jelinek <jakub@redhat.com>
506 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
507 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
511 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
512 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
513 containing a CONST_VECTOR.
515 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
517 * doc/install.texi: Update newlib dependency for nvptx.
519 2018-04-24 Jakub Jelinek <jakub@redhat.com>
522 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
523 instead of INTVAL when shifting x left.
525 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
527 PR tree-optimization/85478
528 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
529 vect_grouped_store_supported for single element vectors.
531 2018-04-24 Richard Biener <rguenther@suse.de>
534 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
535 load cost increase to the case of non-constant step.
537 2018-04-24 Jakub Jelinek <jakub@redhat.com>
540 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
541 destination if any_malformed_asm.
543 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
546 * expr.c (store_field): In the bitfield case, if the value comes from
547 a function call and is returned in registers by means of a PARALLEL,
548 do not change the mode of the temporary unless BLKmode and VOIDmode.
550 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
552 PR rtl-optimization/85423
553 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
554 dependencies to debug insns when the previous insn is non-debug.
556 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
558 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
559 enums into a single definition.
560 (fls): Fix predicates and printing.
563 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
565 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
566 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
567 and short u6 immediate.
568 (check_if_valid_sleep_operand): Remove.
569 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
571 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
573 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
574 flag_always_save_lp condition.
575 * config/nds32/nds32.opt (malways-save-lp): New option.
577 2018-04-22 Shiva Chen <shiva0217@gmail.com>
579 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
580 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
581 * config/nds32/nds32.h
582 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
583 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
585 2018-04-22 Shiva Chen <shiva0217@gmail.com>
587 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
588 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
590 2018-04-22 Shiva Chen <shiva0217@gmail.com>
591 Chung-Ju Wu <jasonwucj@gmail.com>
593 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
595 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
596 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
598 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
600 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
602 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
604 * config/nds32/nds32-protos.h (nds32_data_alignment,
605 nds32_local_alignment): Declare.
606 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
607 nds32_local_alignment): New functions.
608 (TARGET_CONSTANT_ALIGNMENT): Define.
609 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
611 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
613 * config/nds32/nds32.c
614 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
615 (TARGET_MODES_TIEABLE_P): Likewise.
617 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
619 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
622 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
623 Chung-Ju Wu <jasonwucj@gmail.com>
625 * config/nds32/constants.md (unspec_volatile_element): Add enum values
626 for unaligned access.
627 * config/nds32/nds32-intrinsic.c: Implementation of expanding
629 * config/nds32/nds32-intrinsic.md: Likewise.
630 * config/nds32/nds32_intrinsic.h: Likewise.
631 * config/nds32/nds32.h (nds32_builtins): Likewise.
632 * config/nds32/nds32.opt (munaligned-access): New option.
633 * config/nds32/nds32.c (nds32_asm_file_start): Display
634 flag_unaligned_access status.
636 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
638 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
639 -mno-relax is present.
640 * config/riscv/linux.h (LINK_SPEC): Ditto.
642 2018-04-20 Martin Sebor <msebor@redhat.com>
645 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
647 (gimple_fold_builtin_stxcpy_chk): Same.
648 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
650 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
653 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
654 __powikf2 when long double is IEEE 128-bit.
656 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
658 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
659 step to make sure stack always aligned.
661 2018-04-20 Carl Love <cel@us.ibm.com>
664 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
667 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
668 Tom de Vries <tom@codesourcery.com>
671 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
672 Emit insns for calls too.
673 (nvptx_find_par): Always look for worker-level predecessor insn.
674 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
676 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
677 (nvptx_process_pars): Propagate frames for calls.
679 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
682 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
684 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
685 (ix86_handle_option): Don't handle OPT_mibt.
686 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
688 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
690 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
691 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
692 (ix86_target_macros): Define __CET__ with flag_cf_protection
694 * config/i386/i386.c (isa2_opts): Remove -mibt.
695 * config/i386/i386.h (TARGET_IBT): Removed.
696 (TARGET_IBT_P): Likewise.
697 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
698 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
699 * config/i386/i386.opt (mcet): Update help message.
702 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
703 -mcet as an alias for -mshstk.
705 2018-04-20 Richard Biener <rguenther@suse.de>
708 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
709 complexity by forcing a single use of the multiply operand.
711 2018-04-20 Martin Jambor <mjambor@suse.cz>
714 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
715 recursion dependency to only apply to non-clones.
717 2018-04-20 Martin Jambor <mjambor@suse.cz>
720 * ipa-cp.c (create_specialized_node): Check that clones of
721 self-recursive edges exist during IPA-CP.
723 2018-04-19 Toon Moene <toon@moene.org>
725 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
728 2018-04-19 Jakub Jelinek <jakub@redhat.com>
730 PR tree-optimization/85467
731 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
732 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
733 VECTOR_CST element to type.
735 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
738 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
739 * config/i386/i386.md (builtin_setjmp_setup): Removed.
740 (builtin_longjmp): Likewise.
741 (save_stack_nonlocal): New pattern.
742 (restore_stack_nonlocal): Likewise.
744 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
747 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
748 Replace ASM_OUTPUT_LABEL with fprintf.
750 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
753 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
754 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
755 * config/i386/i386-c.c (ix86_target_macros_internal): Also
756 define __IBT__ and __SHSTK__ for -fcf-protection.
757 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
759 (ix86_trampoline_init): Likewise.
760 (x86_output_mi_thunk): Likewise.
761 (ix86_notrack_prefixed_insn_p): Likewise.
762 (ix86_option_override_internal): Don't disallow -fcf-protection.
763 * config/i386/i386.md (rdssp<mode>): Also enable for
765 (incssp<mode>): Likewise.
766 (nop_endbr): Likewise.
767 * config/i386/i386.opt (mcet): Change help message to built-in
771 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
772 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
773 enable CET built-in functions.
775 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
777 * common/config/i386/i386-common.c
778 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
779 OPTION_MASK_ISA_MOVDIRI_UNSET,
780 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
781 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
782 * config.gcc (movdirintrin.h): New header.
783 * config/i386/cpuid.h (bit_MOVDIRI,
784 bit_MOVDIR64B): New bits.
785 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
787 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
788 (VOID, PVOID, PCVOID)): New function types.
789 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
790 __builtin_ia32_directstoreu_u64,
791 __builtin_ia32_movdir64b): New builtins.
792 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
793 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
795 (ix86_valid_target_attribute_inner_p): Ditto.
796 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
797 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
798 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
799 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
800 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
801 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
802 (movdiri<mode>, movdir64b_<mode>): New.
803 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
804 * config/i386/immintrin.h: Include movdirintrin.h.
805 * config/i386/movdirintrin.h: New file.
806 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
808 2018-04-19 Richard Biener <rguenther@suse.de>
811 * cfg.c (clear_bb_flags): When loop state says we have
812 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
814 2018-04-19 Richard Biener <rguenther@suse.de>
816 PR tree-optimization/84737
817 * tree-vect-data-refs.c (vect_copy_ref_info): New function
818 copying restrict info.
819 (vect_setup_realignment): Use it.
820 * tree-vectorizer.h (vect_copy_ref_info): Declare.
821 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
822 the first DR to all generated stores.
823 (vectorizable_load): Likewise for loads.
825 2018-04-19 Jakub Jelinek <jakub@redhat.com>
827 PR tree-optimization/85446
828 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
829 the integral and pointer types to have the same precision.
831 * doc/install.texi: Document --disable-cet being the default and
834 2018-04-18 Martin Liska <mliska@suse.cz>
836 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
839 2018-04-18 Martin Liska <mliska@suse.cz>
842 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
845 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
846 arguments if they are comparable.
848 2018-04-18 Martin Liska <mliska@suse.cz>
851 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
854 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
857 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
860 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
861 ENDBR after calling __morestack.
863 2018-04-18 David Malcolm <dmalcolm@redhat.com>
866 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
867 by using gcc_base_ver to generate a gcc_driver_version, and use
868 it when generating GCC_DRIVER_NAME.
869 * configure: Regenerate.
871 2018-04-18 Jakub Jelinek <jakub@redhat.com>
874 * config.gcc: Obsolete powerpc*-*-*spe*.
876 2018-04-17 Jakub Jelinek <jakub@redhat.com>
879 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
880 (stabstr_D): Change type of unum from unsigned int to
881 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
884 2018-04-17 Jim Wilson <jimw@sifive.com>
887 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
888 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
889 Set arg_pointer_offset after using pretend_args_size.
891 2018-04-17 Jakub Jelinek <jakub@redhat.com>
893 PR rtl-optimization/85431
894 * dse.c (record_store): Ignore zero width stores.
897 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
898 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
899 __builtin_stack_restore rather than after it.
900 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
901 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
902 argument instead of virtual_dynamic_stack_rtx.
904 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
906 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
908 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
909 Add note to error message to explain internal mapping of overloaded
910 built-in function name to non-overloaded built-in function name.
911 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
914 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
917 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
918 where the inputs overlap with the output.
920 2018-04-17 Jakub Jelinek <jakub@redhat.com>
922 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
923 (=v, v) alternative and explicit "memory" attribute.
924 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
925 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
927 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
928 "sselog1" type instead of "sselog".
929 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
930 "sselog". Remove explicit "memory" attribute.
931 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
932 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
934 (vec_extract_hi_v32hi): Merge all alternatives into one, use
935 "sselog1" type instead of "sselog". Remove explicit "memory"
937 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
938 use "sselog1" type instead of "sselog". Remove explicit "memory"
940 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
941 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
943 (vec_extract_hi_v64qi): Merge all alternatives into one, use
944 "sselog1" type instead of "sselog". Remove explicit "memory"
946 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
947 use "sselog1" type instead of "sselog". Remove explicit "memory"
951 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
954 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
957 2018-04-17 Martin Jambor <mjambor@suse.cz>
960 * ipa-cp.c (create_specialized_node): Call
961 expand_all_artificial_thunks if necessary.
963 2018-04-17 Martin Liska <mliska@suse.cz>
966 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
967 in message, remote space in between '_G' and '('.
969 2018-04-17 Jakub Jelinek <jakub@redhat.com>
972 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
973 avx512f_vmcmp<mode>3<round_saeonly_name>,
974 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
975 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
976 avx512f_rndscale<mode><round_saeonly_name>,
977 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
978 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
979 Use %<iptr>2 instead of %2 for -masm=intel.
980 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
981 avx512f_vcvttss2usi<round_saeonly_name>,
982 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
984 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
985 avx512f_vcvttsd2usi<round_saeonly_name>,
986 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
987 Use %q1 instead of %1 for -masm=intel.
988 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
989 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
990 of %3 for -masm=intel.
991 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
993 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
995 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
997 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
998 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1000 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1002 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1003 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1004 %g1 and one with %0 and %1.
1005 (avx512er_vmrcp28<mode><round_saeonly_name>,
1006 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1008 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1009 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1010 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1011 of %0 and %{%4%} for -masm=intel.
1012 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1013 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1014 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1015 order of %0 and %{%5%}%{z%} for -masm=intel.
1017 2018-04-17 Jan Hubicka <jh@suse.cz>
1020 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1022 2018-04-17 Martin Liska <mliska@suse.cz>
1025 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1026 for target_clone error message. Make default implementation
1027 clone to be a local declaration.
1028 (separate_attrs): Add new argument and check for an empty
1030 (expand_target_clones): Handle it.
1031 (ipa_target_clone): Make redirection just for target_clones
1034 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1035 Tom de Vries <tom@codesourcery.com>
1038 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1039 tiled basic blocks without omp continue statements.
1041 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1044 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1045 vec_extract expression as having side effects to make sure it gets
1048 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1051 * config/i386/i386.c (get_builtin_code_for_version): Check
1054 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1057 * gcc/config.gcc: Support "skylake".
1058 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1060 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1061 (processor_target_table): Add "skylake".
1062 (ix86_option_override_internal): Add "skylake".
1063 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1064 PROCESSOR_CANNONLAKE.
1065 (get_builtin_code_for_version): Fix priority for
1066 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1067 PROCESSOR_SKYLAKE-AVX512.
1068 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1069 (processor_type): Add PROCESSOR_SKYLAKE.
1071 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1072 Jason Merrill <jason@redhat.com>
1075 * convert.c (convert_to_integer_1): Use direct recursion for
1076 enumeral types and types with a precision less than the number
1077 of bits in their mode.
1079 2018-04-16 Julia Koval <julia.koval@intel.com>
1082 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1083 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1085 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1088 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1089 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1090 and -mno-direct-move.
1092 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1095 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1096 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1097 Ensure negative shifts result in {0}.
1099 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1101 PR rtl-optimization/79916
1102 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1103 regs (if any) to define how to gnerate SD moves when LRA is in
1106 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1108 PR rtl-optimization/85393
1109 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1110 * except.c (expand_dw2_landing_pad_for_region): Make static.
1111 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1112 a label and unconditional jump to old_bb, rather than
1113 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1116 PR rtl-optimization/85376
1117 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1118 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1119 instead of a specific value.
1121 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1122 Bin Cheng <bin.cheng@arm.com>
1124 PR tree-optimization/82965
1125 PR tree-optimization/83991
1126 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1127 by_profile_only parameter.
1128 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1129 information if the loop was predicted to iterate too many times.
1130 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1132 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1135 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1138 2018-04-13 Martin Liska <mliska@suse.cz>
1139 Jakub Jelinek <jakub@redhat.com>
1142 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1143 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1144 * builtins.c (expand_builtin_memory_copy_args): Use
1145 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1146 handle dest_addr == pc_rtx.
1148 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1151 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1152 asked to not generate direct moves.
1153 (fix_trunc<mode>si2_stfiwx): Similar.
1154 (fix_trunc<mode>si2_internal): Similar.
1156 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1159 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1160 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1161 lookup if dest in some wider mode is known to be const0_rtx and
1162 if so, record permanent equivalence for it to be ZERO_EXTEND of
1163 the narrower mode destination.
1165 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1167 * lto-streamer-out.c (output_function): Revert 259346.
1168 * omp-expand.c (expand_oacc_for): Likewise.
1170 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1172 PR rtl-optimization/85354
1173 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1174 * sel-sched.c (sel_global_init): ... here.
1176 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1179 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1180 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1181 mode for PE-COFF targets.
1182 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1183 (i386_pe_asm_lto_end): Likewise.
1184 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1185 (TARGET_ASM_LTO_END): Likewise.
1186 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1187 (i386_pe_asm_lto_start): New function.
1188 (i386_pe_asm_lto_end): Likewise.
1190 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1191 Richard Biener <rguenther@suse.de>
1194 * lto-streamer-out.c (output_function): Fix CFG loop state before
1196 * omp-expand.c (expand_oacc_for): Handle calls to internal
1197 functions like regular functions.
1199 2018-04-12 Richard Biener <rguenther@suse.de>
1202 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1203 for the early LTO debug to properly generate references to it
1204 during DIE emission. Do not re-use that for the skeleton for
1206 (dwarf2out_early_finish): Likewise.
1208 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1211 * config/i386/sse.md
1212 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1213 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1214 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1215 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1216 and output is a reg, avoid creating invalid lowpart subreg, but
1217 instead split into a 512-bit move. Don't split if not AVX512VL,
1218 input is xmm16+ reg and output is a mem.
1219 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1220 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1221 xmm16+ reg and output is a mem.
1223 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1225 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1226 also for flag_dwarf2_cfi_asm.
1228 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1230 PR rtl-optimization/85342
1231 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1232 a bool scalar var inside of the loop instead. Don't try to update
1233 recog_data.operand after failed apply_change_group.
1235 2018-04-12 Tom de Vries <tom@codesourcery.com>
1238 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1239 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1240 array with flexible array member as array without given dimension.
1241 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1242 argument for undefined param to true.
1244 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1247 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1248 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1249 from PowerPC section.
1250 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1251 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1252 help text that is too long.
1253 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1254 help text that is too long.
1255 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1256 help text that is too long.
1258 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1260 * config/alpha/alpha.md (stack_probe_internal): Rename
1261 from "probe_stack". Update all callers.
1263 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1265 PR rtl-optimization/84566
1266 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1267 sched_macro_fuse_insns.
1269 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1272 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1273 (compute_block_dependences): ... from here.
1275 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1277 PR tree-optimization/85331
1278 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1279 from int to HOST_WIDE_INT.
1281 2018-04-11 Martin Jambor <mjambor@suse.cz>
1284 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1285 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1286 not the same as the source val.
1287 (cgraph_edge_brings_value_p): New parameter.
1288 (gather_edges_for_value): Pass destination value to
1289 cgraph_edge_brings_value_p.
1290 (perhaps_add_new_callers): Likewise.
1291 (get_info_about_necessary_edges): Likewise and exclude values brought
1292 only by self-recursive edges.
1293 (create_specialized_node): Redirect only clones of self-calling edges.
1294 (+self_recursive_pass_through_p): New function.
1295 (find_more_scalar_values_for_callers_subset): Use it.
1296 (find_aggregate_values_for_callers_subset): Likewise.
1297 (known_aggs_to_agg_replacement_list): Removed.
1298 (decide_whether_version_node): Re-calculate known constants for all
1299 remaining context clones.
1301 2018-04-11 Richard Biener <rguenther@suse.de>
1304 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1305 from early DWARF output.
1306 (dwarf2out_early_finish): Output line info unconditionally into
1307 early DWARF and add reference to it.
1309 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1312 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1313 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1314 other than V2DFmode using iptr mode attribute.
1315 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1317 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1319 PR rtl-optimization/84659
1320 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1322 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1325 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1327 (output_loc_list): Pass address of a dummy size variable even in the
1328 locview handling loop.
1329 (index_location_lists): Add comment on why skip_loc_list_entry can't
1332 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1335 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1338 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1341 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1342 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1343 and -mstring-compare-inline-limit.
1345 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1348 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1349 for stack clash protection in a register whenever we need it to be in
1352 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1354 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1355 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1357 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1360 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1362 (mlong-double-): Ditto.
1363 * config/rs6000/sysv4.opt (msdata=): Ditto.
1364 (mtls-size=): Ditto.
1366 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1368 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1369 erroneous entries for
1370 "vector int vec_ldl (int, long int *)", and
1371 "vector unsigned int vec_ldl (int, unsigned long int *)".
1372 Add comments and entries for
1373 "vector bool char vec_ldl (int, bool char *)",
1374 "vector bool short vec_ldl (int, bool short *)",
1375 "vector bool int vec_ldl (int, bool int *)",
1376 "vector bool long long vec_ldl (int, bool long long *)",
1377 "vector pixel vec_ldl (int, pixel *)",
1378 "vector long long vec_ldl (int, long long *)",
1379 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1380 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1381 type tree bool_long_long_type_node and correct definition of
1382 bool_V2DI_type_node to make reference to this new type tree.
1383 (rs6000_mangle_type): Replace erroneous reference to
1384 bool_long_type_node with bool_long_long_type_node.
1385 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1386 comments to emphasize sign distinctions for char and int types and
1387 replace RS6000_BTI_bool_long constant with
1388 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1389 use of RS6000_BTI_pixel.
1390 (bool_long_type_node): Remove this macro definition.
1391 (bool_long_long_type_node): New macro definition
1393 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1395 PR rtl-optimization/85300
1396 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1397 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1398 simplify_unary_operation fails.
1400 2018-04-10 Martin Liska <mliska@suse.cz>
1402 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1403 cgraph_edge and ipa_ref.
1405 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1409 * config/i386/sse.md
1410 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1411 computation of the VEC_MERGE selector from mask.
1412 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1413 Fix decoding of the VEC_MERGE selector into mask.
1415 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1417 PR tree-optimization/85286
1418 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1420 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1422 * final.c (final_1): Set insn_last_address as well as
1423 insn_current_address.
1425 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1428 * explow.c (emit_stack_probe): Call validize_mem on memory location
1429 before passing it to gen_probe_stack. Create address operand and
1430 legitimize it for the probe_stack_address case.
1432 2018-04-09 Jan Hubicka <jh@suse.cz>
1435 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1436 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1437 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1439 (free_lang_data): Rebuild type inheritance graph.
1441 2018-04-09 Martin Sebor <msebor@redhat.com>
1443 * invoke.texi (-finline-small-functions): Mention other optimization
1445 (-findirect-inlining, -fpartial-inlining): Same.
1446 (-finline-functions-called-once): Same.
1447 (-freorder-blocks-and-partition): Same.
1449 2018-04-09 Jan Hubicka <jh@suse.cz>
1452 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1453 jumps; choose last target that matches the criteria (i.e.
1454 no partition changes for non-crossing jumps).
1455 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1456 support for redirecting crossing jumps to non-crossing.
1458 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1460 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1461 also for naked functions.
1463 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1465 * config/arc/arc.md (add_shift): New pattern.
1466 (add_shift2): Likewise.
1467 (sub_shift): Likewise.
1468 (sub_shift_cmp0_noout): Likewise.
1469 (compare_si_ashiftsi): Likewise.
1470 (xbfu_cmp0_noout): New combine pattern.
1471 (xbfu_cmp0"): Likewise.
1472 (movsi_set_cc_insn): Place the predicable variant first.
1473 (commutative_binary_cmp0_noout): Remove clobber.
1474 (commutative_binary_cmp0): New pattern.
1475 (noncommutative_binary_cmp0): Likewise.
1476 (noncommutative_binary_cmp0_noout): Likewise.
1477 (noncommutative_binary_comparison_result_used): Removed.
1478 (rsub_cmp0): New pattern.
1479 (rsub_cmp0_noout): Likewise.
1480 (extzvsi): Changed, keep only meaningful variants.
1481 (SQH, SEZ): New iterators.
1482 (SQH_postfix): New mode attribute.
1483 (SEZ_prefix): New code attribute.
1484 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1485 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1486 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1488 (noncommutative_operator): Check the availability of barrel
1491 2018-04-09 Richard Biener <rguenther@suse.de>
1493 PR tree-optimization/85284
1494 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1495 Only use the niter constraining form of simple_iv when the exit
1498 2018-04-09 Tom de Vries <tom@codesourcery.com>
1501 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1502 (define_expand "*memory_barrier"): New define_expand.
1503 (define_insn "memory_barrier"): New insn.
1505 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1507 PR rtl-optimization/80463
1508 PR rtl-optimization/83972
1509 PR rtl-optimization/83480
1511 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1512 correct producer for the insn.
1513 (tidy_control_flow): Fixup seqnos in case of debug insns.
1515 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1517 PR rtl-optimization/83913
1519 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1520 different sched-times when merging exprs.
1522 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1524 PR rtl-optimization/83962
1526 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1527 tidy_fallthru_edge and tidy_control_flow.
1529 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1531 PR rtl-optimization/83530
1533 * sel-sched.c (force_next_insn): New global variable.
1534 (remove_insn_for_debug): When force_next_insn is true, also leave only
1535 next insn in the ready list.
1536 (sel_sched_region): When the region wasn't scheduled, make another pass
1537 over it with force_next_insn set to 1.
1539 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1541 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1543 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1544 for interrupt control.
1545 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1546 functions for interrupt control.
1547 * config/nds32/nds32-intrinsic.md: Likewise.
1548 * config/nds32/nds32_intrinsic.h: Likewise.
1549 * config/nds32/nds32.h (nds32_builtins): Likewise.
1551 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1553 * config/nds32/nds32.c (nds32_init_machine_status,
1554 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1555 strict_aligned_p field.
1556 (nds32_expand_to_rtl_hook): New function.
1557 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1558 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1560 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1561 Chung-Ju Wu <jasonwucj@gmail.com>
1563 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1564 * config/nds32/nds32-n7.md: New file.
1565 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1566 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1568 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1569 * config/nds32/nds32.md (pipeline_model): Add n7.
1570 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1571 * config/nds32/pipelines.md: Include n7 settings.
1573 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1574 Chung-Ju Wu <jasonwucj@gmail.com>
1576 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1577 * config/nds32/nds32-e8.md: New file.
1578 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1579 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1581 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1582 * config/nds32/nds32.md (pipeline_model): Add e8.
1583 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1584 * config/nds32/pipelines.md: Include e8 settings.
1586 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1587 Chung-Ju Wu <jasonwucj@gmail.com>
1589 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1590 * config/nds32/nds32-n8.md: New file.
1591 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1592 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1594 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1595 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1596 * config/nds32/nds32.md (pipeline_model): Add n8.
1597 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1598 * config/nds32/pipelines.md: Include n8 settings.
1600 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1601 Chung-Ju Wu <jasonwucj@gmail.com>
1603 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1604 * config/nds32/nds32-n9-2r1w.md: New file.
1605 * config/nds32/nds32-n9-3r2w.md: New file.
1606 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1607 nds32_register_ports): New or modify for cpu n9.
1608 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1610 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1611 * config/nds32/nds32-utils.c: New file.
1612 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1613 TARGET_MUL_SLOW): Define.
1614 * config/nds32/nds32.md (pipeline_model): New attribute.
1615 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1616 New options that support cpu n9.
1617 * config/nds32/pipelines.md: Include n9 settings.
1618 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1620 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1622 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1623 information if necessary.
1624 (output_cond_branch_compare_zero): Likewise.
1625 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1626 (nds32_target_alignment): Refine for alignment.
1627 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1628 (FUNCTION_BOUNDARY): Modify.
1629 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1631 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1633 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1635 * config/nds32/constants.md (unspec_volatile_element): Add values for
1636 TLB operation and data prefetch.
1637 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1638 functions for TLB operation and data prefetch.
1639 * config/nds32/nds32-intrinsic.md: Likewise.
1640 * config/nds32/nds32_intrinsic.h: Likewise.
1641 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1642 (nds32_print_operand): Likewise.
1643 * config/nds32/nds32.h (nds32_builtins): Likewise.
1645 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1646 Andrew Pinski <pinsika@gcc.gnu.org>
1649 * match.pd: Use constant_boolean_node of correct type instead of
1650 boolean_true_node or boolean_false_node for simplifying
1651 pointer comparisons to zero.
1653 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1655 PR tree-optimization/80021
1656 * tree.c (verify_type_variant): Make error call in verify_variant_match
1657 translatable and remove final full stop.
1659 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1661 * config/nds32/constants.md (unspec_volatile_element): Add
1662 UNSPEC_VOLATILE_EH_RETURN.
1663 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1664 nds32_output_stack_pop): Support dwarf exception handling process.
1665 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1666 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1667 exception handling process.
1668 (nds32_compute_stack_frame): Likewise.
1669 (nds32_return_addr_rtx): Likewise.
1670 (nds32_initial_elimination_offset): Likewise.
1671 (nds32_expand_prologue): Likewise.
1672 (nds32_expand_epilogue): Likewise.
1673 (nds32_dynamic_chain_address): New function.
1674 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1676 (DYNAMIC_CHAIN_ADDRESS): Define.
1677 (EH_RETURN_DATA_REGNO): Define.
1678 (EH_RETURN_STACKADJ_RTX): Define.
1679 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1680 patterns for dwarf exception handling.
1682 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1684 * config/nds32/nds32.h: Clean up obsolete macros.
1686 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1688 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1689 Add enum values for particular instructions.
1690 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1691 particular intrinsic functions.
1692 * config/nds32/nds32-intrinsic.md: Likewise.
1693 * config/nds32/nds32_intrinsic.h: Likewise.
1694 * config/nds32/nds32.h (nds32_builtins): Likewise.
1695 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1696 (btst, ave): New patterns for particular instructions.
1698 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1700 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1701 Add enum values for atomic load/store and memory sync.
1702 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1704 * config/nds32/nds32-intrinsic.md: Likewise.
1705 * config/nds32/nds32_intrinsic.h: Likewise.
1706 * config/nds32/nds32.h (nds32_builtins): Likewise.
1708 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1710 PR tree-optimization/85257
1711 * fold-const.c (native_encode_vector): If not all elts could fit
1712 and off is -1, return 0 rather than offset.
1713 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1714 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1715 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1716 adjust buffer in native_interpret_expr call.
1718 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1720 * config/nds32/constants.md (unspec_volatile_element): Add cache
1721 control enum values.
1722 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1723 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1724 * config/nds32/nds32.c (nds32_cctl_names): New.
1725 (nds32_print_operand): Handle cache control register names.
1726 * config/nds32/nds32.h (nds32_builtins): New enum values.
1727 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1729 * config/nds32/nds32.md (type): Add mmu.
1730 * config/nds32/pipelines.md (simple_insn): Add mmu.
1732 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1734 * config/nds32/nds32.md (type): Remove call.
1735 * config/nds32/pipelines.md (simple_insn): Likewise.
1737 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1739 * config/nds32/constants.md (unspec_volatile_element): Add
1740 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1741 UNSPEC_VOLATILE_FMFCFG.
1742 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1743 description for fmfcfg and fmfcsr.
1744 (bdesc_1arg): Add fmtcsr.
1745 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1746 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1747 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1748 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1749 unspec_fmfcfg): New patterns.
1750 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1751 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1752 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1753 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1754 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1755 __nds32__fmfcfg): Define.
1757 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1759 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1760 intrinsic register names.
1761 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1762 intrinsic register enum values and macros.
1764 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1766 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1767 for load/store addressing form.
1768 (nds32_print_operand_address): Likewise.
1770 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1773 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1774 based on LABEL_REF. Remove useless assertion.
1775 (pic_address_needs_scratch): Fix formatting.
1776 (sparc_legitimize_pic_address): Minor tweaks.
1777 (sparc_delegitimize_address): Adjust assertion accordingly.
1778 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1779 into symbolic_operand.
1780 (movsi_high_pic_label_ref): Likewise.
1781 (movsi_lo_sum_pic_label_ref): Likewise.
1782 (movdi_pic_label_ref): Likewise.
1783 (movdi_high_pic_label_ref): Likewise.
1784 (movdi_lo_sum_pic_label_ref): Likewise.
1786 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1788 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1789 custom LIB_SPEC setup.
1791 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1792 Kito Cheng <kito.cheng@gmail.com>
1794 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1795 * config/riscv/freebsd.h: New.
1797 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1799 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1800 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1803 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1804 Kito Cheng <kito.cheng@gmail.com>
1806 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1807 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1808 * config/nds32/nds32-protos.h (nds32_output_call,
1809 nds32_output_return): Declare.
1810 * config/nds32/nds32.md: Refine all the call and return patterns.
1812 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1815 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1816 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1818 PR rtl-optimization/84872
1819 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1820 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1823 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1825 * expr.c (copy_blkmode_to_reg): Revert 254862.
1826 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1828 2018-04-06 Richard Biener <rguenther@suse.de>
1831 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1832 after seeing a component reference with an adjacent field. Treat
1833 refs to arrays at struct end of external decls similar to
1834 refs to unconstrained commons.
1836 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1839 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1840 look through SAVE_EXPRs with non-side-effects argument. Adjust
1842 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1845 2018-04-06 Richard Biener <rguenther@suse.de>
1848 * alias.c (find_base_term): New wrapper around find_base_term
1849 unwinding CSELIB_VAL_PTR changes.
1850 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1853 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1855 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1857 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1858 constant definitions.
1859 ("nop"): lr 0,0 -> nopr r0
1860 ("nop_lr0", "nop_lr1"): New insn definitions.
1862 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1864 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1865 NDS32_V3PUSH_AVAILABLE_P macro.
1867 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1868 Chung-Ju Wu <jasonwucj@gmail.com>
1870 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1871 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1872 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1873 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1874 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1875 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1876 * config/nds32/constraints.md: New constraints and checking for hard
1877 float configuration.
1878 * config/nds32/iterators.md: New mode iterator and attribute for hard
1879 float configuration.
1880 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1882 * config/nds32/nds32-fpu.md: New file.
1883 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1884 deal with hard float code generation.
1885 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1887 (abi_type, float_reg_number): New enum type.
1888 * config/nds32/nds32-predicates.c: New predicates for hard float.
1889 * config/nds32/nds32-protos.h: Declare functions for hard float.
1890 * config/nds32/nds32.c: Implementation for hard float configuration.
1891 * config/nds32/nds32.h: Definitions for hard float configuration.
1892 * config/nds32/nds32.md: Include hard float machine description and
1893 modify patterns for hard float configuration.
1894 * config/nds32/nds32.opt: New options for hard float configuration.
1895 * config/nds32/predicates.md: New predicates for hard float
1898 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1900 * common/config/nds32/nds32-common.c
1901 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1903 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1906 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1907 CONSTRUCTOR_ELT (ctor, ...)->value.
1909 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1912 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1914 2018-04-05 Tom de Vries <tom@codesourcery.com>
1917 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1920 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1921 Kito Cheng <kito.cheng@gmail.com>
1923 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1924 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1925 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1926 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1928 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1929 Kito Cheng <kito.cheng@gmail.com>
1931 * config/nds32/constraints.md (Ufe): New memory constraint.
1932 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1933 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1934 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1936 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1937 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1939 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1941 * config/nds32/nds32.md: Use optimize_size in the condition for
1942 alu-shift instructions.
1944 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1946 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1948 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1950 * config/nds32/nds32.md (negsi2): Refine pattern.
1952 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1953 Chung-Ju Wu <jasonwucj@gmail.com>
1955 * config/nds32/iterators.md (shift_rotate): New code iterator.
1956 (shift): New code attribute.
1957 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1958 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1959 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1960 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1961 bit-wise operations.
1962 (andsi3, *andsi3): Ditto.
1963 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1964 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1965 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1966 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1967 nds32_ior_operand, nds32_xor_operand): New predicates.
1969 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1971 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1972 (addsi3, subsi3): ... this.
1974 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1976 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1978 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1980 * config/nds32/nds32.md: Adjust indention.
1982 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1984 * config/nds32/nds32.md (feature): New attribute.
1986 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1988 * config/nds32/nds32.md (subtype): New attribute.
1990 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1993 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1994 expansion to perform a bitwise AND of the argument followed by a
1995 boolean negation of the result.
1997 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1999 PR rtl-optimization/84878
2000 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2001 the basic block. Assert the use reference is not artificial and that
2002 it has an associated insn.
2004 2018-04-04 Michael Matz <matz@suse.de>
2006 * builtins.c (compute_objsize): Pass correct operand
2007 to array_at_struct_end_p.
2009 2018-04-04 Richard Biener <rguenther@suse.de>
2012 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2013 from contexts for DINFO_LEVEL_TERSE and below.
2015 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2017 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2018 resiter_operand condition.
2019 * config/nds32/nds32.md (*move<mode>): Ditto.
2021 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2022 Monk Chiang <sh.chiang04@gmail.com>
2024 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2026 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2028 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2030 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2031 Kito Cheng <kito.cheng@gmail.com>
2033 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2034 nds32_cond_code_str, output_cond_branch,
2035 output_cond_branch_compare_zero, nds32_expand_cbranch,
2036 nds32_expand_cstore, nds32_expand_movcc,
2037 nds32_output_cbranchsi4_equality_zero,
2038 nds32_output_cbranchsi4_equality_reg,
2039 nds32_output_cbranchsi4_equality_reg_or_const_int,
2040 nds32_output_cbranchsi4_greater_less_zero: New functions.
2041 * config/nds32/nds32-protos.h (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): Declare.
2047 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2048 nds32_rimm11s_operand): New predicates.
2049 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2050 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2053 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2055 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2056 * config/nds32/nds32.md: Ditto.
2057 * config/nds32/pipelines.md: Ditto.
2059 2018-04-04 Richard Biener <rguenther@suse.de>
2061 PR tree-optimization/85168
2062 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2063 propagating abnormals.
2065 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2067 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2069 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2070 Kito Cheng <kito.cheng@gmail.com>
2072 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2073 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2074 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2075 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2076 * config/nds32/nds32.md (sibcall_internal): New.
2077 (sibcall_register): Remove.
2078 (sibcall_immediate): Remove.
2079 (sibcall_value_internal): New.
2080 (sibcall_value_register): Remove.
2081 (sibcall_value_immediate): Remove.
2082 * config/nds32/predicates.md (nds32_general_register_operand): New.
2083 (nds32_call_address_operand): New.
2085 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2087 PR rtl-optimization/85167
2088 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2089 bb_defs if *split_p, instead preinitialize it to NULL.
2091 PR tree-optimization/85156
2092 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2093 evaluating the argument multiple times.
2095 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2097 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2099 (_mm_cvtpd_ps): Likewise.
2100 (_mm_cvttpd_epi32): Likewise.
2101 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2102 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2103 vector, pixel, and bool following altivec.h include.
2105 2018-04-03 Martin Sebor <msebor@redhat.com>
2107 * doc/extend.texi (Common Function Attributes): Clarify.
2108 (const attribute): Likewise.
2109 (pure attribute): Likewise.
2111 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2114 * config/i386/i386.c (ix86_expand_vector_set): Use
2115 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2117 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2119 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2120 instructions when changing rounding bits to preserve precision bits
2121 in the x87 control word.
2123 2018-04-03 Martin Liska <mliska@suse.cz>
2125 PR tree-optimization/82491
2126 * rtl.h (strip_offset_and_add): Replace += suboffset with
2127 poly_uint64 () + suboffset.
2129 2018-03-29 Martin Liska <mliska@suse.cz>
2130 Martin Jambor <mjambor@suse.cz>
2133 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2134 param_type is not an integral or pointer type.
2136 2018-04-03 Richard Biener <rguenther@suse.de>
2138 * sese.h (recompute_all_dominators): Remove.
2140 2018-04-02 Martin Sebor <msebor@redhat.com>
2142 * doc/invoke.texi (-Wrestrict): Fix typos.
2144 2018-04-02 Jim Wilson <jimw@sifive.com>
2146 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2147 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2148 (<optab>di3, <optab>si3_extend): Likewise.
2149 (<optab>si3_mask, <optab>si3_mask_1): New.
2150 (<optab>di3_mask, <optab>di3_mask_1): New.
2151 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2152 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2153 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2155 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2157 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2160 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2162 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2163 (nds32_canonicalize_comparison): New function.
2165 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2166 Kito Cheng <kito.cheng@gmail.com>
2167 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2169 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2170 * config/nds32/constants.md (unspec_volatile_element): Add
2171 UNSPEC_VOLATILE_RELAX_GROUP.
2172 * config/nds32/nds32-relax-opt.c: New file.
2173 * config/nds32/nds32-predicates.c
2174 (nds32_symbol_load_store_p): New function.
2175 * config/nds32/nds32-protos.h
2176 (nds32_symbol_load_store_p): Declare function.
2177 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2178 * config/nds32/nds32.c
2179 (nds32_register_pass): New function to register pass.
2180 (nds32_register_passes): New function to register passes.
2181 * config/nds32/nds32.md (relax_group): New pattern.
2182 * config/nds32/nds32.opt (mrelax-hint): New option.
2183 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2185 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2187 * config/nds32/t-nds32: Modify files dependency.
2189 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2191 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2192 (PROFILE_HOOK): Define its implementation.
2194 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2196 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2197 type and 32-bit size.
2199 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2202 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2203 (V_128_256): New mode iterator.
2204 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2205 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2206 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2208 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2209 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2211 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2214 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2215 NaN inputs correctly.
2217 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2220 * config/rs6000/vsx.md (??r): New mode attribute.
2221 (*vsx_mov<mode>_64bit): Use it.
2222 (*vsx_mov<mode>_32bit): Likewise.
2224 2018-03-30 Martin Sebor <msebor@redhat.com>
2226 PR tree-optimization/84818
2227 * builtins.c (check_access): Use warning_n.
2229 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2232 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2234 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2237 2018-03-30 Julia Koval <julia.koval@intel.com>
2240 * x86-tune.def (movx, partial_reg_dependency): Enable for
2243 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2246 * lra-constraints.c (process_alt_operands): Move setting
2247 this_alternative_matches below.
2249 2018-03-29 Martin Liska <mliska@suse.cz>
2252 * doc/invoke.texi: Document how LTO works with debug info.
2253 Describe auto-load support of binutils. Mention 'x86-64'
2254 as valid option value of -march option.
2256 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2258 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2261 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2262 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2263 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2266 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2269 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2270 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2271 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2272 for RS6000_BTM_POWERPC64.
2273 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2274 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2275 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2280 2018-03-28 Carl Love <cel@us.ibm.com>
2283 2017-09-27 Carl Love <cel@us.ibm.com>
2285 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2286 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2287 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2290 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2292 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2293 instead of __vector bool.
2294 (_mm_max_pu8): Likewise.
2295 (_mm_min_pi16): Likewise.
2297 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2300 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2301 (DIVWEUO): Likewise.
2303 (DIVDEUO): Likewise.
2304 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2305 DIVWEUO and DIVDEUO.
2306 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2307 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2308 (div_extend): Likewise.
2309 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2311 (__builtin_divweuo): Likewise.
2312 (__builtin_divdeo): Likewise.
2313 (__builtin_divdeuo): Likewise.
2315 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2318 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2319 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2321 PR tree-optimization/82004
2322 * gimple-match-head.c (optimize_pow_to_exp): New function.
2323 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2324 Don't fold to exp if optimize_pow_to_exp is false.
2326 2018-03-28 Martin Liska <mliska@suse.cz>
2329 * calls.c (initialize_argument_information): Fix trailing space.
2330 * common.opt: Fix typo and provide better explanation for
2331 -fsanitize-coverage option.
2332 * config/i386/i386.opt: Fix typo.
2334 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2335 Martin Liska <mliska@suse.cz>
2338 * gimplify.c (asan_poison_variable): Don't do the check for
2339 gimplify_omp_ctxp here.
2340 (gimplify_decl_expr): Do it here.
2341 (gimplify_target_expr): Likewise.
2343 2018-03-28 Martin Liska <mliska@suse.cz>
2346 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2347 chkp_type_bounds_count if MPX is not enabled.
2349 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2351 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2353 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2356 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2357 function to create the function decl for complex long double
2358 multiply and divide for -mabi=ieeelongdouble.
2359 (init_float128_ieee): Call it.
2361 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2364 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2365 -fcf-protection=branch -mibt.
2366 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2368 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2371 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2373 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2376 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2377 extern array declarations.
2379 2018-03-27 Richard Biener <rguenther@suse.de>
2382 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2383 explicit single_use checks.
2385 2018-03-27 Richard Biener <rguenther@suse.de>
2387 PR tree-optimization/85082
2388 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2391 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2393 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2394 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2395 Turn on fasynchronous-unwind-tables and funwind-tables.
2397 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2400 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2401 (*bmi_blsr_<mode>_ccz): Ditto.
2403 2018-03-26 Tom de Vries <tom@codesourcery.com>
2405 PR tree-optimization/85063
2406 * omp-general.c (offloading_function_p): New function. Factor out
2408 * omp-offload.c (pass_omp_target_link::gate): ... here.
2409 * omp-general.h (offloading_function_p): Declare.
2410 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2411 with attribute omp declare target for offloading functions.
2413 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2415 PR tree-optimization/84005
2416 * tree-data-ref.h (get_base_for_alignment): Declare.
2417 * tree-data-ref.c (get_base_for_alignment_1): New function.
2418 (get_base_for_alignment): Likewise.
2419 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2420 get_base_for_alignment to find a suitable base object, instead
2421 of always using drb->base_address.
2423 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2426 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2427 known size by default.
2429 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2432 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2433 and non BLKmode operands.
2435 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2438 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2439 Clean up attributes.
2441 2018-03-23 Richard Biener <rguenther@suse.de>
2444 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2445 we are going to emit early debug for LTO.
2447 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2450 * function.c (match_asm_constraints_1): Don't optimize if input
2451 doesn't satisfy general_operand predicate for output's mode.
2454 * alias.c (write_dependence_p): Don't require for x_canonicalized
2455 non-VOIDmode if x has VOIDmode.
2458 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2459 just don't try to optimize it rather than assert it never happens.
2461 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2463 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2464 macro expansions for definition of ST_INTERNAL_<mode> and
2465 LD_INTERNAL_<mode> builtins.
2466 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2468 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2470 (altivec_expand_st_builtin): Likewise.
2471 (altivec_expand_builtin): Remove calls to deleted functions.
2472 (rs6000_address_for_altivec): Delete this function.
2473 * config/rs6000/vector.md: Remove expands for
2474 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2476 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2479 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2480 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2481 re-computing once computed.
2482 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2483 (arm_init_machine_status): Initialize
2484 machine->static_chain_stack_bytes.
2486 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2489 * doc/extend.texi: Add four new prototypes for vec_ld.
2490 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2491 definitions for more logical presentation.
2492 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2493 entries for V1TI variants of __builtin_altivec_ld builtin.
2494 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2495 handling of V1TI variant of LVX icode pattern.
2496 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2497 (rs6000_gimple_fold_builtin): Likewise.
2498 (altivec_init_builtins): Add code to define
2499 __builtin_altivec_lvx_v1ti function.
2501 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2504 * function.c (match_asm_constraints_1): Don't do the optimization
2505 if input isn't a REG, SUBREG, MEM or constant.
2507 2018-03-22 Tom de Vries <tom@codesourcery.com>
2509 PR tree-optimization/84956
2510 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2511 bb_has_abnormal_pred.
2513 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2516 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2517 DECL_INITIAL (decl) to decl at the end.
2518 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2521 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2523 * doc/extend.texi (__builtin_tgmath): Document when complex
2524 integer types are treated as _Complex _Float64.
2526 2018-03-21 Tom de Vries <tom@codesourcery.com>
2528 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2530 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2532 PR tree-optimization/84960
2533 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2534 if it is ENTRY block, move them into single succ of ENTRY in that case.
2536 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2538 PR tree-optimization/84811
2539 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2540 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2541 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2542 (known_subrange_p): Update accordingly. Cast each value involved
2543 in the size comparison, rather than casting the result of the
2546 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2548 PR tree-optimization/84982
2549 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2550 by flipping the least significant bit rather than all bits from
2551 bitpos to bitpos + bitsize - 1.
2553 2018-03-21 Nathan Sidwell <nathan@acm.org>
2555 * doc/extend.texi (Deprecated Features): Remove mention of
2556 long-deleted deprecations.
2558 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2561 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2562 * configure: Regenerate.
2564 2018-03-21 Tom de Vries <tom@codesourcery.com>
2566 PR tree-optimization/83126
2567 * tree-parloops.c (num_phis): New function.
2568 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2570 2018-03-21 Nathan Sidwell <nathan@acm.org>
2572 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2573 mention anon-struct/union members and trailing attributes.
2575 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2577 PR tree-optimization/84969
2578 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2579 builtin memset partitions if they set different rhs values.
2581 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2583 PR rtl-optimization/84989
2584 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2585 VEC_DUPLICATE with scalar result mode.
2587 2018-03-21 Martin Liska <mliska@suse.cz>
2590 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2591 not intended return statement.
2593 2018-03-21 Martin Liska <mliska@suse.cz>
2596 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2597 (chkp_find_bound_slots_1): Limit number of iterations.
2599 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2602 * Minor grammar fixes for x86 options.
2604 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2607 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2608 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2610 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2613 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2614 (altivec_lvx_<mode>_1op): Likewise.
2615 (altivec_stvx_<mode>_2op): Likewise.
2616 (altivec_stvx_<mode>_1op): Likewise.
2617 (altivec_lvx_<VM2:mode>): New define_expand.
2618 (altivec_stvx_<VM2:mode>): Likewise.
2619 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2620 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2621 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2622 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2623 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2624 (rs6000_gen_lvx): Likewise.
2625 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2626 (altivec_expand_stv_builtin): Likewise.
2627 (altivec_expand_builtin): Likewise.
2628 * config/rs6000/vector.md: Likewise.
2630 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2633 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2636 2018-03-20 Richard Biener <rguenther@suse.de>
2639 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2640 sign-conversions as zero, fall back to standard scalar_stmt
2643 2018-03-20 Martin Liska <mliska@suse.cz>
2646 * predict.c (rebuild_frequencies): Handle case when we have
2647 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2649 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2652 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2653 flag_section_anchors.
2654 * varasm.c (use_blocks_for_decl_p): Remove hack for
2655 dw2_force_const_mem.
2658 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2660 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2661 be created, use lowpart_subreg of operands[0] rather than operands[0]
2663 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2664 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2665 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2666 and n constraint instead of aarch64_shift_imm_di and Usd.
2667 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2668 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2670 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2673 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2674 to favor GPR over NEON registers.
2675 (<shift>di3_neon): Likewise.
2677 2018-03-20 Tom de Vries <tom@codesourcery.com>
2680 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2681 (nvptx_process_pars): Emit bar.sync asap and alap.
2683 2018-03-20 Tom de Vries <tom@codesourcery.com>
2686 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2687 seen_label if seen_label is already set.
2689 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2692 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2693 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2694 Use 1U instead of 1. Formatting fixes.
2697 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2698 instead of TREE_TYPE (s1) for the return value.
2700 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2702 PR tree-optimization/84946
2703 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2704 bitsize + bitsize in poly_uint64 rather than poly_int64.
2707 * dwarf2asm.c: Include fold-const.c.
2708 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2709 of decl rather than decl itself.
2711 PR rtl-optimization/84643
2712 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2714 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2717 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2718 calling assemble_variable.
2720 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2723 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2724 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2726 2018-03-19 Jim Wilson <jimw@sifive.com>
2729 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2730 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2731 (riscv_first_stack_step): Likewise.
2732 (riscv_option_override): Use STACK_BOUNDARY instead of
2734 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2736 (BIGGEST_ALIGNMENT): Set to 128.
2737 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2738 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2741 2018-03-19 Richard Biener <rguenther@suse.de>
2743 PR tree-optimization/84933
2744 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2745 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2747 2018-03-19 Richard Biener <rguenther@suse.de>
2749 PR tree-optimization/84859
2750 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2751 (cond_if_else_store_replacement): Perform sinking operation on
2752 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2753 Generalize what a BB with a single eligible store is.
2755 2018-03-19 Richard Biener <rguenther@suse.de>
2757 PR tree-optimization/84929
2758 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2759 chrec_is_positive against non-chrec arg.
2761 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2764 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2766 2018-03-18 Martin Liska <mliska@suse.cz>
2768 PR rtl-optimization/84635
2769 * regrename.c (build_def_use): Use matches_mode only when
2772 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2774 PR tree-optimization/84913
2775 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2776 vectorize chains of COND_EXPRs.
2778 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2780 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2782 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2784 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2786 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2788 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2790 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2791 Kito Cheng <kito.cheng@gmail.com>
2793 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2794 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2795 (nds32_adjust_reg_alloc_order): New function.
2796 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2798 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2800 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2801 nds32_print_operand, nds32_print_operand_address): Use
2802 HOST_WIDE_INT_PRINT_DEC instead.
2804 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2806 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2808 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2811 * config/i386/i386.c (initial_ix86_tune_features,
2812 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2814 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2815 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2816 rather than 1u << ix86_tune. Formatting fix.
2817 (ix86_option_override_internal): Change ix86_arch_mask from
2818 unsigned int to unsigned HOST_WIDE_INT, initialize to
2819 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2820 (ix86_function_specific_restore): Likewise.
2822 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2825 * postreload.c (reload_combine_recognize_pattern): Perform
2826 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2827 truncate_int_for_mode the result for the destination's mode.
2830 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2831 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2833 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2836 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2839 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2840 regno_allocno_class_array and sorted_pseudos.
2841 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2842 insns where regno is used.
2844 2018-03-16 Martin Liska <mliska@suse.cz>
2847 * multiple_target.c (create_dispatcher_calls): Redirect
2848 reference in the symbol table.
2850 2018-03-16 Martin Liska <mliska@suse.cz>
2853 * multiple_target.c (create_dispatcher_calls): Redirect also
2856 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2860 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2861 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2862 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2864 2018-03-16 Julia Koval <julia.koval@intel.com>
2866 * doc/invoke.texi (Skylake Server): Add CLWB.
2867 Cannonlake): Remove CLWB.
2869 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2871 PR tree-optimization/84841
2872 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2874 (FLOAT_ONE_CONST_TYPE): Define.
2875 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2876 (sort_by_operand_rank): Put entries with higher constant_type last
2877 rather than first to match comments.
2879 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2881 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2884 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2887 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2888 check and use address of target always.
2890 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2893 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2894 (indirect_thunk_bnd_needed): Likewise.
2895 (indirect_thunks_used): Likewise.
2896 (indirect_thunks_bnd_used): Likewise.
2897 (indirect_return_needed): New.
2898 (indirect_return_bnd_needed): Likewise.
2899 (output_indirect_thunk_function): Add a bool argument for
2901 (output_indirect_thunk_function): Don't generate alias for
2902 function return thunk.
2903 (ix86_code_end): Call output_indirect_thunk_function to generate
2904 function return thunks.
2905 (ix86_output_function_return): Set indirect_return_bnd_needed
2906 and indirect_return_needed instead of indirect_thunk_bnd_needed
2907 and indirect_thunk_needed.
2909 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2911 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2912 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2913 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2915 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2916 Paul Hua <paul.hua.gm@gmail.com>
2919 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2921 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2923 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2924 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2927 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2930 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2931 instead of GET_MODE_SIZE when comparing Units.
2933 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2936 * varasm.c (hash_section): Return an unchangeble hash value
2937 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2938 Return !aarch64_can_use_per_function_literal_pools_p ().
2940 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2943 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2944 rather than address of cmode as last argument to prepare_cmp_insn.
2946 2018-03-15 Julia Koval <julia.koval@intel.com>
2948 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2949 F_AVX512VNNI, F_AVX512BITALG): New.
2951 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2954 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2955 insn for floating-point loads and stores.
2957 2018-03-14 Carl Love <cel@us.ibm.com>
2959 * config/rs6000/rs6000-c.c: Add macro definitions for
2960 ALTIVEC_BUILTIN_VEC_PERMXOR.
2961 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2962 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2963 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2964 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2966 * config/doc/extend.texi: Add prototypes for vec_permxor.
2968 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2971 * diagnostic-show-locus.c (class layout_point): Convert m_line
2972 from int to linenum_type.
2973 (line_span::comparator): Use linenum "compare" function when
2974 comparing line numbers.
2975 (test_line_span): New function.
2976 (layout_range::contains_point): Convert param "row" from int to
2978 (layout_range::intersects_line_p): Likewise.
2979 (layout::will_show_line_p): Likewise.
2980 (layout::print_source_line): Likewise.
2981 (layout::should_print_annotation_line_p): Likewise.
2982 (layout::print_annotation_line): Likewise.
2983 (layout::print_leading_fixits): Likewise.
2984 (layout::annotation_line_showed_range_p): Likewise.
2985 (struct line_corrections): Likewise for field m_row.
2986 (line_corrections::line_corrections): Likewise for param "row".
2987 (layout::print_trailing_fixits): Likewise.
2988 (layout::get_state_at_point): Likewise.
2989 (layout::get_x_bound_for_row): Likewise.
2990 (layout::print_line): Likewise.
2991 (diagnostic_show_locus): Likewise for locals "last_line" and
2993 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2994 * input.c (selftest::test_linenum_comparisons): New function.
2995 (selftest::input_c_tests): Call it.
2996 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2997 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2998 * selftest.h (ASSERT_GT): New macro.
2999 (ASSERT_GT_AT): New macro.
3000 (ASSERT_LT): New macro.
3001 (ASSERT_LT_AT): New macro.
3003 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3005 PR rtl-optimization/84780
3006 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3008 2018-03-14 Martin Liska <mliska@suse.cz>
3010 * tree.c (record_node_allocation_statistics): Use
3011 get_stats_node_kind.
3012 (get_stats_node_kind): New function extracted from
3013 record_node_allocation_statistics.
3014 (free_node): Use get_stats_node_kind.
3016 2018-03-14 Richard Biener <rguenther@suse.de>
3018 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3019 that the value-set of ANTIC_IN doesn't grow.
3022 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3024 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3025 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3027 2018-03-14 Julia Koval <julia.koval@intel.com>
3029 * config.gcc (icelake-client, icelake-server): New.
3031 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3032 (initial_ix86_arch_features): Ditto.
3033 (PTA_SKYLAKE): Add SGX.
3034 (PTA_ICELAKE): Remove.
3035 (PTA_ICELAKE_CLIENT): New.
3036 (PTA_ICELAKE_SERVER): New.
3037 (ix86_option_override_internal): Split up icelake on icelake client and
3039 (get_builtin_code_for_version): Ditto.
3040 (fold_builtin_cpu): Ditto.
3041 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3042 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3043 * config/i386/i386.h (processor_type): Ditto.
3044 * doc/invoke.texi: Ditto.
3046 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3049 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3050 INTEGER_CST offset, add it together with bitpos / 8 and
3051 sign extend based on POINTER_SIZE.
3055 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3058 * config/i386/constraints.md (Yc): New register constraint.
3059 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3060 Use Yc constraint for alternative 2 of operand 0. Remove
3061 preferred_for_speed attribute.
3063 2018-03-14 Richard Biener <rguenther@suse.de>
3065 PR tree-optimization/84830
3066 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3067 with the old one to avoid oscillations.
3069 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3072 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3074 (assign_by_spills): Return a flag of reload assignment failure.
3075 Do not process the reload assignment failures. Do not spill other
3076 reload pseudos if they has the same reg class. Update n if
3078 (lra_assign): Add a return arg. Set up from the result of
3079 assign_by_spills call.
3080 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3081 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3082 usage_insns if it is not NULL.
3083 (spill_hard_reg_in_range): New function.
3084 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3085 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3086 function prototypes.
3087 (lra_assign): Change prototype.
3088 * lra.c (lra): Add code to deal with fails by splitting hard reg
3091 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3093 * config/riscv/riscv.opt (mrelax): New option.
3094 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3095 "norelax" when riscv_mrelax is disabled.
3096 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3098 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3101 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3102 reassociation for int modes.
3104 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3106 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3107 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3109 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3110 * config/aarch64/aarch64-sve.md
3111 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3112 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3113 (*extend<mode><Vwide>2): Rename to...
3114 (aarch64_sve_extend<mode><Vwide>2): ...this.
3115 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3116 renaming the old pattern to...
3117 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3119 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3120 define_expand, renaming the old pattern to...
3121 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3122 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3123 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3124 account when deciding which SVE instruction the optab should use.
3125 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3127 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3129 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3130 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3131 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3132 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3133 (tlsdesc_small_<mode>): Turn a define_expand and use
3134 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3135 (tlsdesc_small_advsimd_<mode>): ...this.
3136 (tlsdesc_small_sve_<mode>): New pattern.
3138 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3140 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3141 (UNSPEC_UMUL_HIGHPART): New constants.
3142 (MUL_HIGHPART): New int iteraor.
3143 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3144 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3146 (*<su>mul<mode>3_highpart): New define_insn.
3148 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3151 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3154 2018-03-13 Martin Liska <mliska@suse.cz>
3157 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3159 (sem_item_optimizer::~sem_item_optimizer): Release it.
3160 (sem_item_optimizer::merge_classes): Register variable aliases.
3161 (sem_item_optimizer::fixup_pt_set): New function.
3162 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3163 * ipa-icf.h: Declare new variables and functions.
3165 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3168 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3169 integer_pow2p@2 and test integer_pow2p in condition.
3170 (A < 0 ? C : 0): Similarly for @1.
3173 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3174 characters starting at p contain '\0' character, don't look beyond
3178 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3179 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3182 * reg-stack.c (change_stack): Change update_end var from int to
3183 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3184 also call set_block_for_insn on the newly added insns and rescan.
3187 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3188 on the last operand.
3191 * tree.c (stabilize_reference_1): Return save_expr (e) for
3192 STATEMENT_LIST even if it doesn't have side-effects.
3194 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3196 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3198 2018-03-12 Renlin Li <renlin.li@arm.com>
3200 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3201 aarch64_output_scalar_simd_mov_immediate.
3203 2018-03-12 Martin Sebor <msebor@redhat.com>
3205 PR tree-optimization/83456
3206 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3207 for perfectly overlapping calls to memcpy.
3208 (gimple_fold_builtin_memory_chk): Same.
3209 (gimple_fold_builtin_strcpy): Handle no-warning.
3210 (gimple_fold_builtin_stxcpy_chk): Same.
3211 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3213 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3215 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3216 parameter. Use it for SFmode.
3217 (rs6000_function_arg_advance_1): Adjust.
3218 (rs6000_function_arg): Adjust.
3219 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3221 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3223 PR rtl-optimization/84169
3224 PR rtl-optimization/84780
3225 * combine.c (can_combine_p): Check for a 2-insn combination whether
3226 the destination register is used between the two insns, too.
3228 2018-03-12 Richard Biener <rguenther@suse.de>
3230 PR tree-optimization/84803
3231 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3232 for refs DR analysis didn't process.
3234 2018-03-12 Richard Biener <rguenther@suse.de>
3236 PR tree-optimization/84777
3237 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3238 force-vectorize loops ignore whether we are optimizing for size.
3240 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3242 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3243 (TARGET_MD_ASM_ADJUST): Define.
3245 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3246 Kito Cheng <kito.cheng@gmail.com>
3247 Chung-Ju Wu <jasonwucj@gmail.com>
3249 * config/nds32/nds32.c (nds32_compute_stack_frame,
3250 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3251 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3252 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3253 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3254 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3255 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3256 * config/nds32/nds32.md (prologue, epilogue): Use macro
3257 NDS32_V3PUSH_AVAILABLE_P to do checking.
3259 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3262 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3263 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3264 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3265 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3266 addition of most attributes on !orig_type_die or the attribute not
3267 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3269 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3270 Chung-Ju Wu <jasonwucj@gmail.com>
3272 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3274 * config/nds32/nds32.opt (mvh): New option.
3276 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3277 Chung-Ju Wu <jasonwucj@gmail.com>
3279 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3281 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3282 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3285 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3286 Chung-Ju Wu <jasonwucj@gmail.com>
3288 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3290 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3291 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3293 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3294 Kito Cheng <kito.cheng@gmail.com>
3295 Chung-Ju Wu <jasonwucj@gmail.com>
3297 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3298 UNSPEC_FFMISM and UNSPEC_FLMISM.
3299 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3300 for ffb, ffmism and flmism.
3301 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3302 (unspec_ffmism): Ditto.
3303 (unspec_flmism): Ditto.
3304 (nds32_expand_builtin_impl): Check if string extension is available.
3305 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3306 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3308 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3311 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3314 * lra-assigns.c (assign_by_spills): Return a flag of reload
3315 assignment failure. Do not process the reload assignment
3316 failures. Do not spill other reload pseudos if they has the same
3318 (lra_assign): Add a return arg. Set up from the result of
3319 assign_by_spills call.
3320 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3321 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3322 usage_insns if it is not NULL.
3323 (spill_hard_reg_in_range): New function.
3324 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3325 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3326 function prototypes.
3327 (lra_assign): Change prototype.
3328 * lra.c (lra): Add code to deal with fails by splitting hard reg
3331 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3334 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3336 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3339 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3340 (dw_val_node): Add val_symbolic_view.
3341 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3342 (symview_upper_bound): New.
3343 (new_line_info_table): Initialize symviews_since_reset.
3344 (dwarf2out_source_line): Count symviews_since_reset and set
3345 symview_upper_bound.
3346 (dw_val_equal_p): Handle symview.
3347 (add_AT_symview): New.
3348 (print_dw_val): Handle symview.
3349 (attr_checksum, attr_checksum_ordered): Likewise.
3350 (same_dw_val_p, size_of_die): Likewise.
3351 (value_format, output_die): Likewise.
3352 (add_high_low_attributes): Use add_AT_symview for entry_view.
3353 (dwarf2out_finish): Reset symview_upper_bound, clear
3356 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3359 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3360 Add strict argument and use it.
3361 (rs6000_split_multireg_move): Update for new strict argument.
3362 (mem_operand_gpr): Disallow all non-offsettable addresses.
3363 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3365 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3368 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3369 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3370 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3373 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3374 decl, use remap_type if we want to use the type.
3376 2018-03-09 Martin Sebor <msebor@redhat.com>
3378 PR tree-optimization/84526
3379 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3381 (builtin_access::generic_overlap): Be prepared to handle non-array
3384 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3386 PR rtl-optimization/84682
3387 * lra-constraints.c (process_address_1): Check is_address flag
3388 for address constraints.
3389 (process_alt_operands): Likewise.
3390 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3391 preprocess_constraints.
3392 * recog.h (preprocess_constraints): Add oploc parameter.
3394 * recog.c (preprocess_constraints): Test address_operand for
3395 CT_ADDRESS constraints.
3397 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3400 * lra-assigns.c (assign_by_spills): Return a flag of reload
3401 assignment failure. Do not process the reload assignment
3402 failures. Do not spill other reload pseudos if they has the same
3404 (lra_assign): Add a return arg. Set up from the result of
3405 assign_by_spills call.
3406 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3407 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3408 usage_insns if it is not NULL.
3409 (spill_hard_reg_in_range): New function.
3410 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3411 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3412 function prototypes.
3413 (lra_assign): Change prototype.
3414 * lra.c (lra): Add code to deal with fails by splitting hard reg
3417 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3420 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3421 Accept complain bool parameter. Only emit errors if it is true.
3422 (arm_parse_cpu_option_name): Likewise.
3423 (arm_target_thumb_only): Adjust callers of the above.
3424 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3425 prototype to take a default true bool parameter.
3426 (arm_parse_arch_option_name): Likewise.
3428 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3429 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3433 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3434 * configure: Regenerate.
3435 * configure.ac ("linker --version-script option"): New.
3436 ("linker soname option"): New.
3438 2018-03-09 Richard Biener <rguenther@suse.de>
3440 PR tree-optimization/84775
3441 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3442 immediate uses of predicate stmts and mark them modified.
3445 PR tree-optimization/84178
3446 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3448 (version_loop_for_if_conversion): Delay update_ssa call.
3449 (tree_if_conversion): Delay update_ssa until after predicate
3452 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3455 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3456 when the function accesses prior frames.
3458 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3461 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3462 gen_llsym, otherwise call maybe_gen_llsym.
3465 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3466 has ',' character inside of it.
3468 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3471 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3472 as clobbering CC_REGNUM.
3474 2018-03-08 Richard Biener <rguenther@suse.de>
3477 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3478 (follow_copies_to_constant): Do not follow SSA names registered
3481 2018-03-08 Richard Biener <rguenther@suse.de>
3483 PR tree-optimization/84178
3484 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3486 (version_loop_for_if_conversion): Delay update_ssa call.
3487 (tree_if_conversion): Delay update_ssa until after predicate
3490 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3492 PR tree-optimization/84178
3493 * tree-if-conv.c (release_bb_predicate): Remove the
3494 the assertion that the stmts have NULL use_ops.
3495 Discard the statements, asserting that they haven't
3496 yet been added to a BB.
3498 2018-03-08 Richard Biener <rguenther@suse.de>
3500 PR tree-optimization/84746
3501 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3502 (phi_translate): Pass in destination ANTIC_OUT set.
3503 (phi_translate_1): Likewise. For a simplified result lookup
3504 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3505 (phi_translate_set): Adjust.
3506 (do_pre_regular_insertion): Likewise.
3507 (do_pre_partial_partial_insertion): Likewise.
3509 2018-03-08 Martin Liska <mliska@suse.cz>
3511 PR gcov-profile/84735
3512 * doc/gcov.texi: Document usage of profile files.
3513 * gcov-io.h: Document changes in the format.
3515 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3519 * dwarf2out.c (struct dw_line_info_table): Update comments for
3521 (FORCE_RESET_NEXT_VIEW): New.
3522 (FORCE_RESETTING_VIEW_P): New.
3523 (RESETTING_VIEW_P): Check for -1 too.
3524 (ZERO_VIEW_P): Likewise.
3525 (new_line_info_table): Force-reset next view.
3526 (dwarf2out_begin_function): Likewise.
3527 (dwarf2out_source_line): Simplify zero_view_p initialization.
3528 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3529 view directly. Omit view when omitting .loc at line 0.
3531 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3533 PR tree-optimization/84740
3534 * tree-switch-conversion.c (process_switch): Call build_constructors
3535 only if info.phi_count is non-zero.
3537 PR tree-optimization/84739
3538 * tree-tailcall.c (find_tail_calls): Check call arguments against
3539 DECL_ARGUMENTS (current_function_decl) rather than
3540 DECL_ARGUMENTS (func) when checking for tail recursion.
3542 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3544 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3545 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3546 Volker Reichelt's entry and add entries for people that perform
3547 GCC fuzzy testing and report numerous bugs.
3549 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3552 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3553 readonly data in sdata, if that is disabled.
3554 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3555 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3556 -mreadonly-in-sdata option.
3558 2018-03-07 Martin Sebor <msebor@redhat.com>
3560 PR tree-optimization/84468
3561 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3562 basic block when looking for nul assignment.
3564 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3567 * except.h (output_function_exception_table): Adjust prototype.
3568 * except.c (output_function_exception_table): Remove FNNAME parameter
3569 and add SECTION parameter. Ouput one part of the table at a time.
3570 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3571 the first part of the exception table and emit unwind directives.
3572 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3573 (i386_pe_seh_cold_init): Likewise.
3574 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3575 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3576 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3577 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3578 * config/i386/winnt.c: Include except.h.
3579 (struct seh_frame_state): Add reg_offset, after_prologue and
3580 in_cold_section fields.
3581 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3582 (i386_pe_seh_cold_init): New function.
3583 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3584 to seh->in_cold_section.
3585 (seh_emit_push): Record the offset of the push.
3586 (seh_emit_save): Record the offet of the save.
3587 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3588 Test seh->after_prologue to disregard the epilogue.
3589 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3590 (i386_pe_end_cold_function): New function.
3592 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3595 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3596 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3599 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3601 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3602 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3605 * multiple_target.c: Include tree-inline.h and intl.h.
3606 (expand_target_clones): Diagnose and fail if node->definition and
3607 !tree_versionable_function_p (node->decl).
3609 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3611 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3613 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3614 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3615 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3617 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3620 * combine.c (try_combine): Use reg_or_subregno instead of handling
3621 just paradoxical SUBREGs and REGs.
3623 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3625 * config/arc/arc.c (arc_finalize_pic): Remove function.
3626 (arc_must_save_register): We use single base PIC register, remove
3627 checks to save/restore the PIC register.
3628 (arc_expand_prologue): Likewise.
3629 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3631 (arc_verify_short): Likewise.
3632 (arc_attr_type): Likewise.
3633 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3634 (walk_stores): Likewise.
3635 (arc_address_cost): Make it static.
3636 (arc_verify_short): Likewise.
3637 (branch_dest): Likewise.
3638 (arc_attr_type): Likewise.
3639 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3640 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3641 (arc_final_prescan_insn): Remove inserting the nops due to
3642 hardware hazards. It is done in reorg step.
3643 (insn_length_variant_t): Remove.
3644 (insn_length_parameters_t): Likewise.
3645 (arc_insn_length_parameters): Likewise.
3646 (arc_get_insn_variants): Likewise.
3647 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3649 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3652 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3655 PR tree-optimization/84687
3656 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3658 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3660 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3662 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3663 Rename to ppc_speculation_barrier.
3664 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3665 __builtin_ppc_speculation_barrier.
3667 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3670 * combine.c (combine_simplify_rtx): Don't try to simplify if
3671 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3674 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3676 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3677 to 32 bytes when compiling for POWER9.
3679 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3682 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3683 regparm >= 3 with no arg reg available also for calls with
3684 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3687 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3689 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3691 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3694 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3696 2018-03-05 Richard Biener <rguenther@suse.de>
3698 PR tree-optimization/84486
3699 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3700 When inserting a __builtin_assume_aligned call set the LHS
3701 SSA name alignment info accordingly.
3703 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3705 PR tree-optimization/84114
3706 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3707 Avoid reassociation of FLOAT_MODE addition.
3709 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3711 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3712 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3713 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3714 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3715 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3716 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3717 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3719 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3720 __builtin_ia32_wbinvd): New builtins.
3721 (SPECIAL_ARGS2): New.
3722 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3723 (SPECIAL_ARGS2): New.
3724 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3725 (ix86_valid_target_attribute_inner_p): Ditto.
3726 (ix86_init_mmx_sse_builtins): Add special_args2.
3727 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3728 TARGET_WBNOINVD_P): New.
3729 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3730 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3731 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3732 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3733 * config/i386/pconfigintrin.h: New file.
3734 * config/i386/wbnoinvdintrin.h: Ditto.
3735 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3736 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3738 2018-03-05 Richard Biener <rguenther@suse.de>
3740 PR tree-optimization/84670
3741 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3743 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3744 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3745 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3746 doesn't grow if all successors (recursively) were visited at least
3749 2018-03-05 Richard Biener <rguenther@suse.de>
3751 PR tree-optimization/84650
3752 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3753 if executed in the loop pipeline.
3755 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3757 * doc/configfiles.texi (Configuration Files): Move info about
3758 conditionalizing $target-protos.h to...
3759 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3760 differs from $target-protos.h.
3762 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3763 Chung-Ju Wu <jasonwucj@gmail.com>
3765 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3766 * config/nds32/nds32-multiple.md (setmemsi): Define.
3767 * config/nds32/nds32-memory-manipulation.c
3768 (nds32_gen_dup_4_byte_to_word_value): New.
3769 (emit_setmem_word_loop): New.
3770 (emit_setmem_byte_loop): New.
3771 (nds32_expand_setmem_loop): New.
3772 (nds32_expand_setmem_loop_v3m): New.
3773 (nds32_expand_setmem_unroll): New.
3774 (nds32_expand_setmem): New.
3776 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3777 Chung-Ju Wu <jasonwucj@gmail.com>
3779 * config/nds32/nds32-memory-manipulation.c
3780 (nds32_emit_load_store): New.
3781 (nds32_emit_post_inc_load_store): New.
3782 (nds32_emit_mem_move): New.
3783 (nds32_emit_mem_move_block): New.
3784 (nds32_expand_movmemsi_loop_unknown_size): New.
3785 (nds32_expand_movmemsi_loop_known_size): New.
3786 (nds32_expand_movmemsi_loop): New.
3787 (nds32_expand_movmemsi_unroll): New.
3788 (nds32_expand_movmemqi): Rename ...
3789 (nds32_expand_movmemsi): ... to this.
3790 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3791 (movmemsi): ... to this.
3792 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3793 (nds32_expand_movmemsi): ... to this.
3795 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3796 Monk Chiang <sh.chiang04@gmail.com>
3797 Chung-Ju Wu <jasonwucj@gmail.com>
3799 * config/nds32/nds32-protos.h
3800 (nds32_expand_load_multiple): New arguments.
3801 (nds32_expand_store_multiple): Ditto.
3802 (nds32_valid_multiple_load_store): Rename ...
3803 (nds32_valid_multiple_load_store_p): ... to this.
3804 * config/nds32/nds32-memory-manipulation.c
3805 (nds32_expand_load_multiple): Refine implementation.
3806 (nds32_expand_store_multiple): Ditto.
3807 * config/nds32/nds32-multiple.md
3808 (load_multiple): Update nds32_expand_load_multiple interface.
3809 (store_multiple): Update nds32_expand_store_multiple interface.
3810 * config/nds32/nds32-predicates.c
3811 (nds32_valid_multiple_load_store): Rename ...
3812 (nds32_valid_multiple_load_store_p): ... to this and refine
3814 * config/nds32/predicates.md
3815 (nds32_load_multiple_and_update_address_operation): New predicate.
3816 (nds32_store_multiple_and_update_address_operation): New predicate.
3818 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3819 Chung-Ju Wu <jasonwucj@gmail.com>
3821 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3822 (combo): New attribute.
3823 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3825 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3827 * config/nds32/nds32.opt: Change -mcmodel= default value.
3829 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3830 Monk Chiang <sh.chiang04@gmail.com>
3831 Chung-Ju Wu <jasonwucj@gmail.com>
3833 * config/nds32/constants.md (unspec_element): New enum.
3834 * config/nds32/constraints.md (Umw): New constraint.
3835 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3836 * config/nds32/nds32-intrinsic.md: Likewise.
3837 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3838 (nds32_valid_smw_lwm_base_p): New.
3839 (nds32_output_smw_single_word): New.
3840 (nds32_output_lmw_single_word): New.
3841 (nds32_expand_unaligned_load): New.
3842 (nds32_expand_unaligned_store): New.
3843 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3844 (nds32_output_smw_single_word): Declare.
3845 (nds32_output_lmw_single_word): Declare.
3846 (nds32_expand_unaligned_load): Declare.
3847 (nds32_expand_unaligned_store): Declare.
3848 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3849 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3850 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3851 NDS32_BUILTIN_UASTORE_DW.
3852 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3855 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3856 Kito Cheng <kito.cheng@gmail.com>
3857 Chung-Ju Wu <jasonwucj@gmail.com>
3859 * config/nds32/nds32-intrinsic.c
3860 (nds32_expand_builtin_null_ftype_reg): Delete.
3861 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3862 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3863 (nds32_read_argument): New.
3864 (nds32_legitimize_target): Ditto.
3865 (nds32_legitimize_argument): Ditto.
3866 (nds32_check_constant_argument): Ditto.
3867 (nds32_expand_unop_builtin): Ditto.
3868 (nds32_expand_unopimm_builtin): Ditto.
3869 (nds32_expand_binop_builtin): Ditto.
3870 (nds32_builtin_decl_impl): Ditto.
3871 (builtin_description): Ditto.
3872 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3873 (nds32_init_builtins_impl): Ditto.
3874 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3875 (nds32_builtin_decl): New.
3876 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3877 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3879 2018-03-02 Jeff Law <law@redhat.com>
3881 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3882 (redundant_insn, fill_simple_delay_slots): Likewise.
3883 (fill_slots_from_thread): Likewise.
3884 * resource.c (mark_referenced_resources): Likewise.
3885 (mark_set_resources, find_dead_or_set_registers): Likewise.
3887 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3889 * substring-locations.h (format_warning_va): Formatting fix for
3891 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3893 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3894 * substring-locations.c: Include intl.h.
3895 (format_warning_va): Turned into small wrapper around
3896 format_warning_n_va, renamed to ...
3897 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3898 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3900 (format_warning_at_substring_n): New function.
3901 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3902 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3903 format_warning_at_substring with just a shorter name instead of
3904 const function pointer.
3905 (fmtwarn_n): New function.
3906 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3907 appropriate, get rid of all the fmtstr temporaries, move conditionals
3908 with G_() wrapped string literals directly into fmtwarn arguments,
3909 cast dir.len to (int), formatting fixes.
3911 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3913 * doc/invoke.texi: Remove "Cilk Plus" references.
3915 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3916 Richard Biener <rguenther@suse.de>
3919 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3920 for error or warning attributes if CALL_FROM_THUNK_P is set.
3923 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3926 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3927 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3930 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3931 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3933 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3934 instead of -1U in last predictors element's probability member.
3936 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3939 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3940 arguments if they are comparable.
3942 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3944 PR tree-optimization/84634
3945 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3946 masks and masked_loop_p with a single loop_masks, making sure it's
3947 null for bb vectorization.
3949 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3951 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3952 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3953 loop->force_vectorize to check whether there is no alias.
3955 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3958 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3960 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3962 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3963 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3964 instead of a loop around prev_real_insn.
3965 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3969 * config/i386/i386.c (ix86_print_operand): Use conditional
3970 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3973 2018-03-02 Richard Biener <rguenther@suse.de>
3975 PR tree-optimization/84427
3976 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3977 (bitmap_set_subtract_values): Rewrite to handle multiple
3980 (prune_clobbered_mems): Likewise.
3981 (phi_translate): Take edge instead of pred/phiblock.
3982 (phi_translate_1): Likewise.
3983 (phi_translate_set): Likewise. Insert all translated
3984 exprs for a value into the set, keeping possibly multiple
3985 expressions per value.
3986 (compute_antic_aux): Adjust for phi_translate changes.
3987 When intersecting union the expressions and prune those
3988 not in the final value set, keeping possibly multiple
3989 expressions per value. Do not use value-insertion
3990 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3991 all expressions. Add verification that the value-sets
3992 only shrink during iteration.
3993 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3994 (do_pre_regular_insertion): Likewise.
3995 (do_pre_partial_partial_insertion): Likewise.
3997 2018-03-02 Richard Biener <rguenther@suse.de>
4000 * config/darwin.c (saved_debug_info_level): New static global.
4001 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4002 (darwin_asm_lto_end): Restore debug info generation settings.
4004 2018-03-01 Martin Liska <mliska@suse.cz>
4007 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4010 2018-03-01 Richard Biener <rguenther@suse.de>
4013 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4014 type annotation with LTO when debug was disabled at compile-time.
4016 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4018 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4020 (mips_final_postscan_insn): Likewise.
4022 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4024 PR rtl-optimization/84528
4025 * alias.c (init_alias_target): Add commentary.
4026 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4027 a unique base value if the frame pointer is not eliminated
4028 to the stack pointer.
4030 2018-03-01 Tom de Vries <tom@codesourcery.com>
4032 PR rtl-optimization/83327
4033 * lra-int.h (hard_regs_spilled_into): Declare.
4034 * lra.c (hard_regs_spilled_into): Define.
4035 (init_reg_info): Init hard_regs_spilled_into.
4036 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4037 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4038 (process_bb_lives): Handle hard_regs_spilled_into.
4039 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4040 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4042 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4044 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4045 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4046 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4047 * config/rs6000/aix72.h: New file.
4049 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4051 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4052 instead of warning_at with conditional singular and plural messages
4056 * stor-layout.c (update_alignment_for_field): For
4057 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4058 && !DECL_PACKED (field), do the alignment update, just use
4059 only desired_align instead of MAX (type_align, desired_align)
4061 (place_field): Don't do known_align < desired_align handling
4062 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4063 is non-NULL, instead do it after rli->prev_field handling and
4064 only if not within a bitfield word. For DECL_PACKED (field)
4065 use type_align of BITS_PER_UNIT.
4067 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4069 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4070 superfluous parentheses and trailing spaces.
4072 2018-02-28 Richard Biener <rguenther@suse.de>
4074 PR tree-optimization/84584
4075 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4076 SCoPs with fake exit edge.
4078 2018-02-28 Martin Liska <mliska@suse.cz>
4081 * timevar.c (timer::print): Fix format to properly print 100%
4084 2018-02-28 Richard Biener <rguenther@suse.de>
4087 * genmatch.c (capture_info::walk_match): Do not mark
4088 captured expressions without operands as expr_p given
4089 they act more like predicates and should be subject to
4090 "lost tail" side-effect preserving.
4092 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4094 PR rtl-optimization/81611
4095 * auto-inc-dec.c (attempt_change): Move dead note from
4096 mem_insn if it's the next use of regno
4097 (find_address): Take address use of reg holding
4098 non-incremented value. Add parm to limit search to the named
4100 (merge_in_block): Attempt to use a mem insn that is the next
4101 use of the original regno.
4103 2018-02-27 Martin Sebor <msebor@redhat.com>
4106 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4107 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4109 2018-02-27 Martin Sebor <msebor@redhat.com>
4111 PR translation/84207
4112 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4113 n argument to unsigned HOST_WIDE_INT.
4114 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4115 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4116 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4117 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4119 2018-02-27 Richard Biener <rguenther@suse.de>
4121 PR tree-optimization/84512
4122 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4123 Do not use the estimate returned from record_stmt_cost for
4124 the scalar iteration cost but sum properly using add_stmt_cost.
4126 2018-02-27 Richard Biener <rguenther@suse.de>
4128 PR tree-optimization/84466
4129 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4130 Adjust last change to less strictly validate use operands.
4132 2018-02-27 Martin Liska <mliska@suse.cz>
4134 PR gcov-profile/84548
4135 * gcov.c (process_file): Allow partial overlap and consider it
4136 also as group functions.
4137 (output_lines): Properly calculate range of lines for a group.
4139 2018-02-27 Martin Liska <mliska@suse.cz>
4141 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4142 'ggc' suffixes. Change first column width.
4143 (timer::print): Fix formatting of the column.
4145 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4147 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4148 preserve inline entry blocks for the sake of debug inline
4149 entry point markers alone.
4150 (remove_unused_locals): Suggest in comments a better place to
4151 force the preservation of inline entry blocks that are
4152 otherwise unused, but do not preserve them.
4154 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4156 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4158 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4161 * config/i386/constraints.md (Bs): Replace
4162 ix86_indirect_branch_register with
4163 TARGET_INDIRECT_BRANCH_REGISTER.
4165 * config/i386/i386.md (indirect_jump): Likewise.
4166 (tablejump): Likewise.
4167 (*sibcall_memory): Likewise.
4168 (*sibcall_value_memory): Likewise.
4169 Peepholes of indirect call and jump via memory: Likewise.
4170 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4171 (*sibcall_value_GOT_32): Likewise.
4172 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4173 (GOT_memory_operand): Likewise.
4174 (call_insn_operand): Likewise.
4175 (sibcall_insn_operand): Likewise.
4176 (GOT32_symbol_operand): Likewise.
4177 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4179 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4181 PR rtl-optimization/83496
4182 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4183 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4184 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4185 redundant insn, if any.
4186 (relax_delay_slots): Likewise.
4187 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4189 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4191 PR tree-optimization/83965
4192 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4193 that grouped statements are part of a reduction chain. Return
4194 true if the statement is not marked as a reduction itself but
4196 (vect_recog_dot_prod_pattern): Don't check whether the statement
4197 is part of a group here.
4198 (vect_recog_sad_pattern): Likewise.
4199 (vect_recog_widen_sum_pattern): Likewise.
4201 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4204 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4206 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4209 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4211 (ix86_output_indirect_function_return): New prototype.
4212 (ix86_split_simple_return_pop_internal): Likewise.
4213 * config/i386/i386.c (indirect_return_via_cx): New.
4214 (indirect_return_via_cx_bnd): Likewise.
4215 (indirect_thunk_name): Handle return va CX_REG.
4216 (output_indirect_thunk_function): Create alias for
4217 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4218 (ix86_output_indirect_jmp): Remove the bool argument.
4219 (ix86_output_indirect_function_return): New function.
4220 (ix86_split_simple_return_pop_internal): Likewise.
4221 * config/i386/i386.md (*indirect_jump): Don't pass false
4222 to ix86_output_indirect_jmp.
4223 (*tablejump_1): Likewise.
4224 (simple_return_pop_internal): Change it to define_insn_and_split.
4225 Call ix86_split_simple_return_pop_internal to split it for
4227 (simple_return_indirect_internal): Call
4228 ix86_output_indirect_function_return instead of
4229 ix86_output_indirect_jmp.
4231 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4234 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4235 memset and value initialization afterwards.
4237 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4239 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4241 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4244 * common/config/aarch64/aarch64-common.c
4245 (aarch_option_optimization_table[]): Switch
4246 off fomit-frame-pointer
4248 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4249 Chung-Ju Wu <jasonwucj@gmail.com>
4251 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4253 (store_multiple): Ditto.
4255 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4257 * config.gcc: Add --with-cpu support for nds32 target.
4258 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4259 * config/nds32/nds32.opt: Add -mcpu= option.
4261 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4263 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4264 isel=yes): Warn for these deprecated options.
4266 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4268 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4269 ISA_2_5_MASKS_EMBEDDED.
4271 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4273 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4274 p->max as pointers rather than using iterative_hash_expr.
4276 2018-02-23 Carl Love <cel@us.ibm.com>
4278 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4279 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4281 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4282 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4283 P8V_BUILTIN_VEC_VUNSIGNED2.
4285 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4288 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4289 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4291 * lra-constraints.c (curr_insn_transform): Set up
4292 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4294 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4295 * lra-eliminations.c (spill_pseudos): Ditto.
4296 (process_insn_for_elimination): Ditto.
4297 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4298 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4299 LRA_NON_CLOBBERED_ALT.
4301 2018-02-22 Martin Sebor <msebor@redhat.com>
4303 PR tree-optimization/84480
4304 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4305 to maybe_diag_stxncpy_trunc. Call it.
4306 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4307 from gimple_fold_builtin_strcpy. Print inlining stack.
4308 (handle_builtin_stxncpy): Print inlining stack.
4309 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4311 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4314 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4315 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4316 and -fcheck-pointer-bounds are used together.
4317 (indirect_thunk_prefix): New enum.
4318 (indirect_thunk_need_prefix): New function.
4319 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4320 "_nt" instead of "_bnd" for NOTRACK prefix.
4321 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4322 (output_indirect_thunk_function): Likewise.
4324 (ix86_code_end): Update output_indirect_thunk_function calls.
4325 (ix86_output_indirect_branch_via_reg): Replace
4326 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4327 (ix86_output_indirect_branch_via_push): Likewise.
4328 (ix86_output_function_return): Likewise.
4329 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4330 incompatible with -fcf-protection=branch and
4331 -fcheck-pointer-bounds.
4333 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4336 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4337 Change gcc_assert call to output_operand_lossage.
4339 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4341 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4343 2018-02-22 DJ Delorie <dj@redhat.com>
4344 Sebastian Perta <sebastian.perta@renesas.com>
4345 Oleg Endo <olegendo@gcc.gnu.org>
4347 * config/rx/rx.c (rx_rtx_costs): New function.
4348 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4350 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4352 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4354 2018-02-22 Martin Liska <mliska@suse.cz>
4357 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4358 Add "native" as a possible value.
4360 2018-02-22 Martin Liska <mliska@suse.cz>
4363 * config/i386/i386.c (ix86_option_override_internal):
4364 Add "native" as a possible value for -march and -mtune.
4366 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4369 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4370 to all type variants.
4372 PR tree-optimization/84503
4373 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4374 width as info->bitpos + info->bitsize - start.
4375 (merged_store_group::merge_overlapping): Simplify width computation.
4376 (check_no_overlap): New function.
4377 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4378 start + width and last_order of the group, fail if check_no_overlap
4380 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4381 to group if check_no_overlap fails.
4383 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4385 * config/rs6000/altivec.md: Delete contraint arguments to
4386 define_expand, define_split, and define_peephole2, and in
4387 define_insn_and_split if always unused.
4388 * config/rs6000/darwin.md: Ditto.
4389 * config/rs6000/dfp.md: Ditto.
4390 * config/rs6000/rs6000.md: Ditto.
4391 * config/rs6000/sync.md: Ditto.
4392 * config/rs6000/vector.md: Ditto.
4393 * config/rs6000/vsx.md: Ditto.
4395 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4397 * config/rs6000/altivec.md: Write output control strings as braced
4398 blocks instead of double-quoted strings.
4399 * config/rs6000/darwin.md: Ditto.
4400 * config/rs6000/rs6000.md: Ditto.
4401 * config/rs6000/vector.md: Ditto.
4402 * config/rs6000/vsx.md: Ditto.
4404 2018-02-21 Jason Merrill <jason@redhat.com>
4406 PR c++/84314 - ICE with templates and fastcall attribute.
4407 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4409 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4411 * ipa-cp.c (determine_versionability): Fix comment typos.
4413 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4416 * ipa-cp.c (determine_versionability): Do not version functions caling
4419 2018-02-21 Martin Liska <mliska@suse.cz>
4422 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4423 Add "native" as a possible value.
4424 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4425 the macro when native cpu detection is available.
4427 2018-02-21 Martin Liska <mliska@suse.cz>
4430 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4431 Add "native" as a possible value.
4432 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4433 when native cpu detection is available.
4435 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4436 Martin Sebor <msebor@redhat.com>
4438 PR tree-optimization/84478
4439 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4441 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4442 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4443 support which is conservatively correct, for 2 only stay conservative
4444 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4445 argument to the 2 argument get_range_strlen, adjust 6 arg
4446 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4448 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4449 (gimple_fold_builtin_strlen): Pass true as last argument to
4452 2018-02-20 Martin Sebor <msebor@redhat.com>
4455 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4456 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4457 (builtin_memref::builtin_memref): Factor out parts into
4458 set_base_and_offset and call it.
4460 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4463 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4464 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4465 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4466 search at the associated MODE_INT.
4468 2018-02-20 Jeff Law <law@redhat.com>
4471 PR tree-optimization/81592
4473 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4474 for range data rather than using global data.
4475 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4476 range data rather than using global data.
4477 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4478 pass it to children as needed.
4479 (struct directive::fmtresult): Similarly.
4480 (struct directive::set_width): Similarly.
4481 (struct directive::set_precision): Similarly.
4482 (format_integer, format_directive, parse_directive): Similarly.
4483 (format_none): Accept unnamed vr_values parameter.
4484 (format_percent, format_floating, format_character): Similarly.
4485 (format_string, format_plain): Similarly.
4486 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4487 the EVRP range analyzer for range data rather than using global data.
4488 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4489 gimple-ssa-evrp-analyze.h
4490 (class sprintf_dom_walker): Add after_dom_children member function.
4491 Add evrp_range_analyzer member.
4492 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4493 range analyzer as needed.
4494 (sprintf_dom_walker::after_dom_children): New member function.
4495 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4497 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4498 (evrp_range_analyzer::pop_to_marker): Likewise.
4500 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4502 PR tree-optimization/84419
4503 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4504 with the required type if its current type is compatible but
4507 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4510 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4511 after vectorization.
4513 2018-02-20 Martin Liska <mliska@suse.cz>
4516 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4517 possible values if we don't have a hint.
4519 2018-02-20 Martin Liska <mliska@suse.cz>
4523 * final.c (shorten_branches): Build align_tab array with one
4525 * opts.c (finish_options): Add alignment option limit check.
4526 (MAX_CODE_ALIGN): Likewise.
4527 (MAX_CODE_ALIGN_VALUE): Likewise.
4528 * doc/invoke.texi: Document maximum allowed option value for
4529 all -falign-* options.
4531 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4534 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4535 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4536 * var-tracking.c (emit_note_insn_var_location): Remove all references
4537 to NOTE_INSN_CALL_ARG_LOCATION.
4538 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4539 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4540 Use copy_rtx_if_shared.
4541 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4542 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4543 (dwarf2out_var_location): Remove handling of
4544 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4546 * final.c (final_scan_insn): Remove all references to
4547 NOTE_INSN_CALL_ARG_LOCATION.
4548 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4549 before dumping final insns.
4550 * except.c (emit_note_eh_region_end): Remove all references to
4551 NOTE_INSN_CALL_ARG_LOCATION.
4552 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4553 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4554 * config/arc/arc.c (hwloop_optimize): Likewise.
4555 * config/arm/arm.c (create_fix_barrier): Likewise.
4556 * config/s390/s390.c (s390_chunkify_start): Likewise.
4557 * config/sh/sh.c (find_barrier): Likewise.
4558 * config/i386/i386.c (rest_of_insert_endbranch,
4559 ix86_seh_fixup_eh_fallthru): Likewise.
4560 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4561 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4562 * config/frv/frv.c (frv_function_prologue): Likewise.
4563 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4565 (note_outside_basic_block_p): Remove all references to
4566 NOTE_INSN_CALL_ARG_LOCATION.
4567 * gengtype.c (adjust_field_rtx_def): Likewise.
4568 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4570 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4571 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4574 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4577 PR tree-optimization/84452
4578 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4579 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4582 2018-02-19 Martin Liska <mliska@suse.cz>
4585 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4587 2018-02-19 Martin Liska <mliska@suse.cz>
4588 Richard Sandiford <richard.sandiford@linaro.org>
4590 PR tree-optimization/82491
4591 * gimple-fold.c (get_base_constructor): Make earlier bail out
4594 2018-02-19 Carl Love <cel@us.ibm.com>
4596 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4597 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4599 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4600 P8V_BUILTIN_VEC_NEG.
4602 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4604 * config/rl78/rl78.md (movdf): New define expand.
4606 2018-02-19 Martin Liska <mliska@suse.cz>
4609 * doc/invoke.texi: Fix typo.
4610 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4612 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4614 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4615 handle rs6000_single_float and rs6000_double_float specially for
4618 2018-02-16 Jeff Law <law@redhat.com>
4620 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4623 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4626 * ipa-inline.c (inline_small_functions): Fix a typo.
4628 2018-02-16 Nathan Sidwell <nathan@acm.org>
4630 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4632 2018-02-16 Carl Love <cel@us.ibm.com>
4634 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4635 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4636 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4637 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4638 expansion to P8V_BUILTIN_VEC_FLOAT2.
4640 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4642 PR rtl-optimization/70023
4643 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4644 src_regno into account.
4646 2018-02-16 Carl Love <cel@us.ibm.com>
4648 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4649 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4650 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4651 * config/rs6000/rs6000.c: Remove case statements for
4652 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4653 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4654 and P9V_BUILTIN_VEC_VINSERT4B.
4655 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4656 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4657 * config/rs6000/vsx.md:
4658 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4661 2018-02-16 Carl Love <cel@us.ibm.com>
4663 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4665 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4667 * config/rs6000/rs6000-c.c: Add the definitions for
4668 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4669 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4670 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4671 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4672 definition for insert4b and define insn *insert3b_internal.
4673 * doc/extend.texi: Add documentation for vec_extract4b.
4675 2018-02-16 Nathan Sidwell <nathan@acm.org>
4677 * doc/extend.texi (Backwards Compatibility): Mention friend
4678 injection. Note for-scope is deprecated.
4679 * doc/invoke.texi (-ffriend-injection): Deprecate.
4681 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4683 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4684 that moved to I2, also allow destinations that are a paradoxical
4685 subreg (instead of a normal reg).
4687 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4690 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4693 2018-02-16 Richard Biener <rguenther@suse.de>
4695 PR tree-optimization/84037
4696 PR tree-optimization/84016
4698 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4699 Adjust vec_construct for the fact we need additional higher latency
4700 128bit inserts for AVX256 and AVX512 vector builds.
4701 (ix86_add_stmt_cost): Scale vector construction cost for
4704 2018-02-16 Richard Biener <rguenther@suse.de>
4706 PR tree-optimization/84417
4707 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4708 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4709 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4711 2018-02-16 Martin Liska <mliska@suse.cz>
4714 * internal-fn.def (ASAN_CHECK): Set proper flags.
4715 (ASAN_MARK): Likewise.
4717 2018-02-16 Julia Koval <julia.koval@intel.com>
4719 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4720 from PTA_CANNONLAKE.
4722 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4725 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4726 Use ++iter rather than iter++ for std::list iterators.
4727 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4728 defer deleting them until all nodes in the forest are processed. Do
4729 free even leaf nodes. Change to_process into auto_vec.
4732 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4733 * vec.h (vec_default_construct): Use memset instead of placement new
4734 if BROKEN_VALUE_INITIALIZATION is defined.
4735 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4736 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4739 PR rtl-optimization/83723
4740 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4741 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4742 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4744 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4746 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4748 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4750 PR rtl-optimization/81443
4751 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4752 from inner REGs to paradoxical SUBREGs.
4754 2018-02-16 Richard Biener <rguenther@suse.de>
4756 PR tree-optimization/84399
4757 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4758 For operands we can analyze at their definition make sure we can
4759 analyze them at each use as well.
4761 2018-02-16 Richard Biener <rguenther@suse.de>
4763 PR tree-optimization/84190
4764 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4765 volatile accesses if the decl isn't volatile.
4767 2018-02-15 Jason Merrill <jason@redhat.com>
4769 PR c++/84314 - ICE with templates and fastcall attribute.
4770 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4771 TYPE_CANONICAL on an existing type.
4773 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4775 PR tree-optimization/84383
4776 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4777 dstoff nor call operand_equal_p if dstbase is NULL.
4779 PR tree-optimization/84334
4780 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4781 also a CONSTANT_CLASS_P, punt.
4783 2018-02-14 Jim Wilson <jimw@sifive.com>
4785 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4786 first SMALL_OPERAND check. New local min_second_step. Move assert
4787 to where locals are set. Add TARGET_RVC support.
4788 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4790 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4792 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4794 2018-02-14 Martin Sebor <msebor@redhat.com>
4796 PR tree-optimization/83698
4797 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4798 arrays constrain the offset range to their bounds.
4799 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4800 (builtin_access::overlap): Avoid setting the size of overlap if it's
4802 (maybe_diag_overlap): Also consider arrays when deciding what values
4803 of offsets to include in diagnostics.
4805 2018-02-14 Martin Sebor <msebor@redhat.com>
4808 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4809 that correspond to the kind of a declaration.
4811 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4814 * config/pa/pa.md: Load address of PIC label using the linkage table
4815 if the label is nonlocal.
4817 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4819 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4820 warning message if user requests -maltivec=be.
4821 * doc/invoke.texi: Document deprecation of -maltivec=be.
4823 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4826 * config/rs6000/rs6000-c.c: Update definitions for
4827 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4828 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4830 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4833 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4834 add _get_ssp intrinsics. Remove argument from
4835 __builtin_ia32_rdssp[d|q].
4836 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4837 * config/i386/i386-builtin.def: Remove argument from
4838 __builtin_ia32_rdssp[d|q].
4839 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4840 ix86_expand_special_args_builtin for _rdssp[d|q].
4841 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4842 Clear register before usage.
4843 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4844 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4846 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4848 PR tree-optimization/84357
4849 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4850 operand 1 of an ARRAY_REF too.
4852 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4855 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4856 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4858 (set_of_reg): New struct.
4859 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4860 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4861 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4863 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4864 Split into bitclr, bitset, bitinvert patterns if appropriate.
4865 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4866 use rx_fuse_in_memory_bitop.
4867 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4868 to named insn, correct maximum insn length.
4870 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4873 * machmode.def: Define a complex mode for PARTIAL_INT.
4874 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4876 * doc/rtl.texi: Document CSPImode.
4877 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4879 (msp430_hard_regno_nregs_with_padding): Likewise.
4881 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4884 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4886 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4888 PR rtl-optimization/84169
4889 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4890 we generated a parallel as new i3 and we split that to new i2 and i3
4891 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4892 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4893 those to i2, not i1. Partially rewrite this scan code.
4895 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4898 * stor-layout.c (place_field): For variable length fields, adjust
4899 offset_align afterwards not just based on the field's alignment,
4900 but also on the size.
4903 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4904 of exps and logs in the use_exp2 case.
4906 2018-02-13 Jeff Law <law@redhat.com>
4908 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4911 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4914 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4918 * common.opt (gas-loc-support, gas-locview-support): New.
4919 (ginline-points, ginternal-reset-location-views): New.
4920 * doc/invoke.texi: Document them. Use @itemx where intended.
4921 (gvariable-location-views): Adjust.
4922 * target.def (reset_location_view): New.
4923 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4924 (TARGET_RESET_LOCATION_VIEW): New.
4925 * doc/tm.texi: Rebuilt.
4926 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4927 (dwarf2out_default_as_locview_support): New.
4928 (output_asm_line_debug_info): Use option variables.
4929 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4930 (output_loc_list): Likewise.
4931 (add_high_low_attributes): Check option variables.
4932 Don't output entry view attribute in strict mode.
4933 (gen_inlined_subroutine_die): Check option variables.
4934 (dwarf2out_inline_entry): Likewise.
4935 (init_sections_and_labels): Likewise.
4936 (dwarf2out_early_finish): Likewise.
4937 (maybe_reset_location_view): New, from...
4938 (dwarf2out_var_location): ... here. Call it.
4939 * debug.h (dwarf2out_default_as_loc_support): Declare.
4940 (dwarf2out_default_as_locview_support): Declare.
4941 * hooks.c (hook_int_rtx_insn_0): New.
4942 * hooks.h (hook_int_rtx_insn_0): Declare.
4943 * toplev.c (process_options): Take -gas-loc-support and
4944 -gas-locview-support from dwarf2out. Enable
4945 -gvariable-location-views by default only with locview
4946 assembler support. Enable -ginternal-reset-location-views by
4947 default only if the target defines the corresponding hook.
4948 Enable -ginline-points by default if location views are
4949 enabled; force it disabled if statement frontiers are
4951 * tree-inline.c (expand_call_inline): Check option variables.
4952 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4954 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4956 PR tree-optimization/84321
4957 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4958 handling. Also check whether the anti-range contains any values
4959 that satisfy the mask; switch to a VR_RANGE if not.
4961 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4964 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4966 2018-02-13 Martin Jambor <mjambor@suse.cz>
4969 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4970 of call statements, also set location of a load to a temporary.
4972 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4974 * config/rl78/rl78.c (add_vector_labels): New function.
4975 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4976 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4977 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4978 which checks that no arguments are passed.
4979 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4980 * doc/extend.texi: Documentation for the new attribute.
4982 2018-02-13 Andreas Schwab <schwab@suse.de>
4984 * config/riscv/linux.h (CPP_SPEC): Define.
4986 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4989 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4990 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4991 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4992 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4993 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4994 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4995 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4996 temporarily for AES and PCLMUL builtins.
4998 PR tree-optimization/84339
4999 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5000 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5004 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5005 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5006 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5008 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5010 * omp-simd-clone.h: New file.
5011 * omp-simd-clone.c: Include omp-simd-clone.h.
5012 (expand_simd_clones): No longer static.
5013 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5014 cgraph.h and omp-simd-clone.h.
5015 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5016 (vect_recog_widen_shift_pattern): Formatting fix.
5017 (vect_pattern_recog_1): Don't check optab for calls.
5020 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5021 operands[2] into a REG before using gen_lowpart on it.
5023 2018-02-12 Jeff Law <law@redhat.com>
5026 * config/sh/sh.c (find_barrier): Consider a sibling call
5029 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5030 successfully back substituting a reg.
5032 2018-02-12 Richard Biener <rguenther@suse.de>
5034 PR tree-optimization/84037
5035 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5036 parameter, move visited init to caller.
5037 (vect_slp_analyze_operations): Separate cost from validity
5038 check, initialize visited once for all instances.
5039 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5041 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5043 (vect_model_promotion_demotion_cost): Likewise.
5044 (vectorizable_bswap): Guard cost modeling with !slp_node
5045 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5047 (vectorizable_call): Likewise.
5048 (vectorizable_conversion): Likewise.
5049 (vectorizable_assignment): Likewise.
5050 (vectorizable_shift): Likewise.
5051 (vectorizable_operation): Likewise.
5052 (vectorizable_store): Likewise.
5053 (vectorizable_load): Likewise.
5054 (vectorizable_condition): Likewise.
5055 (vectorizable_comparison): Likewise.
5057 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5060 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5061 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5062 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5064 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5067 * params.def (inline-min-speedup): Increase from 8 to 15.
5068 (max-inline-insns-auto): Decrease from 40 to 30.
5069 * ipa-split.c (consider_split): Add some buffer for function to
5070 be considered inlining candidate.
5071 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5074 2018-02-12 Richard Biener <rguenther@suse.de>
5076 PR tree-optimization/84037
5077 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5078 matched stmts if we cannot swap the non-matched ones.
5080 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5082 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5083 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5084 _mm_maskz_scalef_round_ss): New intrinsics.
5085 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5086 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5087 __builtin_ia32_scalefss_round): Remove.
5088 (__builtin_ia32_scalefsd_mask_round,
5089 __builtin_ia32_scalefss_mask_round): New intrinsics.
5090 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5091 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5092 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5093 "<round_constraint>")): Changed to ...
5094 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5095 "<round_scalar_constraint>")): ... this.
5096 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5097 %0, %1, %2<round_op3>}"): Changed to ...
5098 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5099 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5100 %2<round_scalar_mask_op3>}"): ... this.
5101 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5103 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5105 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5106 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5107 (_mm_maskz_sqrt_round_ss): New intrinsics.
5108 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5109 (__builtin_ia32_sqrtsd_mask_round)
5110 (__builtin_ia32_sqrtss_mask_round): New builtins.
5111 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5112 (__builtin_ia32_sqrtss_round): Remove.
5113 (__builtin_ia32_sqrtsd_mask_round)
5114 (__builtin_ia32_sqrtss_mask_round): New builtins.
5115 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5116 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5117 ((match_operand:VF_128 1 "vector_operand"
5118 "xBm,<round_constraint>")): Changed to ...
5119 ((match_operand:VF_128 1 "vector_operand"
5120 "xBm,<round_scalar_constraint>")): ... this.
5121 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5122 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5123 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5124 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5125 %<iptr>1<round_scalar_mask_op3>}): ... this.
5126 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5127 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5129 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5132 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5133 Cast vec_cmpeq result to correct type.
5134 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5135 Cast vec_cmpgt result to correct type.
5137 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5139 * final.c (final_scan_insn_1): Renamed from...
5140 (final_scan_insn): ... this. New wrapper, to recover
5141 seen from the outermost call in recursive ones.
5142 * config/sparc/sparc.c (output_return): Drop seen from call.
5143 (output_sibcall): Likewise.
5144 * config/visium/visium.c (output_branch): Likewise.
5146 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5148 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5151 2018-02-10 Alan Modra <amodra@gmail.com>
5154 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5155 Specify LR as an input.
5157 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5160 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5161 remove_member_access_dummy_vars): New functions.
5162 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5163 lower_omp_1, execute_lower_omp): Use them.
5165 PR rtl-optimization/84308
5166 * shrink-wrap.c (spread_components): Release todo vector.
5168 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5170 PR rtl-optimization/57193
5171 * ira-color.c (struct allocno_color_data): Add member
5172 conflict_allocno_hard_prefs.
5173 (update_conflict_allocno_hard_prefs): New.
5174 (bucket_allocno_compare_func): Add a preference based on
5175 conflict_allocno_hard_prefs.
5176 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5177 (color_allocnos): Remove a dead code. Initiate
5178 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5180 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5183 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5184 constraint from =wa to wa. Avoid a subreg on the output operand,
5185 instead use a pseudo and subreg it in a move.
5186 (p9_xxbrd_<mode>): Changed to ...
5187 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5188 (p9_xxbrd_v2df): New expander.
5189 (p9_xxbrw_<mode>): Changed to ...
5190 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5191 (p9_xxbrw_v4sf): New expander.
5193 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5195 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5197 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5200 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5201 multiply in 32-bit mode.
5202 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5203 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5206 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5208 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5209 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5210 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5211 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5213 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5216 * dwarf2out.c (is_trivial_indirect_ref): New function.
5217 (dwarf2out_late_global_decl): Do not generate a location
5218 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5219 and that are not defined in the current unit.
5221 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5223 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5224 instead of a libcall for UNORDERED.
5226 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5229 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5230 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5232 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5235 * config/s390/s390.c (s390_set_current_function): Invoke
5236 s390_indirect_branch_settings also if fndecl didn't change.
5238 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5240 * config/rs6000/rs6000.md (blockage): Set length to zero.
5242 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5244 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5246 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5249 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5250 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5254 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5255 PARALLEL incoming that failed vt_get_decl_and_offset check.
5258 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5259 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5261 (get_variable_section): For decls in named .bss* sections pass true as
5262 second argument to bss_initializer_p.
5264 2018-02-09 Marek Polacek <polacek@redhat.com>
5265 Jakub Jelinek <jakub@redhat.com>
5268 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5269 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5270 Sync some changes from cxx_fold_indirect_ref.
5272 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5274 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5276 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5277 (BLOCK_INLINE_ENTRY_LABEL): New.
5278 (dwarf2out_var_location): Disregard inline entry markers.
5279 (inline_entry_data): New struct.
5280 (inline_entry_data_hasher): New hashtable type.
5281 (inline_entry_data_hasher::hash): New.
5282 (inline_entry_data_hasher::equal): New.
5283 (inline_entry_data_table): New variable.
5284 (add_high_low_attributes): Add DW_AT_entry_pc and
5285 DW_AT_GNU_entry_view attributes if a pending entry is found
5286 in inline_entry_data_table. Add old entry_pc attribute only
5287 if debug nonbinding markers are disabled.
5288 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5289 markers are enabled.
5290 (block_within_block_p, dwarf2out_inline_entry): New.
5291 (dwarf2out_finish): Check that no entries remained in
5292 inline_entry_data_table.
5293 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5294 (final_scan_insn, notice_source_line): Likewise.
5295 (rest_of_clean_state): Skip inline entry markers.
5296 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5298 * gimple.c (gimple_build_debug_inline_entry): New.
5299 * gimple.h (enum gimple_debug_subcode): Add
5300 GIMPLE_DEBUG_INLINE_ENTRY.
5301 (gimple_build_debug_inline_entry): Declare.
5302 (gimple_debug_inline_entry_p): New.
5303 (gimple_debug_nonbind_marker_p): Adjust.
5304 * insn-notes.def (INLINE_ENTRY): New.
5305 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5306 inline entry marker notes.
5307 (print_insn): Likewise.
5308 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5309 (INSN_DEBUG_MARKER_KIND): Likewise.
5310 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5311 * tree-inline.c (expand_call_inline): Build and insert
5312 debug_inline_entry stmt.
5313 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5314 inline entry blocks early, if nonbind markers are enabled.
5315 (dump_scope_block): Dump fragment info.
5316 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5317 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5318 (gimple_build_debug_inline_entry): New.
5319 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5320 Enable/disable inline entry points too.
5321 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5322 (DEBUG_INSN): Describe inline entry markers.
5324 * common.opt (gvariable-location-views): New.
5325 (gvariable-location-views=incompat5): New.
5326 * config.in: Rebuilt.
5327 * configure: Rebuilt.
5328 * configure.ac: Test assembler for view support.
5329 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5330 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5331 * dwarf2out.c (var_loc_view): New typedef.
5332 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5333 (dwarf2out_locviews_in_attribute): New.
5334 (dwarf2out_locviews_in_loclist): New.
5335 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5336 (enum dw_line_info_opcode): Add LI_adv_address.
5337 (struct dw_line_info_table): Add view.
5338 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5339 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5340 (zero_view_p): New variable.
5341 (ZERO_VIEW_P): New macro.
5342 (output_asm_line_debug_info): New.
5343 (struct var_loc_node): Add view.
5344 (add_AT_view_list, AT_loc_list): New.
5345 (add_var_loc_to_decl): Add view param. Test it against last.
5346 (new_loc_list): Add view params. Record them.
5347 (AT_loc_list_ptr): Handle loc and view lists.
5348 (view_list_to_loc_list_val_node): New.
5349 (print_dw_val): Handle dw_val_class_view_list.
5350 (size_of_die): Likewise.
5351 (value_format): Likewise.
5352 (loc_list_has_views): New.
5353 (gen_llsym): Set vl_symbol too.
5354 (maybe_gen_llsym, skip_loc_list_entry): New.
5355 (dwarf2out_maybe_output_loclist_view_pair): New.
5356 (output_loc_list): Output view list or entries too.
5357 (output_view_list_offset): New.
5358 (output_die): Handle dw_val_class_view_list.
5359 (output_dwarf_version): New.
5360 (output_compilation_unit_header): Use it.
5361 (output_skeleton_debug_sections): Likewise.
5362 (output_rnglists, output_line_info): Likewise.
5363 (output_pubnames, output_aranges): Update version comments.
5364 (output_one_line_info_table): Output view numbers in asm comments.
5365 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5366 Call maybe_gen_llsym.
5367 (loc_list_from_tree_1): Adjust.
5368 (add_AT_location_description): Create view list attribute if
5369 needed, check it's absent otherwise.
5370 (convert_cfa_to_fb_loc_list): Adjust.
5371 (maybe_emit_file): Call output_asm_line_debug_info for test.
5372 (dwarf2out_var_location): Reset views as needed. Precompute
5373 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5374 attribute. Set view.
5375 (new_line_info_table): Reset next view.
5376 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5377 (dwarf2out_source_line): Likewise. Output view resets and labels to
5378 the assembler, or select appropriate line info opcodes.
5379 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5380 (optimize_string_length): Catch it. Adjust.
5381 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5382 dw_val_class_view_list, and remove it if no longer needed.
5383 (hash_loc_list): Hash view numbers.
5384 (loc_list_hasher::equal): Compare them.
5385 (optimize_location_lists): Check whether a view list symbol is
5386 needed, and whether the locview attribute is present, and
5387 whether they match. Remove the locview attribute if no longer
5389 (index_location_lists): Call skip_loc_list_entry for test.
5390 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5391 Use output_dwarf_version.
5392 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5393 (struct dw_val_node): Add val_view_list.
5394 * final.c (SEEN_NEXT_VIEW): New.
5395 (set_next_view_needed): New.
5396 (clear_next_view_needed): New.
5397 (maybe_output_next_view): New.
5398 (final_start_function): Rename to...
5399 (final_start_function_1): ... this. Take pointer to FIRST,
5400 add SEEN parameter. Emit param bindings in the initial view.
5401 (final_start_function): Reintroduce SEEN-less interface.
5402 (final): Rename to...
5403 (final_1): ... this. Take SEEN parameter. Output final pending
5404 next view at the end.
5405 (final): Reintroduce seen-less interface.
5406 (final_scan_insn): Output pending next view before switching
5407 sections or ending a block. Mark the next view as needed when
5408 outputting variable locations. Notify debug backend of section
5409 changes, and of location view changes.
5410 (rest_of_handle_final): Adjust.
5411 * toplev.c (process_options): Autodetect value for debug variable
5412 location views option. Warn on incompat5 without -gdwarf-5.
5413 * doc/invoke.texi (gvariable-location-views): New.
5414 (gvariable-location-views=incompat5): New.
5415 (gno-variable-location-views): New.
5417 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5419 PR tree-optimization/84136
5420 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5421 that the result of find_edge is non-NULL.
5423 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5426 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5427 storing integer register in SImode. Fix cost of 256 and 512
5428 byte aligned SSE register store.
5430 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5432 * config/i386/i386.c (ix86_multiplication_cost): Fix
5433 multiplication cost for TARGET_AVX512DQ.
5435 2018-02-08 Marek Polacek <polacek@redhat.com>
5437 PR tree-optimization/84238
5438 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5441 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5443 PR tree-optimization/84265
5444 * tree-vect-stmts.c (vectorizable_store): Don't treat
5445 VMAT_CONTIGUOUS accesses as grouped.
5446 (vectorizable_load): Likewise.
5448 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5450 PR tree-optimization/81635
5451 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5452 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5453 (test_round_for_mask): New functions.
5454 (wide_int_cc_tests): Call test_round_for_mask.
5455 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5456 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5457 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5458 range returned by get_range_info.
5460 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5463 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5464 * symtab.c: Include builtins.h
5465 (symtab_node::output_to_lto_symbol_table_p): Move here
5466 from lto-streamer-out.c:output_symbol_p.
5467 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5468 (output_symbol_p): Move all logic to symtab.c
5469 (produce_symtab): Update.
5471 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5473 * config/s390/s390-opts.h (enum indirect_branch): Define.
5474 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5475 (s390_indirect_branch_via_thunk)
5476 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5477 (enum s390_indirect_branch_type): Define.
5478 * config/s390/s390.c (struct s390_frame_layout, struct
5479 machine_function): Remove.
5480 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5481 (indirect_branch_table_label_no, indirect_branch_table_name):
5483 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5484 (enum s390_indirect_branch_option): Define.
5485 (s390_return_addr_from_memory): New function.
5486 (s390_handle_string_attribute): New function.
5487 (s390_attribute_table): Add new attribute handler.
5488 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5489 (s390_indirect_branch_via_thunk): New function.
5490 (s390_indirect_branch_via_inline_thunk): New function.
5491 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5492 sibling call optimization for non z10 compiles.
5493 (s390_emit_call): Force indirect branch target to be a single
5494 register. Add r1 clobber for non-z10 compiles.
5495 (s390_emit_epilogue): Emit return jump via return_use expander.
5496 (s390_reorg): Handle JUMP_INSNs as execute targets.
5497 (s390_option_override_internal): Perform validity checks for the
5498 new command line options.
5499 (s390_indirect_branch_attrvalue): New function.
5500 (s390_indirect_branch_settings): New function.
5501 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5502 (s390_output_indirect_thunk_function): New function.
5503 (s390_code_end): Implement target hook.
5504 (s390_case_values_threshold): Implement target hook.
5505 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5507 * config/s390/s390.h (struct s390_frame_layout)
5508 (struct machine_function): Move here from s390.c.
5509 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5510 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5511 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5512 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5513 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5514 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5515 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5516 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5517 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5518 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5519 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5520 (mnemonic attribute): Add values which aren't recognized
5522 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5523 pattern for branch conversion. Fix mnemonic attribute.
5524 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5525 indirect branch via thunk if requested.
5526 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5527 ("*indirect_jump"): Disable for branch conversion using out of
5529 ("indirect_jump_via_thunk<mode>_z10")
5530 ("indirect_jump_via_thunk<mode>")
5531 ("indirect_jump_via_inlinethunk<mode>_z10")
5532 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5533 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5534 ("casesi_jump_via_inlinethunk<mode>_z10")
5535 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5536 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5537 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5538 ("*indirect2_jump"): Disable for branch conversion.
5539 ("casesi_jump"): Turn into expander and expand patterns for branch
5541 ("return_use"): New expander.
5542 ("*return"): Emit return via thunk and rename it to ...
5543 ("*return<mode>"): ... this one.
5544 * config/s390/s390.opt: Add new options and and enum for the
5547 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5549 * lra-constraints.c (match_reload): Unconditionally use
5550 gen_lowpart_SUBREG, rather than selecting between that
5551 and equivalent gen_rtx_SUBREG code.
5553 2018-02-08 Richard Biener <rguenther@suse.de>
5555 PR tree-optimization/84233
5556 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5557 changed flag instead of boguously re-using phi_inserted.
5559 2018-02-08 Martin Jambor <mjambor@suse.cz>
5561 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5562 static local variables.
5564 2018-02-08 Richard Biener <rguenther@suse.de>
5566 PR tree-optimization/84278
5567 * tree-vect-stmts.c (vectorizable_store): When looking for
5568 smaller vector types to perform grouped strided loads/stores
5569 make sure the mode is supported by the target.
5570 (vectorizable_load): Likewise.
5572 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5574 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5575 Increase LDP/STP opportunities by adding adjacent callee-saves.
5577 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5579 PR rtl-optimization/84068
5580 PR rtl-optimization/83459
5581 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5583 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5585 PR tree-optimization/84224
5586 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5587 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5590 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5593 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5594 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5595 count, remove one USE.
5597 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5599 * doc/install.texi (Configuration): Document the
5600 --with-long-double-format={ibm,ieee} PowerPC configuration
5604 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5605 Convert from define_expand to be define_insn_and_split. Rework
5606 float/double/_Float128 conversions to QI/HI/SImode to work with
5607 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5608 conversions to QI/HImode types did a store and then a load to
5609 truncate the value. For conversions to VSX registers, don't split
5610 the insn, instead emit the code directly. Use the code iterator
5611 any_fix to combine signed and unsigned conversions.
5612 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5613 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5614 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5615 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5616 (fix_<mode>di2_hw): Likewise.
5617 (fixuns_<mode>di2_hw): Likewise.
5618 (fix_<mode>si2_hw): Likewise.
5619 (fixuns_<mode>si2_hw): Likewise.
5620 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5621 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5622 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5623 fix<uns>_trunc<SFDF:mode>si2_p8.
5624 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5626 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5627 (fix<uns>_<mode>_mem): Likewise.
5628 (fctiw<u>z_<mode>_mem): Likewise.
5629 (fix<uns>_<mode>_mem): Likewise.
5630 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5631 the register allocator from doing a direct move to the GPRs to do
5632 a store, and instead use the ISA 3.0 store byte/half-word from
5633 vector register instruction. For IEEE 128-bit floating point,
5634 also optimize stores of 32-bit ints.
5635 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5637 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5639 * genextract.c (push_pathstr_operand): New function to support
5641 (walk_rtx): Call push_pathstr_operand.
5642 (print_path): Support [a-zA-Z].
5644 2018-02-07 Richard Biener <rguenther@suse.de>
5646 PR tree-optimization/84037
5647 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5648 (cse_and_gimplify_to_preheader): Declare.
5649 (vect_get_place_in_interleaving_chain): Likewise.
5650 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5652 (_loop_vec_info::~_loop_vec_info): Delete it.
5653 (cse_and_gimplify_to_preheader): New function.
5654 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5655 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5656 (vectorizable_load): Likewise. For grouped stores always base
5657 the IV on the first element.
5658 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5659 condition before gimplifying.
5661 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5663 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5664 *DIV_EXPR and *MOD_EXPR.
5666 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5669 * config/i386/i386.c (ix86_option_override_internal): Mask out
5670 the CF_SET bit when checking -fcf-protection.
5672 2018-02-07 Tom de Vries <tom@codesourcery.com>
5675 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5678 2018-02-07 Richard Biener <rguenther@suse.de>
5680 PR tree-optimization/84204
5681 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5684 PR tree-optimization/84205
5685 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5686 special-case isl_ast_op_zdiv_r.
5688 PR tree-optimization/84223
5689 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5690 Only add conditions from within the region.
5691 (gather_bbs::after_dom_children): Adjust.
5693 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5696 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5697 * config/avr/avr.md: Only post-reload split REG-REG moves if
5698 either register is GENERAL_REG_P.
5700 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5702 PR tree-optimization/84235
5703 * tree-ssa-scopedtables.c
5704 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5705 if the subtraction is performed in floating point type where NaNs are
5706 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5707 build 1. Formatting fix.
5709 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5712 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5713 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5714 and skip it regardless of bb boundaries. Use CALL_P macro,
5715 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5716 unnecessarily, formatting fix.
5718 2018-02-06 Michael Collison <michael.collison@arm.com>
5720 * config/arm/thumb2.md:
5721 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5722 (*thumb_mov_notscc): Ditto.
5724 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5727 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5728 unsigned_fix, not "s".
5730 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5732 * configure.ac (gcc_fn_eh_frame_ro): New function.
5733 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5734 correct .eh_frame permissions.
5735 * configure: Regenerate.
5737 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5739 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5742 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5744 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5745 Display warning message for -mno-speculate-indirect-jumps.
5747 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5749 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5751 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5753 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5755 PR tree-optimization/84225
5756 * tree-eh.c (find_trapping_overflow): Only call
5757 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5759 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5762 * config/i386/i386.c: Reimplement the check of possible options
5763 -mibt/-mshstk conbination. Change error messages.
5764 * doc/invoke.texi: Fix a typo: remove extra '='.
5766 2018-02-06 Marek Polacek <polacek@redhat.com>
5768 PR tree-optimization/84228
5769 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5771 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5774 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5775 emitted arch directives.
5776 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5777 __ARM_FEATURE_COPROC before changing architectures.
5779 2018-02-06 Richard Biener <rguenther@suse.de>
5781 * config/i386/i386.c (print_reg): Fix typo.
5782 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5784 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5786 * configure: Regenerate.
5788 2018-02-05 Martin Sebor <msebor@redhat.com>
5790 PR tree-optimization/83369
5791 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5794 2018-02-05 Martin Liska <mliska@suse.cz>
5796 * doc/invoke.texi: Cherry-pick upstream r323995.
5798 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5800 * ira.c (ira_init_register_move_cost): Adjust comment.
5802 2018-02-05 Martin Liska <mliska@suse.cz>
5804 PR gcov-profile/84137
5805 * doc/gcov.texi: Fix typo in documentation.
5807 2018-02-05 Martin Liska <mliska@suse.cz>
5809 PR gcov-profile/83879
5810 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5811 using dlopen functionality.
5813 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5815 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5816 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5817 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5818 _mm_maskz_range_round_ss): New intrinsics.
5819 (__builtin_ia32_rangesd128_round)
5820 (__builtin_ia32_rangess128_round): Remove.
5821 (__builtin_ia32_rangesd128_mask_round,
5822 __builtin_ia32_rangess128_mask_round): New builtins.
5823 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5824 __builtin_ia32_rangess128_round): Remove.
5825 (__builtin_ia32_rangesd128_mask_round,
5826 __builtin_ia32_rangess128_mask_round): New builtins.
5827 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5828 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5829 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5830 "<round_saeonly_constraint>")): Changed to ...
5831 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5832 "<round_saeonly_scalar_constraint>")): ... this.
5833 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5834 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5835 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5836 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5837 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5839 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5841 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5843 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5844 Remove all values except native, 8540 and 8548.
5846 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5848 * config/i386/i386.c (ix86_output_function_return): Pass
5849 INVALID_REGNUM, instead of -1, as invalid register number to
5850 indirect_thunk_name and output_indirect_thunk.
5852 2018-02-02 Julia Koval <julia.koval@intel.com>
5854 * config.gcc: Add -march=icelake.
5855 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5856 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5857 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5858 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5859 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5860 (processor_target_table): Add icelake.
5861 (ix86_option_override_internal): Handle new PTAs.
5862 (get_builtin_code_for_version): Handle icelake.
5863 (M_INTEL_COREI7_ICELAKE): New.
5864 (fold_builtin_cpu): Handle icelake.
5865 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5866 * doc/invoke.texi: Add -march=icelake.
5868 2018-02-02 Julia Koval <julia.koval@intel.com>
5870 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5871 to wide_int_bitmask.
5872 * wide-int-bitmask.h: New.
5874 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5877 * config/i386/i386.md: Replace Pmode with word_mode in
5878 builtin_setjmp_setup and builtin_longjmp to support x32.
5880 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5884 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5886 (rs6000_supported_cpu_names): New static variable.
5887 (linux_cpu_translation_table): Likewise.
5888 (elf_platform) <cpu>: Define new static variable and use it.
5889 Translate kernel AT_PLATFORM name to canonical name if needed.
5890 Error if platform name is unknown.
5892 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5895 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5897 2018-02-01 Jeff Law <law@redhat.com>
5900 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5901 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5902 the scratch if RELEASE_VIA_POP is false.
5903 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5904 If we have to save a temporary register, decrement SIZE appropriately.
5905 Pass new arguments to release_scratch_register_on_entry.
5906 (ix86_adjust_stack_and_probe): Likewise.
5907 (ix86_emit_probe_stack_range): Pass new arguments to
5908 release_scratch_register_on_entry.
5910 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5912 PR rtl-optimization/84157
5913 * combine.c (change_zero_ext): Use REG_P predicate in
5914 front of HARD_REGISTER_P predicate.
5916 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5918 * config/avr/avr.c (avr_option_override): Move disabling of
5919 -fdelete-null-pointer-checks to...
5920 * common/config/avr/avr-common.c (avr_option_optimization_table):
5923 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5925 PR tree-optimization/81635
5926 * tree-data-ref.c (split_constant_offset_1): For types that
5927 wrap on overflow, try to use range info to prove that wrapping
5930 2018-02-01 Renlin Li <renlin.li@arm.com>
5933 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5935 (aarch64_register_move_cost): Likewise.
5936 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5938 (REG_CLASS_NAMES): Likewise.
5939 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5940 TAILCALL_ADDR_REGS. Remove IP registers.
5941 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5943 2018-02-01 Richard Biener <rguenther@suse.de>
5945 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5946 for specifying RPO order and allow NULL for that.
5947 * domwalk.c (dom_walker::dom_walker): Likewise.
5948 (dom_walker::walk): Handle NULL RPO order.
5949 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5951 (rewrite_update_dom_walker): Likewise.
5952 (mark_def_dom_walker): Likewise.
5954 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5956 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5957 (aarch64_maybe_expand_sve_subreg_move): Declare.
5958 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5959 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5961 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5962 that are semantically a reverse operation.
5963 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5964 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5965 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5967 (aarch64_can_change_mode_class): For big-endian, forbid changes
5968 between two SVE modes if they have different element sizes.
5970 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5972 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5973 the TImode handling for big-endian targets.
5975 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5977 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5978 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5980 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5981 Remove BSWAP handing for big-endian targets and use the form of
5982 LD1RQ appropariate for the mode.
5984 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5986 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5987 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5990 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5993 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5994 check for operands that need to go through aarch64_sve_reload_be.
5996 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5998 PR tree-optimization/81661
5999 PR tree-optimization/84117
6000 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6001 * tree-eh.c: Include gimplify.h.
6002 (find_trapping_overflow, replace_trapping_overflow,
6003 rewrite_to_non_trapping_overflow): New functions.
6004 * tree-vect-loop.c: Include tree-eh.h.
6005 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6006 * tree-data-ref.c: Include tree-eh.h.
6007 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6009 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6011 PR rtl-optimization/84123
6012 * combine.c (change_zero_ext): Check if hard register satisfies
6013 can_change_dest_mode before calling gen_lowpart_SUBREG.
6015 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6018 * ira.c (ira_init_register_move_cost): Remove assert.
6020 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6022 PR rtl-optimization/84071
6023 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6024 * doc/tm.texi: Regenerate.
6026 2018-01-31 Richard Biener <rguenther@suse.de>
6028 PR tree-optimization/84132
6029 * tree-data-ref.c (analyze_miv_subscript): Properly
6030 check whether evolution_function_is_affine_multivariate_p
6031 before calling gcd_of_steps_may_divide_p.
6033 2018-01-31 Julia Koval <julia.koval@intel.com>
6036 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6037 * config/i386/i386.md (rdpid_rex64) New.
6038 (rdpid): Make 32bit only.
6040 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6043 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6044 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6046 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6049 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6051 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6053 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6055 PR rtl-optimization/84071
6056 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6057 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6059 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6061 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6062 (arc_attribute_table): Add 'aux' attribute.
6063 (arc_in_small_data_p): Consider aux like variables.
6064 (arc_is_aux_reg_p): New function.
6065 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6066 (arc_get_aux_arg): New function.
6067 (prepare_move_operands): Handle aux-register access.
6068 (arc_handle_aux_attribute): New function.
6069 * doc/extend.texi (ARC Variable attributes): Add subsection.
6071 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6073 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6074 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6075 (arc_attribute_table): Add 'uncached' attribute.
6076 (arc_print_operand): Print '.di' flag for uncached memory
6078 (arc_in_small_data_p): Do not consider for small data the uncached
6080 (arc_is_uncached_mem_p): New function.
6081 * config/arc/predicates.md (compact_store_memory_operand): Check
6082 for uncached memory accesses.
6083 (nonvol_nonimm_operand): Likewise.
6084 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6086 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6089 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6090 falign-loops=): Add Optimization flag.
6092 2018-01-30 Jeff Law <law@redhat.com>
6095 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6096 INT_REGISTERS_SAVED. Check it prior to calling
6097 get_scratch_register_on_entry.
6098 (ix86_adjust_stack_and_probe): Similarly.
6099 (ix86_emit_probe_stack_range): Similarly.
6100 (ix86_expand_prologue): Corresponding changes.
6102 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6105 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6106 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6108 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6111 * lra-constraints.c (curr_insn_transform): Process AND in the
6114 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6116 PR rtl-optimization/83986
6117 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6118 dependence against last_pending_memory_flush in addition to
6121 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6123 PR tree-optimization/81611
6124 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6127 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6130 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6133 2018-01-30 Richard Biener <rguenther@suse.de>
6134 Jakub Jelinek <jakub@redhat.com>
6136 PR tree-optimization/84111
6137 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6138 inner loops added during recursion, as they don't have up-to-date
6141 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6144 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6145 (can_inline_edge_by_limits_p): ... here.
6146 (can_early_inline_edge_p, check_callers,
6147 update_caller_keys, update_callee_keys, recursive_inlining,
6148 add_new_edges_to_heap, speculation_useful_p,
6149 inline_small_functions,
6150 inline_small_functions, flatten_function,
6151 inline_to_all_callers_1): Update.
6153 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6155 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6158 2018-01-30 Richard Biener <rguenther@suse.de>
6160 PR tree-optimization/83008
6161 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6162 invariant and constant vector uses in stmts when they need
6165 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6168 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6169 * configure: Regenerate.
6171 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6173 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6175 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6176 Use gen_rtx_REG rather than gen_lowpart.
6178 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6180 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6181 rather than 0 when creating partial subregs.
6183 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6185 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6188 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6191 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6192 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6193 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6194 flags. This restores the settings used before the 2017-07-24.
6195 Turning off pre increment/decrement/modify allows IVOPTS to
6196 optimize DF/SF loops where the index is an int.
6198 2018-01-29 Richard Biener <rguenther@suse.de>
6199 Kelvin Nilsen <kelvin@gcc.gnu.org>
6202 * tree-vect-stmts.c (vectorizable_call): Don't call
6203 targetm.vectorize_builtin_md_vectorized_function if callee is
6206 2018-01-22 Carl Love <cel@us.ibm.com>
6208 * doc/extend.tex: Fix typo in second arg in
6209 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6211 2018-01-29 Richard Biener <rguenther@suse.de>
6213 PR tree-optimization/84086
6214 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6215 (flush_ssaname_freelist): When SSA names were released reset
6216 the SCEV hash table.
6218 2018-01-29 Richard Biener <rguenther@suse.de>
6220 PR tree-optimization/84057
6221 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6222 removed paths when removing edges.
6224 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6226 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6227 -mfunction-return=@var{choice}.
6229 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6232 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6234 (layout::print_source_line): Likewise.
6235 (test_get_line_width_without_trailing_whitespace): Add test cases.
6237 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6240 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6243 2018-01-26 Jim Wilson <jimw@sifive.com>
6245 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6247 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6250 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6252 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6253 and CMP + SUB-immediate -> SUBS.
6255 2018-01-26 Martin Sebor <msebor@redhat.com>
6257 PR tree-optimization/83896
6258 * tree-ssa-strlen.c (get_string_len): Rename...
6259 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6260 Avoid assuming length is constant.
6261 (handle_char_store): Use HOST_WIDE_INT for string length.
6263 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6266 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6267 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6269 2018-01-26 Richard Biener <rguenther@suse.de>
6271 PR rtl-optimization/84003
6272 * dse.c (record_store): Only record redundant stores when
6273 the earlier store aliases at least all accesses the later one does.
6275 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6277 PR rtl-optimization/83985
6278 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6279 REG_CFA_RESTORE insns.
6280 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6281 REG_CFA_RESTORE insns here.
6284 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6285 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6287 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6289 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6290 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6291 (arc_init): Likewise.
6292 (arc_override_options): Likewise.
6293 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6295 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6297 * config/arc/arc.h (TARGET_DBNZ): Define.
6298 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6299 properly set the tune attribute.
6300 (dbnz): Use TARGET_DBNZ guard.
6301 * config/arc/arc.opt (mtune): Add core3 option.
6303 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6305 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6306 recognize new pic like addresses.
6307 (arc_delegitimize_address): Clean up.
6309 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6311 * config/arc/arc-arches.def: Option mrf16 valid for all
6313 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6314 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6315 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6316 * config/arc/arc-tables.opt: Regenerate.
6317 * config/arc/arc.c (arc_conditional_register_usage): Handle
6318 reduced register file case.
6319 (arc_file_start): Set must have build attributes.
6320 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6322 * config/arc/arc.opt (mrf16): Add new option.
6323 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6324 * config/arc/genmultilib.awk: Handle new mrf16 option.
6325 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6326 * config/arc/t-multilib: Regenerate.
6327 * doc/invoke.texi (ARC Options): Document mrf16 option.
6329 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6331 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6332 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6333 (arc_attribute_table): Add 'secure_call' attribute.
6334 (arc_print_operand): Print secure call operand.
6335 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6337 (arc_is_secure_call_p): New function. * config/arc/arc.md
6338 (call_i): Add support for sjli instruction.
6339 (call_value_i): Likewise.
6340 * config/arc/constraints.md (Csc): New constraint.
6342 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6343 John Eric Martin <John.Martin@emmicro-us.com>
6345 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6346 * config/arc/arc.c (_arc_jli_section): New struct.
6347 (arc_jli_section): New type.
6348 (rc_jli_sections): New static variable.
6349 (arc_handle_jli_attribute): New function.
6350 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6351 (arc_file_end): New function.
6352 (TARGET_ASM_FILE_END): Define.
6353 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6354 (arc_add_jli_section): New function.
6355 (jli_call_scan): Likewise.
6356 (arc_reorg): Call jli_call_scan.
6357 (arc_output_addsi): Remove 'S' from printing asm operand.
6358 (arc_is_jli_call_p): New function.
6359 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6361 (movhi_insn): Likewise.
6362 (movsi_insn): Likewise.
6363 (movsi_set_cc_insn): Likewise.
6364 (loadqi_update): Likewise.
6365 (load_zeroextendqisi_update): Likewise.
6366 (load_signextendqisi_update): Likewise.
6367 (loadhi_update): Likewise.
6368 (load_zeroextendhisi_update): Likewise.
6369 (load_signextendhisi_update): Likewise.
6370 (loadsi_update): Likewise.
6371 (loadsf_update): Likewise.
6372 (movsicc_insn): Likewise.
6373 (bset_insn): Likewise.
6374 (bxor_insn): Likewise.
6375 (bclr_insn): Likewise.
6376 (bmsk_insn): Likewise.
6377 (bicsi3_insn): Likewise.
6378 (cmpsi_cc_c_insn): Likewise.
6379 (movsi_ne): Likewise.
6380 (movsi_cond_exec): Likewise.
6381 (clrsbsi2): Likewise.
6392 (call_i): Remove 'S' asm letter, add jli instruction.
6393 (call_value_i): Likewise.
6394 * config/arc/arc.op (mjli-always): New option.
6395 * config/arc/constraints.md (Cji): New constraint.
6396 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6398 (subsf3_fpx): Likewise.
6399 (mulsf3_fpx): Likewise.
6400 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6402 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6404 * doc/invoke.texi (ARC): Document mjli-always option.
6406 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6408 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6409 avoid addition with 0 and use incw and decw where possible.
6411 2018-01-26 Richard Biener <rguenther@suse.de>
6413 PR tree-optimization/81082
6414 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6415 association if it requires casting to unsigned.
6416 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6417 from fold_plusminus_mult_expr to catch important cases late when
6418 range info is available.
6420 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6422 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6423 * configure.ac (hidden_linkonce): New test.
6424 * configure: Regenerate.
6425 * config.in: Regenerate.
6427 2018-01-26 Julia Koval <julia.koval@intel.com>
6429 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6430 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6431 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6432 _mm_mask_bitshuffle_epi64_mask): Fix type.
6433 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6434 USI_FTYPE_V4DI_V4DI_USI): Remove.
6435 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6436 __builtin_ia32_vpshufbitqmb256_mask,
6437 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6438 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6439 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6441 2018-01-26 Alan Modra <amodra@gmail.com>
6444 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6445 UNSPEC_VBPERMQ. Sort other unspecs.
6447 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6449 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6451 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6454 * predict.c (drop_profile): Do not push/pop cfun; update also
6456 (handle_missing_profiles): Fix logic looking for zero profiles.
6458 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6461 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6462 on functions with #pragma omp declare simd or functions with simd
6464 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6465 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6466 Remove trailing \n from warning_at calls.
6468 2018-01-25 Tom de Vries <tom@codesourcery.com>
6471 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6472 for neutered workers.
6474 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6477 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6478 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6480 2018-01-24 Jeff Law <law@redhat.com>
6483 * i386.c (get_probe_interval): Move to earlier point.
6484 (ix86_compute_frame_layout): If -fstack-clash-protection and
6485 the frame is larger than the probe interval, then use pushes
6486 to save registers rather than reg->mem moves.
6487 (ix86_expand_prologue): Remove conditional for int_registers_saved
6490 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6493 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6494 min/max for never referenced object.
6496 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6499 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6501 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6502 attributes from DECL_ATTRIBUTES (decl) without affecting
6503 DECL_ATTRIBUTES (current_function_decl).
6504 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6505 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6507 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6509 PR tree-optimization/83979
6510 * fold-const.c (fold_comparison): Use constant_boolean_node
6511 instead of boolean_{true,false}_node.
6513 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6515 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6518 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6520 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6521 Simplify the clause that sets the length attribute.
6522 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6523 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6524 clause that sets the length attribute.
6525 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6527 2018-01-24 Tom de Vries <tom@codesourcery.com>
6530 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6531 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6532 Add strict parameter.
6533 (prevent_branch_around_nothing): Insert dummy insn between branch to
6534 label and label with no ptx insn inbetween.
6535 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6537 2018-01-24 Tom de Vries <tom@codesourcery.com>
6540 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6541 for neutered threads in warp.
6542 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6544 2018-01-24 Richard Biener <rguenther@suse.de>
6546 PR tree-optimization/83176
6547 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6550 2018-01-24 Richard Biener <rguenther@suse.de>
6552 PR tree-optimization/82819
6553 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6554 code generating pluses that are no-ops in the target precision.
6556 2018-01-24 Richard Biener <rguenther@suse.de>
6559 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6561 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6563 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6564 to merge probabilities.
6565 * predict.c (probably_never_executed): Also mark as cold functions
6566 with global 0 profile and guessed local profile.
6567 * profile-count.c (profile_probability::combine_with_count): New
6569 * profile-count.h (profile_probability::operator*,
6570 profile_probability::operator*=, profile_probability::operator/,
6571 profile_probability::operator/=): Reduce precision to adjusted
6572 and set value to guessed on contradictory divisions.
6573 (profile_probability::combine_with_freq): Remove.
6574 (profile_probability::combine_wiht_count): Declare.
6575 (profile_count::force_nonzero):: Set to adjusted.
6576 (profile_count::probability_in):: Set quality to adjusted.
6577 * tree-ssa-tail-merge.c (replace_block_by): Use
6580 2018-01-23 Andrew Waterman <andrew@sifive.com>
6581 Jim Wilson <jimw@sifive.com>
6583 * config/riscv/riscv.c (riscv_stack_boundary): New.
6584 (riscv_option_override): Set riscv_stack_boundary. Handle
6585 riscv_preferred_stack_boundary_arg.
6586 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6587 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6588 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6589 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6590 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6591 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6593 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6596 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6597 of struct ix86_frame.
6598 (ix86_expand_epilogue): Likewise. Add a local variable for
6599 the reg_save_offset field in struct ix86_frame.
6601 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6603 PR tree-optimization/82604
6604 * tree-loop-distribution.c (enum partition_kind): New enum item
6605 PKIND_PARTIAL_MEMSET.
6606 (partition_builtin_p): Support above new enum item.
6607 (generate_code_for_partition): Ditto.
6608 (compute_access_range): Differentiate cases that equality can be
6609 proven at all loops, the innermost loops or no loops.
6610 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6611 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6612 (finalize_partitions, distribute_loop): Don't fuse partition of
6613 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6614 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6617 2018-01-23 Martin Liska <mliska@suse.cz>
6619 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6620 order to ignore the predictor.
6621 (PRED_POLYMORPHIC_CALL): Likewise.
6622 (PRED_RECURSIVE_CALL): Likewise.
6624 2018-01-23 Martin Liska <mliska@suse.cz>
6626 * tree-profile.c (tree_profiling): Print function header to
6627 aware reader which function we are working on.
6628 * value-prof.c (gimple_find_values_to_profile): Do not print
6629 not interesting value histograms.
6631 2018-01-23 Martin Liska <mliska@suse.cz>
6633 * profile-count.h (enum profile_quality): Add
6634 profile_uninitialized as the first value. Do not number values
6635 as they are zero based.
6636 (profile_count::verify): Update sanity check.
6637 (profile_probability::verify): Likewise.
6639 2018-01-23 Nathan Sidwell <nathan@acm.org>
6641 * doc/invoke.texi (ffor-scope): Deprecate.
6643 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6645 PR tree-optimization/83510
6646 * domwalk.c (set_all_edges_as_executable): New function.
6647 (dom_walker::dom_walker): Convert bool param
6648 "skip_unreachable_blocks" to enum reachability. Move setup of
6649 edge flags to set_all_edges_as_executable and only do it when
6650 reachability is REACHABLE_BLOCKS.
6651 * domwalk.h (enum dom_walker::reachability): New enum.
6652 (dom_walker::dom_walker): Convert bool param
6653 "skip_unreachable_blocks" to enum reachability.
6654 (set_all_edges_as_executable): New decl.
6655 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6656 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6658 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6659 but converting true to REACHABLE_BLOCKS.
6660 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6662 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6663 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6664 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6666 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6667 if check_all_array_refs will be called.
6669 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6671 * tree.c (selftest::test_location_wrappers): Add more test
6674 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6676 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6677 (selftest::test_bit_in_range): Likewise.
6679 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6682 * doc/sourcebuild.texi (vect_float): Say that the selector
6683 only describes the situation when -funsafe-math-optimizations is on.
6684 (vect_float_strict): Document.
6686 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6688 PR tree-optimization/83965
6689 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6690 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6691 instead of checking only for a reduction.
6692 (vect_recog_widen_sum_pattern): Likewise.
6694 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6696 * predict.c (probably_never_executed): Only use precise profile info.
6697 (compute_function_frequency): Skip after inlining hack since we now
6698 have quality checking.
6700 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6702 * profile-count.h (profile_probability::very_unlikely,
6703 profile_probability::unlikely, profile_probability::even): Set
6704 precision to guessed.
6706 2018-01-23 Richard Biener <rguenther@suse.de>
6708 PR tree-optimization/83963
6709 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6710 Properly terminate dominator walk when crossing the exit edge not
6711 when visiting its source block.
6713 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6716 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6717 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6719 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6721 PR tree-optimization/83957
6722 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6723 semicolon after for body surrounded by braces.
6725 PR tree-optimization/83081
6726 * profile-count.h (profile_probability::split): New method.
6727 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6728 Use profile_probability::split.
6729 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6730 when splitting a single conditional jump into 2.
6732 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6734 PR tree-optimization/69452
6735 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6738 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6740 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6741 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6742 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6744 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6746 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6747 * config/rl78/rl78.md (movdi): New define_expand.
6748 * config/rl78/rl78.c (rl78_split_movdi): New function.
6750 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6753 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6755 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6756 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6757 128-bit to produce an UNSPEC move to get the double word with the
6758 signbit and then a shift directly to do signbit.
6759 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6760 implementation with a new version that just does either a direct
6761 move or a regular move. Move memory interface to separate insns.
6762 Move insns so they are next to the expander.
6763 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6764 with signbit move. Split big and little endian case.
6765 (signbit<mode>2_dm_mem_le): Likewise.
6766 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6767 (signbit<mode>2_dm2): Likewise.
6769 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6771 * config/rl78/rl78.md (anddi3): New define_expand.
6773 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6775 * config/rl78/rl78.md (umindi3): New define_expand.
6777 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6779 * config/rl78/rl78.md (smindi3): New define_expand.
6781 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6783 * config/rl78/rl78.md (smaxdi3): New define_expand.
6785 2018-01-22 Carl Love <cel@us.ibm.com>
6787 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6788 LVX_V1TI): Add macro expansion.
6789 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6790 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6791 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6792 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6793 Change check to determine if the instruction is a byte reversing
6794 entry. Fix typo in comment.
6795 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6796 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6797 Add def_builtin calls for new builtins.
6798 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6799 Add define_insn expansion.
6801 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6803 * config/rl78/rl78.md (umaxdi3): New define_expand.
6805 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6807 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6808 for non-QImode registers.
6810 2018-01-22 Richard Biener <rguenther@suse.de>
6812 PR tree-optimization/83963
6813 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6814 including the loop exit block.
6815 (scop_detection::merge_sese): Likewise.
6816 (scop_detection::add_scop): Do it here instead.
6818 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6820 * doc/sourcebuild.texi (arm_softfloat): Document.
6822 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6825 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6826 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6827 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6829 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6830 David Edelsohn <dje.gcc@gmail.com>
6833 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6834 Change "crset eq" to "crset 2".
6835 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6836 (*call_indirect_aix<mode>_nospec): Likewise.
6837 (*call_value_indirect_aix<mode>_nospec): Likewise.
6838 (*call_indirect_elfv2<mode>_nospec): Likewise.
6839 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6840 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6841 change assembly output from . to $.
6842 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6843 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6844 (*tablejump<mode>_internal1_nospec): Likewise.
6846 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6849 * config/sh/sh_optimize_sett_clrt.cc:
6850 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6852 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6854 PR tree-optimization/83940
6855 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6856 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6857 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6858 use it to pass back the definition type.
6859 (vect_check_store_rhs): Likewise rhs_dt_out.
6860 (vect_build_gather_load_calls): Add a mask_dt argument and use
6861 it instead of a call to vect_is_simple_use.
6862 (vectorizable_store): Update calls to vect_check_load_store_mask
6863 and vect_check_store_rhs. Use the dt returned by the latter instead
6864 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6865 instead of calls to vect_is_simple_use. Pass the scalar rather
6866 than the vector operand to vect_is_simple_use when handling
6867 second and subsequent copies of an rhs value.
6868 (vectorizable_load): Update calls to vect_check_load_store_mask
6869 and vect_build_gather_load_calls. Use the cached mask_dt and
6870 gs_info.offset_dt instead of calls to vect_is_simple_use.
6872 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6875 * tree-emutls.c: Include gimplify.h.
6876 (lower_emutls_2): New function.
6877 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6878 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6879 it before further processing.
6882 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6883 UINTVAL (trueop1) instead of INTVAL (op1).
6885 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6889 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6890 INCOMING_FRAME_SP_OFFSET if not defined.
6891 (scan_trace): Add ENTRY argument. If true and
6892 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6893 emit a note to adjust the CFA offset.
6894 (create_cfi_notes): Adjust scan_trace callers.
6895 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6896 INCOMING_FRAME_SP_OFFSET in the CIE.
6897 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6898 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6900 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6901 * doc/tm.texi: Regenerated.
6903 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6905 PR rtl-optimization/83147
6906 * lra-constraints.c (remove_inheritance_pseudos): Use
6907 lra_substitute_pseudo_within_insn.
6909 2018-01-19 Tom de Vries <tom@codesourcery.com>
6910 Cesar Philippidis <cesar@codesourcery.com>
6913 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6915 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6918 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6919 spaces for function labels.
6921 2018-01-19 Martin Liska <mliska@suse.cz>
6923 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6924 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6925 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6926 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6927 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6928 (PRED_CONST_RETURN): Change from 69 to 65.
6929 (PRED_NULL_RETURN): Change from 91 to 71.
6930 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6931 (PRED_LOOP_GUARD): Change from 66 to 73.
6933 2018-01-19 Martin Liska <mliska@suse.cz>
6935 * predict.c (predict_insn_def): Add new assert.
6936 (struct branch_predictor): Change type to signed integer.
6937 (test_prediction_value_range): Amend test to cover
6939 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6940 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6941 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6942 (PRED_LOOP_IV_COMPARE): Likewise.
6943 * predict.h (PROB_UNINITIALIZED): Define new constant.
6945 2018-01-19 Martin Liska <mliska@suse.cz>
6947 * predict.c (dump_prediction): Add new format for
6948 analyze_brprob.py script which is enabled with -details
6950 * profile-count.h (precise_p): New function.
6952 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6954 PR tree-optimization/83922
6955 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6956 there are no statements that need masking.
6957 (vect_active_double_reduction_p): New function.
6958 (vect_analyze_loop_operations): Use it when handling phis that
6959 are not in the loop header.
6961 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6963 PR tree-optimization/83914
6964 * tree-vect-loop.c (vectorizable_induction): Don't convert
6965 init_expr or apply the peeling adjustment for inductions
6966 that are nested within the vectorized loop.
6968 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6970 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6973 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6977 * function.h (gimplify_parameters): Add gimple_seq * argument.
6978 * function.c: Include gimple.h and options.h.
6979 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6980 for the added local temporaries if needed.
6981 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6982 if there are any parameter cleanups, wrap whole body into a
6983 try/finally with the cleanups.
6985 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6988 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6989 Use GET_MODE_CLASS for scalar floating point.
6991 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6995 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6996 Fix call of call_cgraph_insertion_hooks.
6998 2018-01-18 Martin Sebor <msebor@redhat.com>
7000 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7002 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7005 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7008 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7011 * common.opt: (-ffile-prefix-map): New option.
7012 * opts.c (common_handle_option): Defer it.
7013 * opts-global.c (handle_common_deferred_options): Handle it.
7014 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7015 * file-prefix-map.h: New file.
7016 (remap_debug_filename, add_debug_prefix_map): ...here.
7017 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7018 * final.c (debug_prefix_map, add_debug_prefix_map
7019 remap_debug_filename): Move to...
7020 * file-prefix-map.c: New file.
7021 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7022 generalize, get rid of alloca(), use strrchr() instead of strchr().
7023 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7024 Implement in terms of add_prefix_map().
7025 (remap_macro_filename, remap_debug_filename): Implement in term of
7027 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7028 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7029 * dbxout.c: Include file-prefix-map.h.
7030 * varasm.c: Likewise.
7031 * vmsdbgout.c: Likewise.
7032 * xcoffout.c: Likewise.
7033 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7034 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7035 * doc/invoke.texi (-ffile-prefix-map): Document.
7036 (-fdebug-prefix-map): Update description.
7038 2018-01-18 Martin Liska <mliska@suse.cz>
7040 * config/i386/i386.c (indirect_thunk_name): Document that also
7042 (output_indirect_thunk): Document why both instructions
7043 (pause and lfence) are generated.
7045 2018-01-18 Richard Biener <rguenther@suse.de>
7047 PR tree-optimization/83887
7048 * graphite-scop-detection.c
7049 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7050 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7051 (scop_detection::merge_sese): Re-implement with a flood-fill
7052 algorithm that properly finds a SESE region if it exists.
7054 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7057 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7058 pointer_diff optimizations use view_convert instead of convert.
7060 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7062 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7063 Generate different code for -mno-speculate-indirect-jumps.
7064 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7065 (*call_indirect_aix<mode>): Disable for
7066 -mno-speculate-indirect-jumps.
7067 (*call_indirect_aix<mode>_nospec): New define_insn.
7068 (*call_value_indirect_aix<mode>): Disable for
7069 -mno-speculate-indirect-jumps.
7070 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7071 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7072 -mno-speculate-indirect-jumps.
7073 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7075 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7077 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7078 long double type, set the flags for noting the default long double
7079 type, even if we don't pass or return a long double type.
7081 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7084 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7087 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7089 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7090 support for merge[hl].
7091 (fold_mergehl_helper): New helper function.
7092 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7093 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7094 (altivec_vmrglw_direct): Add xxmrglw insn.
7096 2018-01-17 Andrew Waterman <andrew@sifive.com>
7098 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7099 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7101 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7104 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7105 call the lto_location_cache before reading the
7106 DECL_SOURCE_LOCATION of the types.
7108 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7109 Richard Sandiford <richard.sandiford@linaro.org>
7111 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7112 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7113 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7114 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7115 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7117 * config/aarch64/constraints.md (aarch64_movti_operand):
7119 * config/aarch64/predicates.md (Uti): Add new constraint.
7121 2018-01-17 Carl Love <cel@us.ibm.com>
7123 * config/rs6000/vsx.md (define_expand xl_len_r,
7124 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7125 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7127 (define_expand, define_insn): Move the shift left from the
7128 define_insn to the define_expand for lxvl and stxvl instructions.
7129 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7130 and XL_LEN_R definitions to PURE.
7132 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7134 * config/i386/i386.c (indirect_thunk_name): Declare regno
7135 as unsigned int. Compare regno with INVALID_REGNUM.
7136 (output_indirect_thunk): Ditto.
7137 (output_indirect_thunk_function): Ditto.
7138 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7139 in the call to output_indirect_thunk_function.
7141 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7144 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7145 rather than the size of inner_type to determine the stack slot size
7146 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7148 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7151 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7154 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7156 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7157 endian Linux systems to optionally enable multilibs for selecting
7158 the long double type if the user configured an explicit type.
7159 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7160 have no long double multilibs if not defined.
7161 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7162 warn if the user used -mabi={ieee,ibm}longdouble and we built
7163 multilibs for long double.
7164 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7165 appropriate multilib option.
7166 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7168 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7169 for building long double multilibs.
7170 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7172 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7174 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7177 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7179 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7182 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7185 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7188 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7190 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7191 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7193 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7195 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7196 different rtl trees depending on TARGET_64BIT.
7197 (rs6000_gen_lvx): Likewise.
7199 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7201 * config/visium/visium.md (nop): Tweak comment.
7202 (hazard_nop): Likewise.
7204 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7206 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7207 -mspeculate-indirect-jumps.
7208 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7209 for -mno-speculate-indirect-jumps.
7210 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7211 (*call_value_indirect_elfv2<mode>): Disable for
7212 -mno-speculate-indirect-jumps.
7213 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7214 (indirect_jump): Emit different RTL for
7215 -mno-speculate-indirect-jumps.
7216 (*indirect_jump<mode>): Disable for
7217 -mno-speculate-indirect-jumps.
7218 (*indirect_jump<mode>_nospec): New define_insn.
7219 (tablejump): Emit different RTL for
7220 -mno-speculate-indirect-jumps.
7221 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7222 (tablejumpsi_nospec): New define_expand.
7223 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7224 (tablejumpdi_nospec): New define_expand.
7225 (*tablejump<mode>_internal1): Disable for
7226 -mno-speculate-indirect-jumps.
7227 (*tablejump<mode>_internal1_nospec): New define_insn.
7228 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7231 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7233 * caller-save.c (insert_save): Drop unnecessary parameter. All
7236 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7237 Richard Biener <rguenth@suse.de>
7240 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7241 return early, inline manually is_gimple_sizepos. Make sure if we
7242 call gimplify_expr we don't end up with a gimple constant.
7243 * tree.c (variably_modified_type_p): Don't return true for
7244 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7245 * gimplify.h (is_gimple_sizepos): Remove.
7247 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7249 PR tree-optimization/83857
7250 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7251 vectorizable_live_operation for pure SLP statements.
7252 (vectorizable_live_operation): Handle PHIs.
7254 2018-01-16 Richard Biener <rguenther@suse.de>
7256 PR tree-optimization/83867
7257 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7258 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7260 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7263 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7264 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7265 If off is not INTEGER_CST, issue a may not be aligned warning
7266 rather than isn't aligned. Use isn%'t rather than isn't.
7267 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7269 <case MULT_EXPR>: Improve the case when bottom and one of the
7270 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7271 operand, in that case check if the other operand is multiple of
7272 bottom divided by the INTEGER_CST operand.
7274 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7277 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7278 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7279 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7280 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7281 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7282 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7283 (pa_function_arg_size): New function.
7285 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7287 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7288 in a separate statement.
7290 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7292 PR tree-optimization/83847
7293 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7294 group gathers and scatters.
7296 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7298 PR rtl-optimization/86620
7299 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7301 PR rtl-optimization/83213
7302 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7303 to last if both are JUMP_INSNs.
7305 PR tree-optimization/83843
7306 * gimple-ssa-store-merging.c
7307 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7308 store_immediate_info for bswap/nop orig_stores.
7310 2018-01-15 Andrew Waterman <andrew@sifive.com>
7312 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7314 <UDIV>: Increase cost if !TARGET_DIV.
7316 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7318 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7319 (define_attr "cr_logical_3op"): New.
7320 (cceq_ior_compare): Adjust.
7321 (cceq_ior_compare_complement): Adjust.
7322 (*cceq_rev_compare): Adjust.
7323 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7324 (is_cracked_insn): Adjust.
7325 (insn_must_be_first_in_group): Adjust.
7326 * config/rs6000/40x.md: Adjust.
7327 * config/rs6000/440.md: Adjust.
7328 * config/rs6000/476.md: Adjust.
7329 * config/rs6000/601.md: Adjust.
7330 * config/rs6000/603.md: Adjust.
7331 * config/rs6000/6xx.md: Adjust.
7332 * config/rs6000/7450.md: Adjust.
7333 * config/rs6000/7xx.md: Adjust.
7334 * config/rs6000/8540.md: Adjust.
7335 * config/rs6000/cell.md: Adjust.
7336 * config/rs6000/e300c2c3.md: Adjust.
7337 * config/rs6000/e500mc.md: Adjust.
7338 * config/rs6000/e500mc64.md: Adjust.
7339 * config/rs6000/e5500.md: Adjust.
7340 * config/rs6000/e6500.md: Adjust.
7341 * config/rs6000/mpc.md: Adjust.
7342 * config/rs6000/power4.md: Adjust.
7343 * config/rs6000/power5.md: Adjust.
7344 * config/rs6000/power6.md: Adjust.
7345 * config/rs6000/power7.md: Adjust.
7346 * config/rs6000/power8.md: Adjust.
7347 * config/rs6000/power9.md: Adjust.
7348 * config/rs6000/rs64.md: Adjust.
7349 * config/rs6000/titan.md: Adjust.
7351 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7353 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7354 ix86_indirect_branch_register logic.
7356 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7358 * config/i386/constraints.md (Bs): Update
7359 ix86_indirect_branch_register check. Don't check
7360 ix86_indirect_branch_register with GOT_memory_operand.
7362 * config/i386/predicates.md (GOT_memory_operand): Don't check
7363 ix86_indirect_branch_register here.
7364 (GOT32_symbol_operand): Likewise.
7366 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7368 * config/i386/predicates.md (constant_call_address_operand):
7369 Rewrite ix86_indirect_branch_register logic.
7370 (sibcall_insn_operand): Likewise.
7372 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7374 * config/i386/constraints.md (Bs): Replace
7375 ix86_indirect_branch_thunk_register with
7376 ix86_indirect_branch_register.
7378 * config/i386/i386.md (indirect_jump): Likewise.
7379 (tablejump): Likewise.
7380 (*sibcall_memory): Likewise.
7381 (*sibcall_value_memory): Likewise.
7382 Peepholes of indirect call and jump via memory: Likewise.
7383 * config/i386/i386.opt: Likewise.
7384 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7385 (GOT_memory_operand): Likewise.
7386 (call_insn_operand): Likewise.
7387 (sibcall_insn_operand): Likewise.
7388 (GOT32_symbol_operand): Likewise.
7390 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7393 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7394 type rather than type addr's type points to.
7395 (expand_omp_atomic_mutex): Likewise.
7396 (expand_omp_atomic): Likewise.
7398 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7401 * config/i386/i386.c (output_indirect_thunk_function): Use
7402 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7403 for __x86_return_thunk.
7405 2018-01-15 Richard Biener <rguenther@suse.de>
7408 * expmed.c (extract_bit_field_1): Fix typo.
7410 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7413 * config/arm/iterators.md (VF): New mode iterator.
7414 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7415 Remove integer-related logic from pattern.
7416 (neon_vabd<mode>_3): Likewise.
7418 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7421 * common.opt (fstrict-overflow): No longer an alias.
7422 (fwrapv-pointer): New option.
7423 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7424 also for pointer types based on flag_wrapv_pointer.
7425 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7426 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7427 opts->x_flag_wrapv got set.
7428 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7429 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7430 POINTER_TYPE_OVERFLOW_UNDEFINED.
7431 * match.pd: Likewise in address comparison pattern.
7432 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7434 2018-01-15 Richard Biener <rguenther@suse.de>
7437 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7438 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7439 Reset type names to their identifier if their TYPE_DECL doesn't
7440 have linkage (and thus is used for ODR and devirt).
7441 (save_debug_info_for_decl): Remove.
7442 (save_debug_info_for_type): Likewise.
7443 (add_tree_to_fld_list): Adjust.
7444 * tree-pretty-print.c (dump_generic_node): Make dumping of
7445 type names more robust.
7447 2018-01-15 Richard Biener <rguenther@suse.de>
7449 * BASE-VER: Bump to 8.0.1.
7451 2018-01-14 Martin Sebor <msebor@redhat.com>
7454 * builtins.c (check_access): Avoid warning when the no-warning bit
7457 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7459 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7460 * ira-color (allocno_hard_regs_compare): Likewise.
7462 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7465 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7466 Use .pushsection/.popsection.
7468 2018-01-14 Martin Sebor <msebor@redhat.com>
7471 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7473 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7475 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7476 entry from extra_headers.
7477 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7478 extra_headers, make the list bitwise identical to the i?86-*-* one.
7480 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7482 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7483 -mcmodel=large with -mindirect-branch=thunk,
7484 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7485 -mfunction-return=thunk-extern.
7486 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7487 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7488 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7490 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7492 * config/i386/i386.c (print_reg): Print the name of the full
7493 integer register without '%'.
7494 (ix86_print_operand): Handle 'V'.
7495 * doc/extend.texi: Document 'V' modifier.
7497 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7499 * config/i386/constraints.md (Bs): Disallow memory operand for
7500 -mindirect-branch-register.
7502 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7503 (GOT_memory_operand): Likewise.
7504 (call_insn_operand): Likewise.
7505 (sibcall_insn_operand): Likewise.
7506 (GOT32_symbol_operand): Likewise.
7507 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7508 for -mindirect-branch-register.
7509 (tablejump): Likewise.
7510 (*sibcall_memory): Likewise.
7511 (*sibcall_value_memory): Likewise.
7512 Disallow peepholes of indirect call and jump via memory for
7513 -mindirect-branch-register.
7514 (*call_pop): Replace m with Bw.
7515 (*call_value_pop): Likewise.
7516 (*sibcall_pop_memory): Replace m with Bs.
7517 * config/i386/i386.opt (mindirect-branch-register): New option.
7518 * doc/invoke.texi: Document -mindirect-branch-register option.
7520 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7522 * config/i386/i386-protos.h (ix86_output_function_return): New.
7523 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7524 set function_return_type.
7525 (indirect_thunk_name): Add ret_p to indicate thunk for function
7527 (output_indirect_thunk_function): Pass false to
7528 indirect_thunk_name.
7529 (ix86_output_indirect_branch_via_reg): Likewise.
7530 (ix86_output_indirect_branch_via_push): Likewise.
7531 (output_indirect_thunk_function): Create alias for function
7532 return thunk if regno < 0.
7533 (ix86_output_function_return): New function.
7534 (ix86_handle_fndecl_attribute): Handle function_return.
7535 (ix86_attribute_table): Add function_return.
7536 * config/i386/i386.h (machine_function): Add
7537 function_return_type.
7538 * config/i386/i386.md (simple_return_internal): Use
7539 ix86_output_function_return.
7540 (simple_return_internal_long): Likewise.
7541 * config/i386/i386.opt (mfunction-return=): New option.
7542 (indirect_branch): Mention -mfunction-return=.
7543 * doc/extend.texi: Document function_return function attribute.
7544 * doc/invoke.texi: Document -mfunction-return= option.
7546 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7548 * config/i386/i386-opts.h (indirect_branch): New.
7549 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7550 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7551 with local indirect jump when converting indirect call and jump.
7552 (ix86_set_indirect_branch_type): New.
7553 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7554 (indirectlabelno): New.
7555 (indirect_thunk_needed): Likewise.
7556 (indirect_thunk_bnd_needed): Likewise.
7557 (indirect_thunks_used): Likewise.
7558 (indirect_thunks_bnd_used): Likewise.
7559 (INDIRECT_LABEL): Likewise.
7560 (indirect_thunk_name): Likewise.
7561 (output_indirect_thunk): Likewise.
7562 (output_indirect_thunk_function): Likewise.
7563 (ix86_output_indirect_branch_via_reg): Likewise.
7564 (ix86_output_indirect_branch_via_push): Likewise.
7565 (ix86_output_indirect_branch): Likewise.
7566 (ix86_output_indirect_jmp): Likewise.
7567 (ix86_code_end): Call output_indirect_thunk_function if needed.
7568 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7570 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7571 (ix86_attribute_table): Add indirect_branch.
7572 * config/i386/i386.h (machine_function): Add indirect_branch_type
7573 and has_local_indirect_jump.
7574 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7576 (tablejump): Likewise.
7577 (*indirect_jump): Use ix86_output_indirect_jmp.
7578 (*tablejump_1): Likewise.
7579 (simple_return_indirect_internal): Likewise.
7580 * config/i386/i386.opt (mindirect-branch=): New option.
7581 (indirect_branch): New.
7584 (thunk-inline): Likewise.
7585 (thunk-extern): Likewise.
7586 * doc/extend.texi: Document indirect_branch function attribute.
7587 * doc/invoke.texi: Document -mindirect-branch= option.
7589 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7592 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7594 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7596 * ipa-inline.c (want_inline_small_function_p): Return false if
7597 inlining has already failed with CIF_FINAL_ERROR.
7598 (update_caller_keys): Call want_inline_small_function_p before
7600 (update_callee_keys): Likewise.
7602 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7604 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7606 (rs6000_quadword_masked_address_p): Likewise.
7607 (quad_aligned_load_p): Likewise.
7608 (quad_aligned_store_p): Likewise.
7609 (const_load_sequence_p): Add comment to describe the outer-most loop.
7610 (mimic_memory_attributes_and_flags): New function.
7611 (rs6000_gen_stvx): Likewise.
7612 (replace_swapped_aligned_store): Likewise.
7613 (rs6000_gen_lvx): Likewise.
7614 (replace_swapped_aligned_load): Likewise.
7615 (replace_swapped_load_constant): Capitalize argument name in
7616 comment describing this function.
7617 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7618 and stores that access quad-word aligned addresses and replace
7619 with stvx or lvx instructions when appropriate.
7620 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7621 New function prototype.
7622 (rs6000_quadword_masked_address_p): Likewise.
7623 (rs6000_gen_lvx): Likewise.
7624 (rs6000_gen_stvx): Likewise.
7625 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7626 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7627 when memory address is aligned.
7628 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7629 this split to select lvx instruction when memory address is aligned.
7630 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7631 instruction when memory address is aligned.
7632 (*vsx_le_perm_load_v16qi): Likewise.
7633 (four unnamed splitters): Modify to select the stvx instruction
7634 when memory is aligned.
7636 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7638 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7639 which appears in the queue multiple times.
7641 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7642 Alan Hayward <alan.hayward@arm.com>
7643 David Sherwood <david.sherwood@arm.com>
7645 * tree-vectorizer.h (vec_lower_bound): New structure.
7646 (_loop_vec_info): Add check_nonzero and lower_bounds.
7647 (LOOP_VINFO_CHECK_NONZERO): New macro.
7648 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7649 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7650 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7651 fields. Make seg_len the distance travelled, not including the
7653 (dr_direction_indicator): Declare.
7654 (dr_zero_step_indicator): Likewise.
7655 (dr_known_forward_stride_p): Likewise.
7656 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7658 (runtime_alias_check_p): Allow runtime alias checks with
7660 (operator ==): Compare access_size and align.
7661 (prune_runtime_alias_test_list): Rework for new distinction between
7662 the access_size and seg_len.
7663 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7665 (get_segment_min_max): New function.
7666 (create_intersect_range_checks): Use it.
7667 (dr_step_indicator): New function.
7668 (dr_direction_indicator): Likewise.
7669 (dr_zero_step_indicator): Likewise.
7670 (dr_known_forward_stride_p): Likewise.
7671 * tree-loop-distribution.c (data_ref_segment_size): Return
7672 DR_STEP * (niters - 1).
7673 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7675 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7676 (vect_preserves_scalar_order_p): New function, split out from...
7677 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7678 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7679 (vect_vfa_access_size): New function.
7680 (vect_vfa_align): Likewise.
7681 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7682 (dump_lower_bound): New function.
7683 (vect_check_lower_bound): Likewise.
7684 (vect_small_gap_p): Likewise.
7685 (vectorizable_with_step_bound_p): Likewise.
7686 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7687 depencies if the vectorization factor is 1. Convert the checks
7688 for nonzero steps into checks on the bounds of DR_STEP. Try using
7689 a bunds check for variable steps if the minimum required step is
7690 relatively small. Update calls to the dr_with_seg_len
7691 constructor and to vect_compile_time_alias.
7692 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7694 (vect_loop_versioning): Call it.
7695 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7697 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7699 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7700 Alan Hayward <alan.hayward@arm.com>
7701 David Sherwood <david.sherwood@arm.com>
7703 * doc/sourcebuild.texi (vect_scatter_store): Document.
7704 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7706 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7708 * genopinit.c (main): Add supports_vec_scatter_store and
7709 supports_vec_scatter_store_cached to target_optabs.
7710 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7711 IFN_MASK_SCATTER_STORE.
7712 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7714 * internal-fn.h (internal_store_fn_p): Declare.
7715 (internal_fn_stored_value_index): Likewise.
7716 * internal-fn.c (scatter_store_direct): New macro.
7717 (expand_scatter_store_optab_fn): New function.
7718 (direct_scatter_store_optab_supported_p): New macro.
7719 (internal_store_fn_p): New function.
7720 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7721 IFN_MASK_SCATTER_STORE.
7722 (internal_fn_mask_index): Likewise.
7723 (internal_fn_stored_value_index): New function.
7724 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7726 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7727 * optabs-query.c (supports_vec_scatter_store_p): New function.
7728 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7729 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7730 true for scatter stores.
7731 (vect_gather_scatter_fn_p): Handle scatter stores too.
7732 (vect_check_gather_scatter): Consider using scatter stores if
7733 supports_vec_scatter_store_p.
7734 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7736 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7737 internal_fn_stored_value_index.
7738 (check_load_store_masking): Handle scatter stores too.
7739 (vect_get_store_rhs): Make public.
7740 (vectorizable_call): Use internal_store_fn_p.
7741 (vectorizable_store): Handle scatter store internal functions.
7742 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7743 when deciding whether the end of the group has been reached.
7744 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7745 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7746 (mask_scatter_store<mode>): New insns.
7748 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7749 Alan Hayward <alan.hayward@arm.com>
7750 David Sherwood <david.sherwood@arm.com>
7752 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7753 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7754 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7756 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7757 Use vect_truncate_gather_scatter_offset if we can't treat the
7758 operation as a normal gather load or scatter store.
7759 (get_group_load_store_type): Take the gather_scatter_info
7760 as argument. Try using a gather load or scatter store for
7761 single-element groups.
7762 (get_load_store_type): Update calls to get_group_load_store_type
7763 and vect_use_strided_gather_scatters_p.
7765 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7766 Alan Hayward <alan.hayward@arm.com>
7767 David Sherwood <david.sherwood@arm.com>
7769 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7770 optional tree argument.
7771 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7773 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7774 but continue to use the current value as a fallback.
7775 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7776 to compare the updates.
7777 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7778 (get_load_store_type): Use it when handling a strided access.
7779 (vect_get_strided_load_store_ops): New function.
7780 (vect_get_data_ptr_increment): Likewise.
7781 (vectorizable_load): Handle strided gather loads. Always pass
7782 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7784 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7785 Alan Hayward <alan.hayward@arm.com>
7786 David Sherwood <david.sherwood@arm.com>
7788 * doc/md.texi (gather_load@var{m}): Document.
7789 (mask_gather_load@var{m}): Likewise.
7790 * genopinit.c (main): Add supports_vec_gather_load and
7791 supports_vec_gather_load_cached to target_optabs.
7792 * optabs-tree.c (init_tree_optimization_optabs): Use
7793 ggc_cleared_alloc to allocate target_optabs.
7794 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7795 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7797 * internal-fn.h (internal_load_fn_p): Declare.
7798 (internal_gather_scatter_fn_p): Likewise.
7799 (internal_fn_mask_index): Likewise.
7800 (internal_gather_scatter_fn_supported_p): Likewise.
7801 * internal-fn.c (gather_load_direct): New macro.
7802 (expand_gather_load_optab_fn): New function.
7803 (direct_gather_load_optab_supported_p): New macro.
7804 (direct_internal_fn_optab): New function.
7805 (internal_load_fn_p): Likewise.
7806 (internal_gather_scatter_fn_p): Likewise.
7807 (internal_fn_mask_index): Likewise.
7808 (internal_gather_scatter_fn_supported_p): Likewise.
7809 * optabs-query.c (supports_at_least_one_mode_p): New function.
7810 (supports_vec_gather_load_p): Likewise.
7811 * optabs-query.h (supports_vec_gather_load_p): Declare.
7812 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7813 and memory_type field.
7814 (NUM_PATTERNS): Bump to 15.
7815 * tree-vect-data-refs.c: Include internal-fn.h.
7816 (vect_gather_scatter_fn_p): New function.
7817 (vect_describe_gather_scatter_call): Likewise.
7818 (vect_check_gather_scatter): Try using internal functions for
7819 gather loads. Recognize existing calls to a gather load function.
7820 (vect_analyze_data_refs): Consider using gather loads if
7821 supports_vec_gather_load_p.
7822 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7823 (vect_get_gather_scatter_offset_type): Likewise.
7824 (vect_convert_mask_for_vectype): Likewise.
7825 (vect_add_conversion_to_patterm): Likewise.
7826 (vect_try_gather_scatter_pattern): Likewise.
7827 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7828 (vect_vect_recog_func_ptrs): Add it.
7829 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7830 internal_fn_mask_index and internal_gather_scatter_fn_p.
7831 (check_load_store_masking): Take the gather_scatter_info as an
7832 argument and handle gather loads.
7833 (vect_get_gather_scatter_ops): New function.
7834 (vectorizable_call): Check internal_load_fn_p.
7835 (vectorizable_load): Likewise. Handle gather load internal
7837 (vectorizable_store): Update call to check_load_store_masking.
7838 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7839 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7840 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7841 (aarch64_gather_scale_operand_d): New predicates.
7842 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7843 (mask_gather_load<mode>): New insns.
7845 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7846 Alan Hayward <alan.hayward@arm.com>
7847 David Sherwood <david.sherwood@arm.com>
7849 * optabs.def (fold_left_plus_optab): New optab.
7850 * doc/md.texi (fold_left_plus_@var{m}): Document.
7851 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7852 * internal-fn.c (fold_left_direct): Define.
7853 (expand_fold_left_optab_fn): Likewise.
7854 (direct_fold_left_optab_supported_p): Likewise.
7855 * fold-const-call.c (fold_const_fold_left): New function.
7856 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7857 * tree-parloops.c (valid_reduction_p): New function.
7858 (gather_scalar_reductions): Use it.
7859 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7860 (vect_finish_replace_stmt): Declare.
7861 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7862 (needs_fold_left_reduction_p): New function, split out from...
7863 (vect_is_simple_reduction): ...here. Accept reductions that
7864 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7865 (vect_force_simple_reduction): Also store the reduction type in
7866 the assignment's STMT_VINFO_REDUC_TYPE.
7867 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7868 (merge_with_identity): New function.
7869 (vect_expand_fold_left): Likewise.
7870 (vectorize_fold_left_reduction): Likewise.
7871 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7872 scalar phi in place for it. Check for target support and reject
7873 cases that would reassociate the operation. Defer the transform
7874 phase to vectorize_fold_left_reduction.
7875 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7876 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7877 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7879 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7881 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7882 call to ifc_temp_var.
7884 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7885 Alan Hayward <alan.hayward@arm.com>
7886 David Sherwood <david.sherwood@arm.com>
7888 * target.def (legitimize_address_displacement): Take the original
7889 offset as a poly_int.
7890 * targhooks.h (default_legitimize_address_displacement): Update
7892 * targhooks.c (default_legitimize_address_displacement): Likewise.
7893 * doc/tm.texi: Regenerate.
7894 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7895 as an argument, moving assert of ad->disp == ad->disp_term to...
7896 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7897 Try calling targetm.legitimize_address_displacement before expanding
7898 the address rather than afterwards, and adjust for the new interface.
7899 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7900 Match the new hook interface. Handle SVE addresses.
7901 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7904 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7906 * Makefile.in (OBJS): Add early-remat.o.
7907 * target.def (select_early_remat_modes): New hook.
7908 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7909 * doc/tm.texi: Regenerate.
7910 * targhooks.h (default_select_early_remat_modes): Declare.
7911 * targhooks.c (default_select_early_remat_modes): New function.
7912 * timevar.def (TV_EARLY_REMAT): New timevar.
7913 * passes.def (pass_early_remat): New pass.
7914 * tree-pass.h (make_pass_early_remat): Declare.
7915 * early-remat.c: New file.
7916 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7918 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7920 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7921 Alan Hayward <alan.hayward@arm.com>
7922 David Sherwood <david.sherwood@arm.com>
7924 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7925 vfm1 with a bound_epilog parameter.
7926 (vect_do_peeling): Update calls accordingly, and move the prologue
7927 call earlier in the function. Treat the base bound_epilog as 0 for
7928 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7929 this base when peeling for gaps.
7930 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7931 with fully-masked loops.
7932 (vect_estimate_min_profitable_iters): Handle the single peeled
7933 iteration in that case.
7935 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7936 Alan Hayward <alan.hayward@arm.com>
7937 David Sherwood <david.sherwood@arm.com>
7939 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7940 single-element interleaving even if the size is not a power of 2.
7941 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7942 accesses for single-element interleaving if the group size is
7945 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7946 Alan Hayward <alan.hayward@arm.com>
7947 David Sherwood <david.sherwood@arm.com>
7949 * doc/md.texi (fold_extract_last_@var{m}): Document.
7950 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7951 * optabs.def (fold_extract_last_optab): New optab.
7952 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7953 * internal-fn.c (fold_extract_direct): New macro.
7954 (expand_fold_extract_optab_fn): Likewise.
7955 (direct_fold_extract_optab_supported_p): Likewise.
7956 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7957 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7958 EXTRACT_LAST_REDUCTION.
7959 (get_initial_def_for_reduction): Do not create an initial vector
7960 for EXTRACT_LAST_REDUCTION reductions.
7961 (vectorizable_reduction): Leave the scalar phi in place for
7962 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7963 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7964 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7965 transform phase to vectorizable_condition.
7966 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7968 (vect_finish_stmt_generation): ...here.
7969 (vect_finish_replace_stmt): New function.
7970 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7971 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7973 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7975 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7976 Alan Hayward <alan.hayward@arm.com>
7977 David Sherwood <david.sherwood@arm.com>
7979 * doc/md.texi (extract_last_@var{m}): Document.
7980 * optabs.def (extract_last_optab): New optab.
7981 * internal-fn.def (EXTRACT_LAST): New internal function.
7982 * internal-fn.c (cond_unary_direct): New macro.
7983 (expand_cond_unary_optab_fn): Likewise.
7984 (direct_cond_unary_optab_supported_p): Likewise.
7985 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7986 loops using EXTRACT_LAST.
7987 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7988 (extract_last_<mode>): ...this optab.
7989 (vec_extract<mode><Vel>): Update accordingly.
7991 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7992 Alan Hayward <alan.hayward@arm.com>
7993 David Sherwood <david.sherwood@arm.com>
7995 * target.def (empty_mask_is_expensive): New hook.
7996 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7997 * doc/tm.texi: Regenerate.
7998 * targhooks.h (default_empty_mask_is_expensive): Declare.
7999 * targhooks.c (default_empty_mask_is_expensive): New function.
8000 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8001 if the target says that empty masks are expensive.
8002 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8004 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8006 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8007 Alan Hayward <alan.hayward@arm.com>
8008 David Sherwood <david.sherwood@arm.com>
8010 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8011 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8012 (vect_use_loop_mask_for_alignment_p): New function.
8013 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8014 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8015 niters_skip argument. Make sure that the first niters_skip elements
8016 of the first iteration are inactive.
8017 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8018 Update call to vect_set_loop_masks_directly.
8019 (get_misalign_in_elems): New function, split out from...
8020 (vect_gen_prolog_loop_niters): ...here.
8021 (vect_update_init_of_dr): Take a code argument that specifies whether
8022 the adjustment should be added or subtracted.
8023 (vect_update_init_of_drs): Likewise.
8024 (vect_prepare_for_masked_peels): New function.
8025 (vect_do_peeling): Skip prologue peeling if we're using a mask
8026 instead. Update call to vect_update_inits_of_drs.
8027 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8029 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8030 alignment. Do not include the number of peeled iterations in
8031 the minimum threshold in that case.
8032 (vectorizable_induction): Adjust the start value down by
8033 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8034 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8035 Take the number of skipped iterations into account when calculating
8037 * tree-vect-stmts.c (vect_gen_while_not): New function.
8039 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8040 Alan Hayward <alan.hayward@arm.com>
8041 David Sherwood <david.sherwood@arm.com>
8043 * doc/sourcebuild.texi (vect_fully_masked): Document.
8044 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8046 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8048 (vect_analyze_loop_2): ...here. Don't check the vectorization
8049 factor against the number of loop iterations if the loop is
8052 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8053 Alan Hayward <alan.hayward@arm.com>
8054 David Sherwood <david.sherwood@arm.com>
8056 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8057 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8058 (dump_groups): Update accordingly.
8059 (iv_use::mem_type): New member variable.
8060 (address_p): New function.
8061 (record_use): Add a mem_type argument and initialize the new
8063 (record_group_use): Add a mem_type argument. Use address_p.
8064 Remove obsolete null checks of base_object. Update call to record_use.
8065 (find_interesting_uses_op): Update call to record_group_use.
8066 (find_interesting_uses_cond): Likewise.
8067 (find_interesting_uses_address): Likewise.
8068 (get_mem_type_for_internal_fn): New function.
8069 (find_address_like_use): Likewise.
8070 (find_interesting_uses_stmt): Try find_address_like_use before
8071 calling find_interesting_uses_op.
8072 (addr_offset_valid_p): Use the iv mem_type field as the type
8073 of the addressed memory.
8074 (add_autoinc_candidates): Likewise.
8075 (get_address_cost): Likewise.
8076 (split_small_address_groups_p): Use address_p.
8077 (split_address_groups): Likewise.
8078 (add_iv_candidate_for_use): Likewise.
8079 (autoinc_possible_for_pair): Likewise.
8080 (rewrite_groups): Likewise.
8081 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8082 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8083 (get_alias_ptr_type_for_ptr_address): New function.
8084 (rewrite_use_address): Rewrite address uses in calls that were
8085 identified by find_address_like_use.
8087 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8088 Alan Hayward <alan.hayward@arm.com>
8089 David Sherwood <david.sherwood@arm.com>
8091 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8093 * gimple-expr.h (is_gimple_addressable: Likewise.
8094 * gimple-expr.c (is_gimple_address): Likewise.
8095 * internal-fn.c (expand_call_mem_ref): New function.
8096 (expand_mask_load_optab_fn): Use it.
8097 (expand_mask_store_optab_fn): Likewise.
8099 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8100 Alan Hayward <alan.hayward@arm.com>
8101 David Sherwood <david.sherwood@arm.com>
8103 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8104 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8105 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8106 (cond_umax@var{mode}): Document.
8107 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8108 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8109 (cond_umin_optab, cond_umax_optab): New optabs.
8110 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8111 (COND_IOR, COND_XOR): New internal functions.
8112 * internal-fn.h (get_conditional_internal_fn): Declare.
8113 * internal-fn.c (cond_binary_direct): New macro.
8114 (expand_cond_binary_optab_fn): Likewise.
8115 (direct_cond_binary_optab_supported_p): Likewise.
8116 (get_conditional_internal_fn): New function.
8117 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8118 Cope with reduction statements that are vectorized as calls rather
8120 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8121 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8122 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8123 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8124 (UNSPEC_COND_EOR): New unspecs.
8125 (optab): Add mappings for them.
8126 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8127 (sve_int_op, sve_fp_op): New int attributes.
8129 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8130 Alan Hayward <alan.hayward@arm.com>
8131 David Sherwood <david.sherwood@arm.com>
8133 * optabs.def (while_ult_optab): New optab.
8134 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8135 * internal-fn.def (WHILE_ULT): New internal function.
8136 * internal-fn.h (direct_internal_fn_supported_p): New override
8137 that takes two types as argument.
8138 * internal-fn.c (while_direct): New macro.
8139 (expand_while_optab_fn): New function.
8140 (convert_optab_supported_p): Likewise.
8141 (direct_while_optab_supported_p): New macro.
8142 * wide-int.h (wi::udiv_ceil): New function.
8143 * tree-vectorizer.h (rgroup_masks): New structure.
8144 (vec_loop_masks): New typedef.
8145 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8147 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8148 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8149 (vect_max_vf): New function.
8150 (slpeel_make_loop_iterate_ntimes): Delete.
8151 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8152 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8153 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8154 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8155 internal-fn.h, stor-layout.h and optabs-query.h.
8156 (vect_set_loop_mask): New function.
8157 (add_preheader_seq): Likewise.
8158 (add_header_seq): Likewise.
8159 (interleave_supported_p): Likewise.
8160 (vect_maybe_permute_loop_masks): Likewise.
8161 (vect_set_loop_masks_directly): Likewise.
8162 (vect_set_loop_condition_masked): Likewise.
8163 (vect_set_loop_condition_unmasked): New function, split out from
8164 slpeel_make_loop_iterate_ntimes.
8165 (slpeel_make_loop_iterate_ntimes): Rename to..
8166 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8167 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8168 (vect_do_peeling): Update call accordingly.
8169 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8171 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8172 mask_compare_type, can_fully_mask_p and fully_masked_p.
8173 (release_vec_loop_masks): New function.
8174 (_loop_vec_info): Use it to free the loop masks.
8175 (can_produce_all_loop_masks_p): New function.
8176 (vect_get_max_nscalars_per_iter): Likewise.
8177 (vect_verify_full_masking): Likewise.
8178 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8179 retries, and free the mask rgroups before retrying. Check loop-wide
8180 reasons for disallowing fully-masked loops. Make the final decision
8181 about whether use a fully-masked loop or not.
8182 (vect_estimate_min_profitable_iters): Do not assume that peeling
8183 for the number of iterations will be needed for fully-masked loops.
8184 (vectorizable_reduction): Disable fully-masked loops.
8185 (vectorizable_live_operation): Likewise.
8186 (vect_halve_mask_nunits): New function.
8187 (vect_double_mask_nunits): Likewise.
8188 (vect_record_loop_mask): Likewise.
8189 (vect_get_loop_mask): Likewise.
8190 (vect_transform_loop): Handle the case in which the final loop
8191 iteration might handle a partial vector. Call vect_set_loop_condition
8192 instead of slpeel_make_loop_iterate_ntimes.
8193 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8194 (check_load_store_masking): New function.
8195 (prepare_load_store_mask): Likewise.
8196 (vectorizable_store): Handle fully-masked loops.
8197 (vectorizable_load): Likewise.
8198 (supportable_widening_operation): Use vect_halve_mask_nunits for
8200 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8201 (vect_gen_while): New function.
8202 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8203 (aarch64_uqdec<mode>): New insn.
8205 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8206 Alan Hayward <alan.hayward@arm.com>
8207 David Sherwood <david.sherwood@arm.com>
8209 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8210 (reduc_xor_scal_optab): New optabs.
8211 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8212 (reduc_xor_scal_@var{m}): Document.
8213 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8214 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8216 * fold-const-call.c (fold_const_call): Handle them.
8217 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8218 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8219 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8220 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8221 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8222 (UNSPEC_XORV): New unspecs.
8223 (optab): Add entries for them.
8224 (BITWISEV): New int iterator.
8225 (bit_reduc_op): New int attributes.
8227 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8228 Alan Hayward <alan.hayward@arm.com>
8229 David Sherwood <david.sherwood@arm.com>
8231 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8232 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8233 * optabs.def (vec_shl_insert_optab): New optab.
8234 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8235 (duplicate_and_interleave): Likewise.
8236 * tree-vect-loop.c: Include internal-fn.h.
8237 (neutral_op_for_slp_reduction): New function, split out from
8238 get_initial_defs_for_reduction.
8239 (get_initial_def_for_reduction): Handle option 2 for variable-length
8240 vectors by loading the neutral value into a vector and then shifting
8241 the initial value into element 0.
8242 (get_initial_defs_for_reduction): Replace the code argument with
8243 the neutral value calculated by neutral_op_for_slp_reduction.
8244 Use gimple_build_vector for constant-length vectors.
8245 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8246 but the first group_size elements have a neutral value.
8247 Use duplicate_and_interleave otherwise.
8248 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8249 Update call to get_initial_defs_for_reduction. Handle SLP
8250 reductions for variable-length vectors by creating one vector
8251 result for each scalar result, with the elements associated
8252 with other scalar results stubbed out with the neutral value.
8253 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8254 Require IFN_VEC_SHL_INSERT for double reductions on
8255 variable-length vectors, or SLP reductions that have
8256 a neutral value. Require can_duplicate_and_interleave_p
8257 support for variable-length unchained SLP reductions if there
8258 is no neutral value, such as for MIN/MAX reductions. Also require
8259 the number of vector elements to be a multiple of the number of
8260 SLP statements when doing variable-length unchained SLP reductions.
8261 Update call to vect_create_epilog_for_reduction.
8262 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8263 and remove initial values.
8264 (duplicate_and_interleave): Make public.
8265 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8266 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8268 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8269 Alan Hayward <alan.hayward@arm.com>
8270 David Sherwood <david.sherwood@arm.com>
8272 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8273 (can_duplicate_and_interleave_p): New function.
8274 (vect_get_and_check_slp_defs): Take the vector of statements
8275 rather than just the current one. Remove excess parentheses.
8276 Restriction rejectinon of vect_constant_def and vect_external_def
8277 for variable-length vectors to boolean types, or types for which
8278 can_duplicate_and_interleave_p is false.
8279 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8280 (duplicate_and_interleave): New function.
8281 (vect_get_constant_vectors): Use gimple_build_vector for
8282 constant-length vectors and suitable variable-length constant
8283 vectors. Use duplicate_and_interleave for other variable-length
8284 vectors. Don't defer the update when inserting new statements.
8286 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8287 Alan Hayward <alan.hayward@arm.com>
8288 David Sherwood <david.sherwood@arm.com>
8290 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8291 min_profitable_iters doesn't go negative.
8293 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8294 Alan Hayward <alan.hayward@arm.com>
8295 David Sherwood <david.sherwood@arm.com>
8297 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8298 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8299 * optabs.def (vec_mask_load_lanes_optab): New optab.
8300 (vec_mask_store_lanes_optab): Likewise.
8301 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8302 (MASK_STORE_LANES): Likewise.
8303 * internal-fn.c (mask_load_lanes_direct): New macro.
8304 (mask_store_lanes_direct): Likewise.
8305 (expand_mask_load_optab_fn): Handle masked operations.
8306 (expand_mask_load_lanes_optab_fn): New macro.
8307 (expand_mask_store_optab_fn): Handle masked operations.
8308 (expand_mask_store_lanes_optab_fn): New macro.
8309 (direct_mask_load_lanes_optab_supported_p): Likewise.
8310 (direct_mask_store_lanes_optab_supported_p): Likewise.
8311 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8313 (vect_load_lanes_supported): Likewise.
8314 * tree-vect-data-refs.c (strip_conversion): New function.
8315 (can_group_stmts_p): Likewise.
8316 (vect_analyze_data_ref_accesses): Use it instead of checking
8317 for a pair of assignments.
8318 (vect_store_lanes_supported): Take a masked_p parameter.
8319 (vect_load_lanes_supported): Likewise.
8320 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8321 vect_store_lanes_supported and vect_load_lanes_supported.
8322 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8323 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8324 parameter. Don't allow gaps for masked accesses.
8325 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8326 and vect_load_lanes_supported.
8327 (get_load_store_type): Take a masked_p parameter and update
8328 call to get_group_load_store_type.
8329 (vectorizable_store): Update call to get_load_store_type.
8330 Handle IFN_MASK_STORE_LANES.
8331 (vectorizable_load): Update call to get_load_store_type.
8332 Handle IFN_MASK_LOAD_LANES.
8334 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8335 Alan Hayward <alan.hayward@arm.com>
8336 David Sherwood <david.sherwood@arm.com>
8338 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8340 * config/aarch64/aarch64-protos.h
8341 (aarch64_sve_struct_memory_operand_p): Declare.
8342 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8343 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8344 (VPRED, vpred): Handle SVE structure modes.
8345 * config/aarch64/constraints.md (Utx): New constraint.
8346 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8347 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8348 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8349 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8350 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8351 structure modes. Split into pieces after RA.
8352 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8353 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8355 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8356 SVE structure modes.
8357 (aarch64_classify_address): Likewise.
8358 (sizetochar): Move earlier in file.
8359 (aarch64_print_operand): Handle SVE register lists.
8360 (aarch64_array_mode): New function.
8361 (aarch64_sve_struct_memory_operand_p): Likewise.
8362 (TARGET_ARRAY_MODE): Redefine.
8364 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8365 Alan Hayward <alan.hayward@arm.com>
8366 David Sherwood <david.sherwood@arm.com>
8368 * target.def (array_mode): New target hook.
8369 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8370 * doc/tm.texi: Regenerate.
8371 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8372 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8373 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8375 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8378 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8379 Alan Hayward <alan.hayward@arm.com>
8380 David Sherwood <david.sherwood@arm.com>
8382 * fold-const.c (fold_binary_loc): Check the argument types
8383 rather than the result type when testing for a vector operation.
8385 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8387 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8388 * doc/tm.texi: Regenerate.
8390 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8391 Alan Hayward <alan.hayward@arm.com>
8392 David Sherwood <david.sherwood@arm.com>
8394 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8395 (sve): Document new AArch64 extension.
8396 * doc/md.texi (w): Extend the description of the AArch64
8397 constraint to include SVE vectors.
8398 (Upl, Upa): Document new AArch64 predicate constraints.
8399 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8401 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8402 (msve-vector-bits=): New option.
8403 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8404 SVE when these are disabled.
8405 (sve): New extension.
8406 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8407 modes. Adjust their number of units based on aarch64_sve_vg.
8408 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8409 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8410 aarch64_addr_query_type.
8411 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8412 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8413 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8414 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8415 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8416 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8417 (aarch64_simd_imm_zero_p): Delete.
8418 (aarch64_check_zero_based_sve_index_immediate): Declare.
8419 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8420 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8421 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8422 (aarch64_sve_float_mul_immediate_p): Likewise.
8423 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8425 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8426 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8427 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8428 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8429 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8430 (aarch64_regmode_natural_size): Likewise.
8431 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8432 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8434 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8435 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8436 for VG and the SVE predicate registers.
8437 (V_ALIASES): Add a "z"-prefixed alias.
8438 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8439 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8440 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8441 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8442 (REG_CLASS_NAMES): Add entries for them.
8443 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8444 and the predicate registers.
8445 (aarch64_sve_vg): Declare.
8446 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8447 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8448 (REGMODE_NATURAL_SIZE): Define.
8449 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8451 * config/aarch64/aarch64.c: Include cfgrtl.h.
8452 (simd_immediate_info): Add a constructor for series vectors,
8453 and an associated step field.
8454 (aarch64_sve_vg): New variable.
8455 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8456 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8457 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8458 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8459 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8460 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8461 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8462 (aarch64_get_mask_mode): New functions.
8463 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8464 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8465 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8466 predicate modes and predicate registers. Explicitly restrict
8467 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8468 to store a vector mode if it is recognized by
8469 aarch64_classify_vector_mode.
8470 (aarch64_regmode_natural_size): New function.
8471 (aarch64_hard_regno_caller_save_mode): Return the original mode
8473 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8474 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8475 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8476 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8478 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8479 does not overlap dest if the function is frame-related. Handle
8481 (aarch64_split_add_offset): New function.
8482 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8483 them aarch64_add_offset.
8484 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8485 and update call to aarch64_sub_sp.
8486 (aarch64_add_cfa_expression): New function.
8487 (aarch64_expand_prologue): Pass extra temporary registers to the
8488 functions above. Handle the case in which we need to emit new
8489 DW_CFA_expressions for registers that were originally saved
8490 relative to the stack pointer, but now have to be expressed
8491 relative to the frame pointer.
8492 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8494 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8495 IP0 and IP1 values for SVE frames.
8496 (aarch64_expand_vec_series): New function.
8497 (aarch64_expand_sve_widened_duplicate): Likewise.
8498 (aarch64_expand_sve_const_vector): Likewise.
8499 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8500 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8501 into the register, rather than emitting a SET directly.
8502 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8503 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8504 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8505 (offset_9bit_signed_scaled_p): New functions.
8506 (aarch64_replicate_bitmask_imm): New function.
8507 (aarch64_bitmask_imm): Use it.
8508 (aarch64_cannot_force_const_mem): Reject expressions involving
8509 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8510 (aarch64_classify_index): Handle SVE indices, by requiring
8511 a plain register index with a scale that matches the element size.
8512 (aarch64_classify_address): Handle SVE addresses. Assert that
8513 the mode of the address is VOIDmode or an integer mode.
8514 Update call to aarch64_classify_symbol.
8515 (aarch64_classify_symbolic_expression): Update call to
8516 aarch64_classify_symbol.
8517 (aarch64_const_vec_all_in_range_p): New function.
8518 (aarch64_print_vector_float_operand): Likewise.
8519 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8520 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8521 and the FP immediates 1.0 and 0.5.
8522 (aarch64_print_address_internal): Handle SVE addresses.
8523 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8524 (aarch64_regno_regclass): Handle predicate registers.
8525 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8527 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8528 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8529 (aarch64_convert_sve_vector_bits): New function.
8530 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8531 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8533 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8534 Handle SVE vector and predicate modes. Accept VL-based constants
8535 that need only one temporary register, and VL offsets that require
8536 no temporary registers.
8537 (aarch64_conditional_register_usage): Mark the predicate registers
8538 as fixed if SVE isn't available.
8539 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8540 Return true for SVE vector and predicate modes.
8541 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8542 rather than an unsigned int. Handle SVE modes.
8543 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8545 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8547 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8548 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8549 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8550 (aarch64_sve_float_mul_immediate_p): New functions.
8551 (aarch64_sve_valid_immediate): New function.
8552 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8553 Explicitly reject structure modes. Check for INDEX constants.
8554 Handle PTRUE and PFALSE constants.
8555 (aarch64_check_zero_based_sve_index_immediate): New function.
8556 (aarch64_simd_imm_zero_p): Delete.
8557 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8558 vector modes. Accept constants in the range of CNT[BHWD].
8559 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8560 ask for an Advanced SIMD mode.
8561 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8562 (aarch64_simd_vector_alignment): Handle SVE predicates.
8563 (aarch64_vectorize_preferred_vector_alignment): New function.
8564 (aarch64_simd_vector_alignment_reachable): Use it instead of
8566 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8567 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8569 (MAX_VECT_LEN): Delete.
8570 (expand_vec_perm_d): Add a vec_flags field.
8571 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8572 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8573 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8575 (aarch64_evpc_rev): Rename to...
8576 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8577 (aarch64_evpc_rev_global): New function.
8578 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8579 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8581 (aarch64_evpc_sve_tbl): New function.
8582 (aarch64_expand_vec_perm_const_1): Update after rename of
8583 aarch64_evpc_rev. Handle SVE permutes too, trying
8584 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8585 than aarch64_evpc_tbl.
8586 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8587 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8588 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8589 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8590 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8591 (aarch64_expand_sve_vcond): New functions.
8592 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8593 of aarch64_vector_mode_p.
8594 (aarch64_dwarf_poly_indeterminate_value): New function.
8595 (aarch64_compute_pressure_classes): Likewise.
8596 (aarch64_can_change_mode_class): Likewise.
8597 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8598 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8599 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8600 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8601 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8602 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8603 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8604 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8606 (Dn, Dl, Dr): Accept const as well as const_vector.
8607 (Dz): Likewise. Compare against CONST0_RTX.
8608 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8609 of "vector" where appropriate.
8610 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8611 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8612 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8613 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8614 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8615 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8616 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8617 (v_int_equiv): Extend to SVE modes.
8618 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8620 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8621 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8622 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8623 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8624 (SVE_COND_FP_CMP): New int iterators.
8625 (perm_hilo): Handle the new unpack unspecs.
8626 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8628 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8629 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8630 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8631 (aarch64_equality_operator, aarch64_constant_vector_operand)
8632 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8633 (aarch64_sve_nonimmediate_operand): Likewise.
8634 (aarch64_sve_general_operand): Likewise.
8635 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8636 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8637 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8638 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8639 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8640 (aarch64_sve_float_arith_immediate): Likewise.
8641 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8642 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8643 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8644 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8645 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8646 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8647 (aarch64_sve_float_arith_operand): Likewise.
8648 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8649 (aarch64_sve_float_mul_operand): Likewise.
8650 (aarch64_sve_vec_perm_operand): Likewise.
8651 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8652 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8653 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8654 as well as const_vector.
8655 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8656 in file. Use CONST0_RTX and CONSTM1_RTX.
8657 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8658 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8659 Use aarch64_simd_imm_zero.
8660 * config/aarch64/aarch64-sve.md: New file.
8661 * config/aarch64/aarch64.md: Include it.
8662 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8663 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8664 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8665 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8666 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8667 (sve): New attribute.
8668 (enabled): Disable instructions with the sve attribute unless
8670 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8671 aarch64_expand_mov_immediate.
8672 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8673 CNT[BHSD] immediates.
8674 (movti): Split CONST_POLY_INT moves into two halves.
8675 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8676 Split additions that need a temporary here if the destination
8677 is the stack pointer.
8678 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8679 (*add<mode>3_poly_1): New instruction.
8680 (set_clobber_cc): New expander.
8682 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8684 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8685 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8686 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8687 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8688 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8689 Change innermode from fixed_mode_size to machine_mode.
8690 (simplify_subreg): Update call accordingly. Handle a constant-sized
8691 subreg of a variable-length CONST_VECTOR.
8693 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8694 Alan Hayward <alan.hayward@arm.com>
8695 David Sherwood <david.sherwood@arm.com>
8697 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8698 (add_offset_to_base): New function, split out from...
8699 (create_mem_ref): ...here. When handling a scale other than 1,
8700 check first whether the address is valid without the offset.
8701 Add it into the base if so, leaving the index and scale as-is.
8703 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8706 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8707 fold_for_warn before checking if arg2 is INTEGER_CST.
8709 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8711 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8712 (store_multiple_operation): Delete.
8713 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8714 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8715 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8716 guarded by TARGET_STRING.
8717 (rs6000_output_load_multiple): Delete.
8718 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8719 OPTION_MASK_STRING / TARGET_STRING handling.
8720 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8721 (const rs6000_opt_masks) <"string">: Change mask to 0.
8722 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8723 (MASK_STRING): Delete.
8724 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8726 (load_multiple): Delete.
8733 (store_multiple): Delete.
8740 (movmemsi_8reg): Delete.
8741 (corresponding unnamed define_insn): Delete.
8742 (movmemsi_6reg): Delete.
8743 (corresponding unnamed define_insn): Delete.
8744 (movmemsi_4reg): Delete.
8745 (corresponding unnamed define_insn): Delete.
8746 (movmemsi_2reg): Delete.
8747 (corresponding unnamed define_insn): Delete.
8748 (movmemsi_1reg): Delete.
8749 (corresponding unnamed define_insn): Delete.
8750 * config/rs6000/rs6000.opt (mno-string): New.
8751 (mstring): Replace by deprecation warning stub.
8752 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8754 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8756 * regrename.c (regrename_do_replace): If replacing the same
8757 reg multiple times, try to reuse last created gen_raw_REG.
8760 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8761 main to workaround a bug in GDB.
8763 2018-01-12 Tom de Vries <tom@codesourcery.com>
8766 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8768 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8770 PR rtl-optimization/80481
8771 * ira-color.c (get_cap_member): New function.
8772 (allocnos_conflict_by_live_ranges_p): Use it.
8773 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8774 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8776 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8779 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8780 (*saddl_se_1): Ditto.
8782 (*ssubl_se_1): Ditto.
8784 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8786 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8787 rather than wi::to_widest for DR_INITs.
8788 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8789 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8790 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8792 (vect_analyze_group_access_1): Note that here.
8794 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8796 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8797 polynomial type sizes.
8799 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8801 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8802 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8803 (gimple_add_tmp_var): Likewise.
8805 2018-01-12 Martin Liska <mliska@suse.cz>
8807 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8808 (gimple_alloc_sizes): Likewise.
8809 (dump_gimple_statistics): Use PRIu64 in printf format.
8810 * gimple.h: Change uint64_t to int.
8812 2018-01-12 Martin Liska <mliska@suse.cz>
8814 * tree-core.h: Use uint64_t instead of int.
8815 * tree.c (tree_node_counts): Likewise.
8816 (tree_node_sizes): Likewise.
8817 (dump_tree_statistics): Use PRIu64 in printf format.
8819 2018-01-12 Martin Liska <mliska@suse.cz>
8821 * Makefile.in: As qsort_chk is implemented in vec.c, add
8822 vec.o to linkage of gencfn-macros.
8823 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8824 passing the info to record_node_allocation_statistics.
8825 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8827 * ggc-common.c (struct ggc_usage): Add operator== and use
8828 it in operator< and compare function.
8829 * mem-stats.h (struct mem_usage): Likewise.
8830 * vec.c (struct vec_usage): Remove operator< and compare
8831 function. Can be simply inherited.
8833 2018-01-12 Martin Jambor <mjambor@suse.cz>
8836 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8837 * tree-ssa-math-opts.c: Include domwalk.h.
8838 (convert_mult_to_fma_1): New function.
8839 (fma_transformation_info): New type.
8840 (fma_deferring_state): Likewise.
8841 (cancel_fma_deferring): New function.
8842 (result_of_phi): Likewise.
8843 (last_fma_candidate_feeds_initial_phi): Likewise.
8844 (convert_mult_to_fma): Added deferring logic, split actual
8845 transformation to convert_mult_to_fma_1.
8846 (math_opts_dom_walker): New type.
8847 (math_opts_dom_walker::after_dom_children): New method, body moved
8848 here from pass_optimize_widening_mul::execute, added deferring logic
8850 (pass_optimize_widening_mul::execute): Moved most of code to
8851 math_opts_dom_walker::after_dom_children.
8852 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8853 * config/i386/i386.c (ix86_option_override_internal): Added
8854 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8856 2018-01-12 Richard Biener <rguenther@suse.de>
8859 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8860 inline instance vars.
8862 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8865 * config/rx/rx.c (rx_is_restricted_memory_address):
8868 2018-01-12 Richard Biener <rguenther@suse.de>
8870 PR tree-optimization/80846
8871 * target.def (split_reduction): New target hook.
8872 * targhooks.c (default_split_reduction): New function.
8873 * targhooks.h (default_split_reduction): Declare.
8874 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8875 target requests first reduce vectors by combining low and high
8877 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8878 (get_vectype_for_scalar_type_and_size): Export.
8879 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8880 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8881 * doc/tm.texi: Regenerate.
8882 * config/i386/i386.c (ix86_split_reduction): Implement
8883 TARGET_VECTORIZE_SPLIT_REDUCTION.
8885 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8888 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8889 in PIC mode except for TARGET_VXWORKS_RTP.
8890 * config/sparc/sparc.c: Include cfgrtl.h.
8891 (TARGET_INIT_PIC_REG): Define.
8892 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8893 (sparc_pic_register_p): New predicate.
8894 (sparc_legitimate_address_p): Use it.
8895 (sparc_legitimize_pic_address): Likewise.
8896 (sparc_delegitimize_address): Likewise.
8897 (sparc_mode_dependent_address_p): Likewise.
8898 (gen_load_pcrel_sym): Remove 4th parameter.
8899 (load_got_register): Adjust call to above. Remove obsolete stuff.
8900 (sparc_expand_prologue): Do not call load_got_register here.
8901 (sparc_flat_expand_prologue): Likewise.
8902 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8903 (sparc_use_pseudo_pic_reg): New function.
8904 (sparc_init_pic_reg): Likewise.
8905 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8906 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8908 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8910 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8911 Add item for branch_cost.
8913 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8915 PR rtl-optimization/83565
8916 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8917 not extend the result to a larger mode for rotate operations.
8918 (num_sign_bit_copies1): Likewise.
8920 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8923 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8925 Use values-Xc.o for -pedantic.
8926 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8928 2018-01-12 Martin Liska <mliska@suse.cz>
8931 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8933 (possible_polymorphic_call_targets): Use it.
8934 (ipa_devirt): Likewise.
8936 2018-01-12 Martin Liska <mliska@suse.cz>
8938 * profile-count.h (enum profile_quality): Use 0 as invalid
8939 enum value of profile_quality.
8941 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8943 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8944 -mext-string options.
8946 2018-01-12 Richard Biener <rguenther@suse.de>
8948 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8949 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8950 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8952 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8954 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8956 * configure.ac (--with-long-double-format): Add support for the
8957 configuration option to change the default long double format on
8959 * config.gcc (powerpc*-linux*-*): Likewise.
8960 * configure: Regenerate.
8961 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8962 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8963 used without modification.
8965 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8967 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8968 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8969 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8970 MISC_BUILTIN_SPEC_BARRIER.
8971 (rs6000_init_builtins): Likewise.
8972 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8974 (speculation_barrier): New define_insn.
8975 * doc/extend.texi: Document __builtin_speculation_barrier.
8977 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8980 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8981 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8982 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8984 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8985 integral modes instead of "ss" and "sd".
8986 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8987 vectors with 32-bit and 64-bit elements.
8988 (vecdupssescalarmodesuffix): New mode attribute.
8989 (vec_dup<mode>): Use it.
8991 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8994 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8995 frame if argument is passed on stack.
8997 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9000 * ree.c (combine_reaching_defs): Optimize also
9001 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9002 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9004 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9007 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9009 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9012 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9013 after they are computed.
9015 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9017 PR tree-optimization/83695
9018 * gimple-loop-linterchange.cc
9019 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9020 reset cached scev information after interchange.
9021 (pass_linterchange::execute): Remove call to scev_reset_htab.
9023 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9025 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9026 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9027 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9028 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9029 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9030 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9031 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9032 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9033 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9034 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9035 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9036 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9037 (V_lane_reg): Likewise.
9038 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9040 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9041 (vfmal_lane_low<mode>_intrinsic,
9042 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9043 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9044 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9045 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9046 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9047 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9049 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9051 * config/arm/arm-cpus.in (fp16fml): New feature.
9052 (ALL_SIMD): Add fp16fml.
9053 (armv8.2-a): Add fp16fml as an option.
9054 (armv8.3-a): Likewise.
9055 (armv8.4-a): Add fp16fml as part of fp16.
9056 * config/arm/arm.h (TARGET_FP16FML): Define.
9057 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9059 * config/arm/arm-modes.def (V2HF): Define.
9060 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9061 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9062 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9063 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9064 vfmsl_low, vfmsl_high): New set of builtins.
9065 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9066 (vfml_op): New code attribute.
9067 (VFMLHALVES): New int iterator.
9068 (VFML, VFMLSEL): New mode attributes.
9069 (V_reg): Define mapping for V2HF.
9070 (V_hi, V_lo): New mode attributes.
9071 (VF_constraint): Likewise.
9072 (vfml_half, vfml_half_selector): New int attributes.
9073 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9075 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9076 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9078 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9079 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9080 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9081 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9083 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9084 Document new effective target and option set.
9086 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9088 * config/arm/arm-cpus.in (armv8_4): New feature.
9089 (ARMv8_4a): New fgroup.
9090 (armv8.4-a): New arch.
9091 * config/arm/arm-tables.opt: Regenerate.
9092 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9093 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9094 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9095 Add matching rules for -march=armv8.4-a and extensions.
9096 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9098 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9101 * config/rx/rx.md (BW): New mode attribute.
9102 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9104 2018-01-11 Richard Biener <rguenther@suse.de>
9106 PR tree-optimization/83435
9107 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9108 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9109 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9111 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9112 Alan Hayward <alan.hayward@arm.com>
9113 David Sherwood <david.sherwood@arm.com>
9115 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9117 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9118 (aarch64_print_address_internal): Use it to check for a zero offset.
9120 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9121 Alan Hayward <alan.hayward@arm.com>
9122 David Sherwood <david.sherwood@arm.com>
9124 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9125 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9126 Return a poly_int64 rather than a HOST_WIDE_INT.
9127 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9128 rather than a HOST_WIDE_INT.
9129 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9130 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9131 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9132 final_offset from HOST_WIDE_INT to poly_int64.
9133 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9134 to_constant when getting the number of units in an Advanced SIMD
9136 (aarch64_builtin_vectorized_function): Check for a constant number
9138 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9140 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9141 attribute instead of GET_MODE_NUNITS.
9142 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9143 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9144 GET_MODE_SIZE for fixed-size registers.
9145 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9146 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9147 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9148 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9149 (aarch64_print_operand, aarch64_print_address_internal)
9150 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9151 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9152 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9153 Handle polynomial GET_MODE_SIZE.
9154 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9155 wider than SImode without modification.
9156 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9157 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9158 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9159 passing and returning SVE modes.
9160 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9161 rather than GEN_INT.
9162 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9163 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9164 (aarch64_allocate_and_probe_stack_space): Likewise.
9165 (aarch64_layout_frame): Cope with polynomial offsets.
9166 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9167 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9169 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9170 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9171 poly_int64 rather than a HOST_WIDE_INT.
9172 (aarch64_get_separate_components, aarch64_process_components)
9173 (aarch64_expand_prologue, aarch64_expand_epilogue)
9174 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9175 (aarch64_anchor_offset): New function, split out from...
9176 (aarch64_legitimize_address): ...here.
9177 (aarch64_builtin_vectorization_cost): Handle polynomial
9178 TYPE_VECTOR_SUBPARTS.
9179 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9181 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9182 number of elements from the PARALLEL rather than the mode.
9183 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9184 rather than GET_MODE_BITSIZE.
9185 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9186 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9187 (aarch64_expand_vec_perm_const_1): Handle polynomial
9188 d->perm.length () and d->perm elements.
9189 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9190 Apply to_constant to d->perm elements.
9191 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9192 polynomial CONST_VECTOR_NUNITS.
9193 (aarch64_move_pointer): Take amount as a poly_int64 rather
9195 (aarch64_progress_pointer): Avoid temporary variable.
9196 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9197 the mode attribute instead of GET_MODE.
9199 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9200 Alan Hayward <alan.hayward@arm.com>
9201 David Sherwood <david.sherwood@arm.com>
9203 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9204 x exists before using it.
9205 (aarch64_add_constant_internal): Rename to...
9206 (aarch64_add_offset_1): ...this. Replace regnum with separate
9207 src and dest rtxes. Handle the case in which they're different,
9208 including when the offset is zero. Replace scratchreg with an rtx.
9209 Use 2 additions if there is no spare register into which we can
9210 move a 16-bit constant.
9211 (aarch64_add_constant): Delete.
9212 (aarch64_add_offset): Replace reg with separate src and dest
9213 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9214 Use aarch64_add_offset_1.
9215 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9216 an rtx rather than an int. Take the delta as a poly_int64
9217 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9218 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9219 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9220 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9221 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9223 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9224 aarch64_add_constant.
9226 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9228 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9229 Use scalar_float_mode.
9231 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9233 * config/aarch64/aarch64-simd.md
9234 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9235 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9236 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9237 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9238 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9239 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9240 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9241 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9242 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9243 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9245 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9248 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9249 targ_options->x_arm_arch_string is non NULL.
9251 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9253 * config/aarch64/aarch64.h
9254 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9256 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9259 * expmed.c (emit_store_flag_force): Swap if const op0
9260 and change VOIDmode to mode of op0.
9262 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9264 PR rtl-optimization/83761
9265 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9266 than bytes to mode_for_size.
9268 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9271 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9272 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9275 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9278 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9279 when in layout mode.
9280 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9281 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9284 2018-01-10 Michael Collison <michael.collison@arm.com>
9286 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9287 * config/aarch64/aarch64-option-extension.def: Add
9288 AARCH64_OPT_EXTENSION of 'fp16fml'.
9289 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9290 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9291 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9292 * config/aarch64/constraints.md (Ui7): New constraint.
9293 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9294 (VFMLA_SEL_W): Ditto.
9297 (VFMLA16_LOW): New int iterator.
9298 (VFMLA16_HIGH): Ditto.
9299 (UNSPEC_FMLAL): New unspec.
9300 (UNSPEC_FMLSL): Ditto.
9301 (UNSPEC_FMLAL2): Ditto.
9302 (UNSPEC_FMLSL2): Ditto.
9303 (f16mac): New code attribute.
9304 * config/aarch64/aarch64-simd-builtins.def
9305 (aarch64_fmlal_lowv2sf): Ditto.
9306 (aarch64_fmlsl_lowv2sf): Ditto.
9307 (aarch64_fmlalq_lowv4sf): Ditto.
9308 (aarch64_fmlslq_lowv4sf): Ditto.
9309 (aarch64_fmlal_highv2sf): Ditto.
9310 (aarch64_fmlsl_highv2sf): Ditto.
9311 (aarch64_fmlalq_highv4sf): Ditto.
9312 (aarch64_fmlslq_highv4sf): Ditto.
9313 (aarch64_fmlal_lane_lowv2sf): Ditto.
9314 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9315 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9316 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9317 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9318 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9319 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9320 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9321 (aarch64_fmlal_lane_highv2sf): Ditto.
9322 (aarch64_fmlsl_lane_highv2sf): Ditto.
9323 (aarch64_fmlal_laneq_highv2sf): Ditto.
9324 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9325 (aarch64_fmlalq_lane_highv4sf): Ditto.
9326 (aarch64_fmlsl_lane_highv4sf): Ditto.
9327 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9328 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9329 * config/aarch64/aarch64-simd.md:
9330 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9331 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9332 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9333 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9334 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9335 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9336 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9337 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9338 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9339 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9340 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9341 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9342 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9343 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9344 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9345 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9346 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9347 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9348 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9349 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9350 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9351 (vfmlsl_low_u32): Ditto.
9352 (vfmlalq_low_u32): Ditto.
9353 (vfmlslq_low_u32): Ditto.
9354 (vfmlal_high_u32): Ditto.
9355 (vfmlsl_high_u32): Ditto.
9356 (vfmlalq_high_u32): Ditto.
9357 (vfmlslq_high_u32): Ditto.
9358 (vfmlal_lane_low_u32): Ditto.
9359 (vfmlsl_lane_low_u32): Ditto.
9360 (vfmlal_laneq_low_u32): Ditto.
9361 (vfmlsl_laneq_low_u32): Ditto.
9362 (vfmlalq_lane_low_u32): Ditto.
9363 (vfmlslq_lane_low_u32): Ditto.
9364 (vfmlalq_laneq_low_u32): Ditto.
9365 (vfmlslq_laneq_low_u32): Ditto.
9366 (vfmlal_lane_high_u32): Ditto.
9367 (vfmlsl_lane_high_u32): Ditto.
9368 (vfmlal_laneq_high_u32): Ditto.
9369 (vfmlsl_laneq_high_u32): Ditto.
9370 (vfmlalq_lane_high_u32): Ditto.
9371 (vfmlslq_lane_high_u32): Ditto.
9372 (vfmlalq_laneq_high_u32): Ditto.
9373 (vfmlslq_laneq_high_u32): Ditto.
9374 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9375 (AARCH64_FL_FOR_ARCH8_4): New.
9376 (AARCH64_ISA_F16FML): New ISA flag.
9377 (TARGET_F16FML): New feature flag for fp16fml.
9378 (doc/invoke.texi): Document new fp16fml option.
9380 2018-01-10 Michael Collison <michael.collison@arm.com>
9382 * config/aarch64/aarch64-builtins.c:
9383 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9384 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9385 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9386 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9387 (AARCH64_ISA_SHA3): New ISA flag.
9388 (TARGET_SHA3): New feature flag for sha3.
9389 * config/aarch64/iterators.md (sha512_op): New int attribute.
9390 (CRYPTO_SHA512): New int iterator.
9391 (UNSPEC_SHA512H): New unspec.
9392 (UNSPEC_SHA512H2): Ditto.
9393 (UNSPEC_SHA512SU0): Ditto.
9394 (UNSPEC_SHA512SU1): Ditto.
9395 * config/aarch64/aarch64-simd-builtins.def
9396 (aarch64_crypto_sha512hqv2di): New builtin.
9397 (aarch64_crypto_sha512h2qv2di): Ditto.
9398 (aarch64_crypto_sha512su0qv2di): Ditto.
9399 (aarch64_crypto_sha512su1qv2di): Ditto.
9400 (aarch64_eor3qv8hi): Ditto.
9401 (aarch64_rax1qv2di): Ditto.
9402 (aarch64_xarqv2di): Ditto.
9403 (aarch64_bcaxqv8hi): Ditto.
9404 * config/aarch64/aarch64-simd.md:
9405 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9406 (aarch64_crypto_sha512su0qv2di): Ditto.
9407 (aarch64_crypto_sha512su1qv2di): Ditto.
9408 (aarch64_eor3qv8hi): Ditto.
9409 (aarch64_rax1qv2di): Ditto.
9410 (aarch64_xarqv2di): Ditto.
9411 (aarch64_bcaxqv8hi): Ditto.
9412 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9413 (vsha512h2q_u64): Ditto.
9414 (vsha512su0q_u64): Ditto.
9415 (vsha512su1q_u64): Ditto.
9416 (veor3q_u16): Ditto.
9417 (vrax1q_u64): Ditto.
9419 (vbcaxq_u16): Ditto.
9420 * config/arm/types.md (crypto_sha512): New type attribute.
9421 (crypto_sha3): Ditto.
9422 (doc/invoke.texi): Document new sha3 option.
9424 2018-01-10 Michael Collison <michael.collison@arm.com>
9426 * config/aarch64/aarch64-builtins.c:
9427 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9428 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9429 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9430 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9431 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9432 (AARCH64_ISA_SM4): New ISA flag.
9433 (TARGET_SM4): New feature flag for sm4.
9434 * config/aarch64/aarch64-simd-builtins.def
9435 (aarch64_sm3ss1qv4si): Ditto.
9436 (aarch64_sm3tt1aq4si): Ditto.
9437 (aarch64_sm3tt1bq4si): Ditto.
9438 (aarch64_sm3tt2aq4si): Ditto.
9439 (aarch64_sm3tt2bq4si): Ditto.
9440 (aarch64_sm3partw1qv4si): Ditto.
9441 (aarch64_sm3partw2qv4si): Ditto.
9442 (aarch64_sm4eqv4si): Ditto.
9443 (aarch64_sm4ekeyqv4si): Ditto.
9444 * config/aarch64/aarch64-simd.md:
9445 (aarch64_sm3ss1qv4si): Ditto.
9446 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9447 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9448 (aarch64_sm4eqv4si): Ditto.
9449 (aarch64_sm4ekeyqv4si): Ditto.
9450 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9451 (sm3part_op): Ditto.
9452 (CRYPTO_SM3TT): Ditto.
9453 (CRYPTO_SM3PART): Ditto.
9454 (UNSPEC_SM3SS1): New unspec.
9455 (UNSPEC_SM3TT1A): Ditto.
9456 (UNSPEC_SM3TT1B): Ditto.
9457 (UNSPEC_SM3TT2A): Ditto.
9458 (UNSPEC_SM3TT2B): Ditto.
9459 (UNSPEC_SM3PARTW1): Ditto.
9460 (UNSPEC_SM3PARTW2): Ditto.
9461 (UNSPEC_SM4E): Ditto.
9462 (UNSPEC_SM4EKEY): Ditto.
9463 * config/aarch64/constraints.md (Ui2): New constraint.
9464 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9465 * config/arm/types.md (crypto_sm3): New type attribute.
9466 (crypto_sm4): Ditto.
9467 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9468 (vsm3tt1aq_u32): Ditto.
9469 (vsm3tt1bq_u32): Ditto.
9470 (vsm3tt2aq_u32): Ditto.
9471 (vsm3tt2bq_u32): Ditto.
9472 (vsm3partw1q_u32): Ditto.
9473 (vsm3partw2q_u32): Ditto.
9474 (vsm4eq_u32): Ditto.
9475 (vsm4ekeyq_u32): Ditto.
9476 (doc/invoke.texi): Document new sm4 option.
9478 2018-01-10 Michael Collison <michael.collison@arm.com>
9480 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9481 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9482 (AARCH64_FL_FOR_ARCH8_4): New.
9483 (AARCH64_FL_V8_4): New flag.
9484 (doc/invoke.texi): Document new armv8.4-a option.
9486 2018-01-10 Michael Collison <michael.collison@arm.com>
9488 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9489 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9490 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9491 * config/aarch64/aarch64-option-extension.def: Add
9492 AARCH64_OPT_EXTENSION of 'sha2'.
9493 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9494 (crypto): Disable sha2 and aes if crypto disabled.
9495 (crypto): Enable aes and sha2 if enabled.
9496 (simd): Disable sha2 and aes if simd disabled.
9497 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9499 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9500 (TARGET_SHA2): New feature flag for sha2.
9501 (TARGET_AES): New feature flag for aes.
9502 * config/aarch64/aarch64-simd.md:
9503 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9504 conditional on TARGET_AES.
9505 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9506 (aarch64_crypto_sha1hsi): Make pattern conditional
9508 (aarch64_crypto_sha1hv4si): Ditto.
9509 (aarch64_be_crypto_sha1hv4si): Ditto.
9510 (aarch64_crypto_sha1su1v4si): Ditto.
9511 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9512 (aarch64_crypto_sha1su0v4si): Ditto.
9513 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9514 (aarch64_crypto_sha256su0v4si): Ditto.
9515 (aarch64_crypto_sha256su1v4si): Ditto.
9516 (doc/invoke.texi): Document new aes and sha2 options.
9518 2018-01-10 Martin Sebor <msebor@redhat.com>
9520 PR tree-optimization/83781
9521 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9524 2018-01-11 Martin Sebor <msebor@gmail.com>
9525 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9527 PR tree-optimization/83501
9528 PR tree-optimization/81703
9530 * tree-ssa-strlen.c (get_string_cst): Rename...
9531 (get_string_len): ...to this. Handle global constants.
9532 (handle_char_store): Adjust.
9534 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9535 Jim Wilson <jimw@sifive.com>
9537 * config/riscv/riscv-protos.h (riscv_output_return): New.
9538 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9539 (riscv_attribute_table, riscv_output_return),
9540 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9541 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9542 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9544 (riscv_expand_prologue): Add early return for naked function.
9545 (riscv_expand_epilogue): Likewise.
9546 (riscv_function_ok_for_sibcall): Return false for naked function.
9547 (riscv_set_current_function): New.
9548 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9549 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9550 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9551 * doc/extend.texi (RISC-V Function Attributes): New.
9553 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9555 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9556 check for 128-bit long double before checking TCmode.
9557 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9558 128-bit long doubles before checking TFmode or TCmode.
9559 (FLOAT128_IBM_P): Likewise.
9561 2018-01-10 Martin Sebor <msebor@redhat.com>
9563 PR tree-optimization/83671
9564 * builtins.c (c_strlen): Unconditionally return zero for the empty
9566 Use -Warray-bounds for warnings.
9567 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9568 for non-constant array indices with COMPONENT_REF, arrays of
9569 arrays, and pointers to arrays.
9570 (gimple_fold_builtin_strlen): Determine and set length range for
9571 non-constant character arrays.
9573 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9576 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9579 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9581 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9583 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9586 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9587 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9588 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9589 indexed_or_indirect_operand predicate.
9590 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9591 (*vsx_le_perm_load_v8hi): Likewise.
9592 (*vsx_le_perm_load_v16qi): Likewise.
9593 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9594 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9595 (*vsx_le_perm_store_v8hi): Likewise.
9596 (*vsx_le_perm_store_v16qi): Likewise.
9597 (eight unnamed splitters): Likewise.
9599 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9601 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9602 * config/rs6000/emmintrin.h: Likewise.
9603 * config/rs6000/mmintrin.h: Likewise.
9604 * config/rs6000/xmmintrin.h: Likewise.
9606 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9609 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9611 * tree.c (tree_nop_conversion): Return true for location wrapper
9613 (maybe_wrap_with_location): New function.
9614 (selftest::check_strip_nops): New function.
9615 (selftest::test_location_wrappers): New function.
9616 (selftest::tree_c_tests): Call it.
9617 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9618 (maybe_wrap_with_location): New decl.
9619 (EXPR_LOCATION_WRAPPER_P): New macro.
9620 (location_wrapper_p): New inline function.
9621 (tree_strip_any_location_wrapper): New inline function.
9623 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9626 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9627 stack_realign_offset for the largest alignment of stack slot
9629 (ix86_find_max_used_stack_alignment): New function.
9630 (ix86_finalize_stack_frame_flags): Use it. Set
9631 max_used_stack_alignment if we don't realign stack.
9632 * config/i386/i386.h (machine_function): Add
9633 max_used_stack_alignment.
9635 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9637 * config/arm/arm.opt (-mbranch-cost): New option.
9638 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9641 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9644 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9645 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9647 2018-01-10 Richard Biener <rguenther@suse.de>
9650 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9651 early out so it also covers the case where we have a non-NULL
9654 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9656 PR tree-optimization/83753
9657 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9658 for non-strided grouped accesses if the number of elements is 1.
9660 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9663 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9664 * i386.h (TARGET_USE_GATHER): Define.
9665 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9667 2018-01-10 Martin Liska <mliska@suse.cz>
9670 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9671 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9673 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9674 CLEANUP_NO_PARTITIONING is not set.
9676 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9678 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9679 for vectors, as a partial revert of r254296.
9680 * rtl.h (const_vec_p): Delete.
9681 (const_vec_duplicate_p): Don't test for vector CONSTs.
9682 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9683 * expmed.c (make_tree): Likewise.
9686 * common.md (E, F): Use CONSTANT_P instead of checking for
9688 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9689 checking for CONST_VECTOR.
9691 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9694 * predict.c (force_edge_cold): Handle in more sane way edges
9697 2018-01-09 Carl Love <cel@us.ibm.com>
9699 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9701 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9702 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9703 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9704 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9705 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9706 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9707 * config/rs6000/rs6000-protos.h: Add extern defition for
9708 rs6000_generate_float2_double_code.
9709 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9711 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9712 (float2_v2df): Add define_expand.
9714 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9717 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9718 op_mode in the force_to_mode call.
9720 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9722 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9723 instead of checking each element individually.
9724 (aarch64_evpc_uzp): Likewise.
9725 (aarch64_evpc_zip): Likewise.
9726 (aarch64_evpc_ext): Likewise.
9727 (aarch64_evpc_rev): Likewise.
9728 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9729 instead of checking each element individually. Return true without
9731 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9732 whether all selected elements come from the same input, instead of
9733 checking each element individually. Remove calls to gen_rtx_REG,
9734 start_sequence and end_sequence and instead assert that no rtl is
9737 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9739 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9740 order of HIGH and CONST checks.
9742 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9744 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9745 if the destination isn't an SSA_NAME.
9747 2018-01-09 Richard Biener <rguenther@suse.de>
9749 PR tree-optimization/83668
9750 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9752 (canonicalize_loop_form): ... here, renamed from ...
9753 (canonicalize_loop_closed_ssa_form): ... this and amended to
9754 swap successor edges for loop exit blocks to make us use
9755 the RPO order we need for initial schedule generation.
9757 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9759 PR tree-optimization/64811
9760 * match.pd: When optimizing comparisons with Inf, avoid
9761 introducing or losing exceptions from comparisons with NaN.
9763 2018-01-09 Martin Liska <mliska@suse.cz>
9766 * asan.c (shadow_mem_size): Add gcc_assert.
9768 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9770 Don't save registers in main().
9773 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9774 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9775 * config/avr/avr.c (avr_set_current_function): Don't error if
9776 naked, OS_task or OS_main are specified at the same time.
9777 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9779 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9781 * common/config/avr/avr-common.c (avr_option_optimization_table):
9782 Switch on -mmain-is-OS_task for optimizing compilations.
9784 2018-01-09 Richard Biener <rguenther@suse.de>
9786 PR tree-optimization/83572
9787 * graphite.c: Include cfganal.h.
9788 (graphite_transform_loops): Connect infinite loops to exit
9789 and remove fake edges at the end.
9791 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9793 * ipa-inline.c (edge_badness): Revert accidental checkin.
9795 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9798 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9799 symbols; not inline clones.
9801 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9804 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9805 hard registers. Formatting fixes.
9807 PR preprocessor/83722
9808 * gcc.c (try_generate_repro): Pass
9809 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9810 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9813 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9814 Kito Cheng <kito.cheng@gmail.com>
9816 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9817 (riscv_leaf_function_p): Delete.
9818 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9820 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9822 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9824 (do_ifelse): New function.
9825 (do_isel): New function.
9826 (do_sub3): New function.
9827 (do_add3): New function.
9828 (do_load_mask_compare): New function.
9829 (do_overlap_load_compare): New function.
9830 (expand_compare_loop): New function.
9831 (expand_block_compare): Call expand_compare_loop() when appropriate.
9832 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9834 (-mblock-compare-inline-loop-limit): New option.
9836 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9839 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9840 Reverse order of second and third operands in first alternative.
9841 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9842 of first and second elements in UNSPEC_VPERMR vector.
9843 (altivec_expand_vec_perm_le): Likewise.
9845 2018-01-08 Jeff Law <law@redhat.com>
9847 PR rtl-optimizatin/81308
9848 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9849 (process_switch): If group_case_labels makes a change, then set
9851 (pass_convert_switch::execute): If a switch is converted, then
9852 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9854 PR rtl-optimization/81308
9855 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9858 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9860 PR target/83663 - Revert r255946
9861 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9862 generation for cases where splatting a value is not useful.
9863 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9864 across a vec_duplicate and a paradoxical subreg forming a vector
9865 mode to a vec_concat.
9867 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9869 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9870 -march=armv8.3-a variants.
9871 * config/arm/t-multilib: Likewise.
9872 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9874 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9876 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9878 (cceq_ior_compare_complement): Give it a name so I can use it, and
9879 change boolean_or_operator predicate to boolean_operator so it can
9880 be used to generate a crand.
9881 (eqne): New code iterator.
9882 (bd/bd_neg): New code_attrs.
9883 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9884 a single define_insn.
9885 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9886 decrement (bdnzt/bdnzf/bdzt/bdzf).
9887 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9888 with the new names of the branch decrement patterns, and added the
9889 names of the branch decrement conditional patterns.
9891 2018-01-08 Richard Biener <rguenther@suse.de>
9893 PR tree-optimization/83563
9894 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9897 2018-01-08 Richard Biener <rguenther@suse.de>
9900 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9902 2018-01-08 Richard Biener <rguenther@suse.de>
9904 PR tree-optimization/83685
9905 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9906 references to abnormals.
9908 2018-01-08 Richard Biener <rguenther@suse.de>
9911 * dwarf2out.c (output_indirect_strings): Handle empty
9912 skeleton_debug_str_hash.
9913 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9915 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9917 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9918 (emit_store_direct): Likewise.
9919 (arc_trampoline_adjust_address): Likewise.
9920 (arc_asm_trampoline_template): New function.
9921 (arc_initialize_trampoline): Use asm_trampoline_template.
9922 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9923 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9924 * config/arc/arc.md (flush_icache): Delete pattern.
9926 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9928 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9929 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9932 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9935 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9936 by not USED_FOR_TARGET.
9937 (make_pass_resolve_sw_modes): Likewise.
9939 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9941 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9944 2018-01-08 Richard Biener <rguenther@suse.de>
9947 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9949 2018-01-08 Richard Biener <rguenther@suse.de>
9952 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9954 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9957 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9958 basic blocks with a small number of successors.
9959 (convert_control_dep_chain_into_preds): Improve handling of
9961 (dump_predicates): Split apart into...
9962 (dump_pred_chain): ...here...
9963 (dump_pred_info): ...and here.
9964 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9965 (can_chain_union_be_invalidated_p): Improve check for invalidation
9967 (uninit_uses_cannot_happen): Avoid unnecessary if
9968 convert_control_dep_chain_into_preds yielded nothing.
9970 2018-01-06 Martin Sebor <msebor@redhat.com>
9972 PR tree-optimization/83640
9973 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9974 subtracting negative offset from size.
9975 (builtin_access::overlap): Adjust offset bounds of the access to fall
9976 within the size of the object if possible.
9978 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9980 PR rtl-optimization/83699
9981 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9982 extract_bit_field_as_subreg to cases in which the extracted
9983 value is also a vector.
9985 * lra-constraints.c (process_alt_operands): Test for the equivalence
9986 substitutions when detecting a possible reload cycle.
9988 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9991 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9992 by default if flag_selective_schedling{,2}. Formatting fixes.
9994 PR rtl-optimization/83682
9995 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9996 if it has non-VECTOR_MODE element mode.
9997 (vec_duplicate_p): Likewise.
10000 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10001 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10003 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10006 * config/i386/i386-builtin.def
10007 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10008 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10009 Require also OPTION_MASK_ISA_AVX512F in addition to
10010 OPTION_MASK_ISA_GFNI.
10011 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10012 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10013 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10014 to OPTION_MASK_ISA_GFNI.
10015 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10016 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10017 OPTION_MASK_ISA_AVX512BW.
10018 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10019 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10020 addition to OPTION_MASK_ISA_GFNI.
10021 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10022 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10023 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10024 to OPTION_MASK_ISA_GFNI.
10025 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10026 a requirement for all ISAs rather than any of them with a few
10028 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10030 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10031 bitmasks to be enabled with 3 exceptions, instead of requiring any
10032 enabled ISA with lots of exceptions.
10033 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10034 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10035 Change avx512bw in isa attribute to avx512f.
10036 * config/i386/sgxintrin.h: Add license boilerplate.
10037 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10038 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10039 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10040 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10042 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10043 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10044 temporarily sse2 rather than sse if not enabled already.
10047 * config/i386/sse.md (VI248_VLBW): Rename to ...
10048 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10049 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10050 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10051 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10052 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10053 mode iterator instead of VI248_VLBW.
10055 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10057 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10058 (record_modified): Skip clobbers; add debug output.
10059 (param_change_prob): Use sreal frequencies.
10061 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10063 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10064 punt for user-aligned variables.
10066 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10068 * tree-chrec.c (chrec_contains_symbols): Return true for
10071 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10074 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10075 of (x|y) == x for BICS pattern.
10077 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10079 PR tree-optimization/83605
10080 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10081 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10084 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10086 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10087 * config/epiphany/rtems.h: New file.
10089 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10090 Uros Bizjak <ubizjak@gmail.com>
10093 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10094 QIreg_operand instead of register_operand predicate.
10095 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10096 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10097 comments instead of -fmitigate[-_]rop.
10099 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10102 * cgraphunit.c (symbol_table::compile): Switch to text_section
10103 before calling assembly_start debug hook.
10104 * run-rtl-passes.c (run_rtl_passes): Likewise.
10107 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10109 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10110 range_int_cst_p rather than !symbolic_range_p before calling
10111 extract_range_from_multiplicative_op_1.
10113 2018-01-04 Jeff Law <law@redhat.com>
10115 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10116 redundant test in assertion.
10118 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10120 * doc/rtl.texi: Document machine_mode wrapper classes.
10122 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10124 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10125 using tree_to_uhwi.
10127 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10129 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10130 the VEC_PERM_EXPR fold to fail.
10132 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10135 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10136 to switched_sections.
10138 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10141 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10142 test for d.testing.
10144 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10147 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10148 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10150 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10153 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10154 is BLKmode and bitpos not zero or mode change is needed.
10156 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10159 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10162 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10165 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10166 instead of MULT rtx. Update all corresponding splitters.
10167 (*saddl_se): Ditto.
10168 (*ssub<modesuffix>): Ditto.
10169 (*ssubl_se): Ditto.
10170 (*cmp_sadd_di): Update split patterns.
10171 (*cmp_sadd_si): Ditto.
10172 (*cmp_sadd_sidi): Ditto.
10173 (*cmp_ssub_di): Ditto.
10174 (*cmp_ssub_si): Ditto.
10175 (*cmp_ssub_sidi): Ditto.
10176 * config/alpha/predicates.md (const23_operand): New predicate.
10177 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10178 Look for ASHIFT, not MULT inner operand.
10179 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10181 2018-01-04 Martin Liska <mliska@suse.cz>
10183 PR gcov-profile/83669
10184 * gcov.c (output_intermediate_file): Add version to intermediate
10186 * doc/gcov.texi: Document new field 'version' in intermediate
10187 file format. Fix location of '-k' option of gcov command.
10189 2018-01-04 Martin Liska <mliska@suse.cz>
10192 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10194 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10196 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10198 2018-01-03 Martin Sebor <msebor@redhat.com>
10200 PR tree-optimization/83655
10201 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10202 checking calls with invalid arguments.
10204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10206 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10207 (vectorizable_mask_load_store): Delete.
10208 (vectorizable_call): Return false for masked loads and stores.
10209 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10210 instead of gimple_assign_rhs1.
10211 (vectorizable_load): Handle IFN_MASK_LOAD.
10212 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10216 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10218 (vectorizable_mask_load_store): ...here.
10219 (vectorizable_load): ...and here.
10221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10223 * tree-vect-stmts.c (vect_build_all_ones_mask)
10224 (vect_build_zero_merge_argument): New functions, split out from...
10225 (vectorizable_load): ...here.
10227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10229 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10231 (vectorizable_mask_load_store): ...here.
10232 (vectorizable_store): ...and here.
10234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10236 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10238 (vectorizable_mask_load_store): ...here.
10240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10242 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10243 (vect_model_store_cost): Take a vec_load_store_type instead of a
10245 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10246 (vect_model_store_cost): Take a vec_load_store_type instead of a
10248 (vectorizable_mask_load_store): Update accordingly.
10249 (vectorizable_store): Likewise.
10250 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10254 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10255 IFN_MASK_LOAD calls here rather than...
10256 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10258 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10259 Alan Hayward <alan.hayward@arm.com>
10260 David Sherwood <david.sherwood@arm.com>
10262 * expmed.c (extract_bit_field_1): For vector extracts,
10263 fall back to extract_bit_field_as_subreg if vec_extract
10266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10267 Alan Hayward <alan.hayward@arm.com>
10268 David Sherwood <david.sherwood@arm.com>
10270 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10271 they are variable or constant sized.
10272 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10273 slots for constant-sized data.
10275 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10276 Alan Hayward <alan.hayward@arm.com>
10277 David Sherwood <david.sherwood@arm.com>
10279 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10280 handling COND_EXPRs with boolean comparisons, try to find a better
10281 basis for the mask type than the boolean itself.
10283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10285 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10286 is calculated and how it can be overridden.
10287 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10288 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10290 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10294 Alan Hayward <alan.hayward@arm.com>
10295 David Sherwood <david.sherwood@arm.com>
10297 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10298 Remove the mode argument.
10299 (aarch64_simd_valid_immediate): Remove the mode and inverse
10301 * config/aarch64/iterators.md (bitsize): New iterator.
10302 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10303 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10304 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10305 aarch64_simd_valid_immediate.
10306 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10307 (aarch64_reg_or_bic_imm): Likewise.
10308 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10309 with an insn_type enum and msl with a modifier_type enum.
10310 Replace element_width with a scalar_mode. Change the shift
10311 to unsigned int. Add constructors for scalar_float_mode and
10312 scalar_int_mode elements.
10313 (aarch64_vect_float_const_representable_p): Delete.
10314 (aarch64_can_const_movi_rtx_p)
10315 (aarch64_simd_scalar_immediate_valid_for_move)
10316 (aarch64_simd_make_constant): Update call to
10317 aarch64_simd_valid_immediate.
10318 (aarch64_advsimd_valid_immediate_hs): New function.
10319 (aarch64_advsimd_valid_immediate): Likewise.
10320 (aarch64_simd_valid_immediate): Remove mode and inverse
10321 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10322 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10323 and aarch64_float_const_representable_p on the result.
10324 (aarch64_output_simd_mov_immediate): Remove mode argument.
10325 Update call to aarch64_simd_valid_immediate and use of
10326 simd_immediate_info.
10327 (aarch64_output_scalar_simd_mov_immediate): Update call
10330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10331 Alan Hayward <alan.hayward@arm.com>
10332 David Sherwood <david.sherwood@arm.com>
10334 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10335 (mode_nunits): Likewise CONST_MODE_NUNITS.
10336 * machmode.def (ADJUST_NUNITS): Document.
10337 * genmodes.c (mode_data::need_nunits_adj): New field.
10338 (blank_mode): Update accordingly.
10339 (adj_nunits): New variable.
10340 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10342 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10343 listed in adj_nunits.
10344 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10345 listed in adj_nunits. Don't emit case statements for such modes.
10346 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10347 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10348 nothing if adj_nunits is nonnull.
10349 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10350 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10351 (emit_mode_fbit): Update use of print_maybe_const_decl.
10352 (emit_move_size): Likewise. Treat the array as non-const
10354 (emit_mode_adjustments): Handle adj_nunits.
10356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10358 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10359 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10360 (VECTOR_MODES): Use it.
10361 (make_vector_modes): Take the prefix as an argument.
10363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10364 Alan Hayward <alan.hayward@arm.com>
10365 David Sherwood <david.sherwood@arm.com>
10367 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10368 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10369 for MODE_VECTOR_BOOL.
10370 * machmode.def (VECTOR_BOOL_MODE): Document.
10371 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10372 (make_vector_bool_mode): New function.
10373 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10375 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10376 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10378 * stor-layout.c (int_mode_for_mode): Likewise.
10379 * tree.c (build_vector_type_for_mode): Likewise.
10380 * varasm.c (output_constant_pool_2): Likewise.
10381 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10382 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10383 for MODE_VECTOR_BOOL.
10384 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10385 of mode class checks.
10386 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10387 instead of a list of mode class checks.
10388 (expand_vector_scalar_condition): Likewise.
10389 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10392 Alan Hayward <alan.hayward@arm.com>
10393 David Sherwood <david.sherwood@arm.com>
10395 * machmode.h (mode_size): Change from unsigned short to
10397 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10398 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10399 or if measurement_type is not polynomial.
10400 (fixed_size_mode::includes_p): Check for constant-sized modes.
10401 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10402 return a poly_uint16 rather than an unsigned short.
10403 (emit_mode_size): Change the type of mode_size from unsigned short
10404 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10405 (emit_mode_adjustments): Cope with polynomial vector sizes.
10406 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10408 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10410 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10411 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10412 * caller-save.c (setup_save_areas): Likewise.
10413 (replace_reg_with_saved_mem): Likewise.
10414 * calls.c (emit_library_call_value_1): Likewise.
10415 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10416 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10417 (gen_lowpart_for_combine): Likewise.
10418 * convert.c (convert_to_integer_1): Likewise.
10419 * cse.c (equiv_constant, cse_insn): Likewise.
10420 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10421 (cselib_subst_to_values): Likewise.
10422 * dce.c (word_dce_process_block): Likewise.
10423 * df-problems.c (df_word_lr_mark_ref): Likewise.
10424 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10425 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10426 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10427 (rtl_for_decl_location): Likewise.
10428 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10429 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10430 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10431 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10432 (expand_expr_real_1): Likewise.
10433 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10434 (pad_below): Likewise.
10435 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10436 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10437 * ira.c (get_subreg_tracking_sizes): Likewise.
10438 * ira-build.c (ira_create_allocno_objects): Likewise.
10439 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10440 (ira_sort_regnos_for_alter_reg): Likewise.
10441 * ira-costs.c (record_operand_costs): Likewise.
10442 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10443 (resolve_simple_move): Likewise.
10444 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10445 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10446 (lra_constraints): Likewise.
10447 (CONST_POOL_OK_P): Reject variable-sized modes.
10448 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10449 (add_pseudo_to_slot, lra_spill): Likewise.
10450 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10451 * optabs-query.c (get_best_extraction_insn): Likewise.
10452 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10453 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10454 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10455 * recog.c (offsettable_address_addr_space_p): Likewise.
10456 * regcprop.c (maybe_mode_change): Likewise.
10457 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10458 * regrename.c (build_def_use): Likewise.
10459 * regstat.c (dump_reg_info): Likewise.
10460 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10461 (find_reloads, find_reloads_subreg_address): Likewise.
10462 * reload1.c (eliminate_regs_1): Likewise.
10463 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10464 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10465 (simplify_binary_operation_1, simplify_subreg): Likewise.
10466 * targhooks.c (default_function_arg_padding): Likewise.
10467 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10468 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10469 (verify_gimple_assign_ternary): Likewise.
10470 * tree-inline.c (estimate_move_cost): Likewise.
10471 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10472 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10473 (get_address_cost_ainc): Likewise.
10474 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10475 (vect_supportable_dr_alignment): Likewise.
10476 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10477 (vectorizable_reduction): Likewise.
10478 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10479 (vectorizable_operation, vectorizable_load): Likewise.
10480 * tree.c (build_same_sized_truth_vector_type): Likewise.
10481 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10482 * var-tracking.c (emit_note_insn_var_location): Likewise.
10483 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10484 (ADDR_VEC_ALIGN): Likewise.
10486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10487 Alan Hayward <alan.hayward@arm.com>
10488 David Sherwood <david.sherwood@arm.com>
10490 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10492 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10493 or if measurement_type is polynomial.
10494 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10495 * combine.c (make_extraction): Likewise.
10496 * dse.c (find_shift_sequence): Likewise.
10497 * dwarf2out.c (mem_loc_descriptor): Likewise.
10498 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10499 (extract_bit_field, extract_low_bits): Likewise.
10500 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10501 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10502 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10503 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10504 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10505 * reload.c (find_reloads): Likewise.
10506 * reload1.c (alter_reg): Likewise.
10507 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10508 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10509 * tree-if-conv.c (predicate_mem_writes): Likewise.
10510 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10511 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10512 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10513 * valtrack.c (dead_debug_insert_temp): Likewise.
10514 * varasm.c (mergeable_constant_section): Likewise.
10515 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10518 Alan Hayward <alan.hayward@arm.com>
10519 David Sherwood <david.sherwood@arm.com>
10521 * expr.c (expand_assignment): Cope with polynomial mode sizes
10522 when assigning to a CONCAT.
10524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10525 Alan Hayward <alan.hayward@arm.com>
10526 David Sherwood <david.sherwood@arm.com>
10528 * machmode.h (mode_precision): Change from unsigned short to
10530 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10532 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10533 or if measurement_type is not polynomial.
10534 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10535 in which the mode is already known to be a scalar_int_mode.
10536 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10537 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10539 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10540 for GET_MODE_PRECISION.
10541 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10542 for GET_MODE_PRECISION.
10543 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10545 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10546 (expand_field_assignment, make_extraction): Likewise.
10547 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10548 (get_last_value): Likewise.
10549 * convert.c (convert_to_integer_1): Likewise.
10550 * cse.c (cse_insn): Likewise.
10551 * expr.c (expand_expr_real_1): Likewise.
10552 * lra-constraints.c (simplify_operand_subreg): Likewise.
10553 * optabs-query.c (can_atomic_load_p): Likewise.
10554 * optabs.c (expand_atomic_load): Likewise.
10555 (expand_atomic_store): Likewise.
10556 * ree.c (combine_reaching_defs): Likewise.
10557 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10558 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10559 * tree.h (type_has_mode_precision_p): Likewise.
10560 * ubsan.c (instrument_si_overflow): Likewise.
10562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10563 Alan Hayward <alan.hayward@arm.com>
10564 David Sherwood <david.sherwood@arm.com>
10566 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10567 polynomial numbers of units.
10568 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10569 (valid_vector_subparts_p): New function.
10570 (build_vector_type): Remove temporary shim and take the number
10571 of units as a poly_uint64 rather than an int.
10572 (build_opaque_vector_type): Take the number of units as a
10573 poly_uint64 rather than an int.
10574 * tree.c (build_vector_from_ctor): Handle polynomial
10575 TYPE_VECTOR_SUBPARTS.
10576 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10577 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10578 (build_vector_from_val): If the number of units is variable,
10579 use build_vec_duplicate_cst for constant operands and
10580 VEC_DUPLICATE_EXPR otherwise.
10581 (make_vector_type): Remove temporary is_constant ().
10582 (build_vector_type, build_opaque_vector_type): Take the number of
10583 units as a poly_uint64 rather than an int.
10584 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10586 * cfgexpand.c (expand_debug_expr): Likewise.
10587 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10588 (store_constructor, expand_expr_real_1): Likewise.
10589 (const_scalar_mask_from_tree): Likewise.
10590 * fold-const-call.c (fold_const_reduction): Likewise.
10591 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10592 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10593 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10594 (fold_relational_const): Likewise.
10595 (native_interpret_vector): Likewise. Change the size from an
10596 int to an unsigned int.
10597 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10598 TYPE_VECTOR_SUBPARTS.
10599 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10600 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10601 duplicating a non-constant operand into a variable-length vector.
10602 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10603 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10604 * ipa-icf.c (sem_variable::equals): Likewise.
10605 * match.pd: Likewise.
10606 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10607 * print-tree.c (print_node): Likewise.
10608 * stor-layout.c (layout_type): Likewise.
10609 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10610 * tree-cfg.c (verify_gimple_comparison): Likewise.
10611 (verify_gimple_assign_binary): Likewise.
10612 (verify_gimple_assign_ternary): Likewise.
10613 (verify_gimple_assign_single): Likewise.
10614 * tree-pretty-print.c (dump_generic_node): Likewise.
10615 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10616 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10617 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10618 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10619 (vect_shift_permute_load_chain): Likewise.
10620 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10621 (expand_vector_condition, optimize_vector_constructor): Likewise.
10622 (lower_vec_perm, get_compute_type): Likewise.
10623 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10624 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10625 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10626 (vect_recog_mask_conversion_pattern): Likewise.
10627 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10628 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10629 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10630 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10631 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10632 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10633 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10634 (supportable_widening_operation): Likewise.
10635 (supportable_narrowing_operation): Likewise.
10636 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10638 * varasm.c (output_constant): Likewise.
10640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10641 Alan Hayward <alan.hayward@arm.com>
10642 David Sherwood <david.sherwood@arm.com>
10644 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10645 so that both the length == 3 and length != 3 cases set up their
10646 own permute vectors. Add comments explaining why we know the
10647 number of elements is constant.
10648 (vect_permute_load_chain): Likewise.
10650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10651 Alan Hayward <alan.hayward@arm.com>
10652 David Sherwood <david.sherwood@arm.com>
10654 * machmode.h (mode_nunits): Change from unsigned char to
10656 (ONLY_FIXED_SIZE_MODES): New macro.
10657 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10658 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10659 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10661 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10662 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10663 or if measurement_type is not polynomial.
10664 * genmodes.c (ZERO_COEFFS): New macro.
10665 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10667 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10668 Use ZERO_COEFFS when emitting initializers.
10669 * data-streamer.h (bp_pack_poly_value): New function.
10670 (bp_unpack_poly_value): Likewise.
10671 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10672 for GET_MODE_NUNITS.
10673 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10674 for GET_MODE_NUNITS.
10675 * tree.c (make_vector_type): Remove temporary shim and make
10676 the real function take the number of units as a poly_uint64
10677 rather than an int.
10678 (build_vector_type_for_mode): Handle polynomial nunits.
10679 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10680 * emit-rtl.c (const_vec_series_p_1): Likewise.
10681 (gen_rtx_CONST_VECTOR): Likewise.
10682 * fold-const.c (test_vec_duplicate_folding): Likewise.
10683 * genrecog.c (validate_pattern): Likewise.
10684 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10685 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10686 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10687 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10688 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10689 * rtlanal.c (subreg_get_info): Likewise.
10690 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10691 (vect_grouped_load_supported): Likewise.
10692 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10693 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10694 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10695 (simplify_const_unary_operation, simplify_binary_operation_1)
10696 (simplify_const_binary_operation, simplify_ternary_operation)
10697 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10698 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10699 instead of CONST_VECTOR_NUNITS.
10700 * varasm.c (output_constant_pool_2): Likewise.
10701 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10702 explicit-encoded elements in the XVEC for variable-length vectors.
10704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10706 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10709 Alan Hayward <alan.hayward@arm.com>
10710 David Sherwood <david.sherwood@arm.com>
10712 * coretypes.h (fixed_size_mode): Declare.
10713 (fixed_size_mode_pod): New typedef.
10714 * builtins.h (target_builtins::x_apply_args_mode)
10715 (target_builtins::x_apply_result_mode): Change type to
10716 fixed_size_mode_pod.
10717 * builtins.c (apply_args_size, apply_result_size, result_vector)
10718 (expand_builtin_apply_args_1, expand_builtin_apply)
10719 (expand_builtin_return): Update accordingly.
10721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10723 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10724 * cselib.c (cselib_hash_rtx): Likewise.
10725 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10726 CONST_VECTOR encoding.
10728 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10729 Jeff Law <law@redhat.com>
10732 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10733 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10734 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10735 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10738 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10739 explicitly probe *sp in a noreturn function if there were any callee
10740 register saves or frame pointer is needed.
10742 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10745 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10746 BLKmode for ternary, binary or unary expressions.
10749 * var-tracking.c (delete_vta_debug_insn): New inline function.
10750 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10751 insns from get_insns () to NULL instead of each bb separately.
10752 Use delete_vta_debug_insn. No longer static.
10753 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10754 delete_vta_debug_insns callers.
10755 * rtl.h (delete_vta_debug_insns): Declare.
10756 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10757 instead of variable_tracking_main.
10759 2018-01-03 Martin Sebor <msebor@redhat.com>
10761 PR tree-optimization/83603
10762 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10763 arguments past the endof the argument list in functions declared
10764 without a prototype.
10765 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10766 Avoid checking when arguments are null.
10768 2018-01-03 Martin Sebor <msebor@redhat.com>
10771 * doc/extend.texi (attribute const): Fix a typo.
10772 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10773 issuing -Wsuggest-attribute for void functions.
10775 2018-01-03 Martin Sebor <msebor@redhat.com>
10777 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10778 offset_int::from instead of wide_int::to_shwi.
10779 (maybe_diag_overlap): Remove assertion.
10780 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10781 * gimple-ssa-sprintf.c (format_directive): Same.
10782 (parse_directive): Same.
10783 (sprintf_dom_walker::compute_format_length): Same.
10784 (try_substitute_return_value): Same.
10786 2018-01-03 Jeff Law <law@redhat.com>
10788 PR middle-end/83654
10789 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10790 non-constant residual for zero at runtime and avoid probing in
10791 that case. Reorganize code for trailing problem to mirror handling
10794 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10796 PR tree-optimization/83501
10797 * tree-ssa-strlen.c (get_string_cst): New.
10798 (handle_char_store): Call get_string_cst.
10800 2018-01-03 Martin Liska <mliska@suse.cz>
10802 PR tree-optimization/83593
10803 * tree-ssa-strlen.c: Include tree-cfg.h.
10804 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10805 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10806 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10808 (strlen_dom_walker::before_dom_children): Call
10809 gimple_purge_dead_eh_edges. Dump tranformation with details
10811 (strlen_dom_walker::before_dom_children): Update call by adding
10812 new argument cleanup_eh.
10813 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10815 2018-01-03 Martin Liska <mliska@suse.cz>
10818 * cif-code.def (VARIADIC_THUNK): New enum value.
10819 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10822 2018-01-03 Jan Beulich <jbeulich@suse.com>
10824 * sse.md (mov<mode>_internal): Tighten condition for when to use
10825 vmovdqu<ssescalarsize> for TI and OI modes.
10827 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10829 Update copyright years.
10831 2018-01-03 Martin Liska <mliska@suse.cz>
10834 * ipa-visibility.c (function_and_variable_visibility): Skip
10835 functions with noipa attribure.
10837 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10839 * gcc.c (process_command): Update copyright notice dates.
10840 * gcov-dump.c (print_version): Ditto.
10841 * gcov.c (print_version): Ditto.
10842 * gcov-tool.c (print_version): Ditto.
10843 * gengtype.c (create_file): Ditto.
10844 * doc/cpp.texi: Bump @copying's copyright year.
10845 * doc/cppinternals.texi: Ditto.
10846 * doc/gcc.texi: Ditto.
10847 * doc/gccint.texi: Ditto.
10848 * doc/gcov.texi: Ditto.
10849 * doc/install.texi: Ditto.
10850 * doc/invoke.texi: Ditto.
10852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10854 * vector-builder.h (vector_builder::m_full_nelts): Change from
10855 unsigned int to poly_uint64.
10856 (vector_builder::full_nelts): Update prototype accordingly.
10857 (vector_builder::new_vector): Likewise.
10858 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10859 (vector_builder::operator ==): Likewise.
10860 (vector_builder::finalize): Likewise.
10861 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10862 Take the number of elements as a poly_uint64 rather than an
10864 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10865 from unsigned int to poly_uint64.
10866 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10867 (vec_perm_indices::new_vector): Likewise.
10868 (vec_perm_indices::length): Likewise.
10869 (vec_perm_indices::nelts_per_input): Likewise.
10870 (vec_perm_indices::input_nelts): Likewise.
10871 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10872 number of elements per input as a poly_uint64 rather than an
10873 unsigned int. Use the original encoding for variable-length
10874 vectors, rather than clamping each individual element.
10875 For the second and subsequent elements in each pattern,
10876 clamp the step and base before clamping their sum.
10877 (vec_perm_indices::series_p): Handle polynomial element counts.
10878 (vec_perm_indices::all_in_range_p): Likewise.
10879 (vec_perm_indices_to_tree): Likewise.
10880 (vec_perm_indices_to_rtx): Likewise.
10881 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10882 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10883 (tree_vector_builder::new_binary_operation): Handle polynomial
10884 element counts. Return false if we need to know the number
10885 of elements at compile time.
10886 * fold-const.c (fold_vec_perm): Punt if the number of elements
10887 isn't known at compile time.
10889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10891 * vec-perm-indices.h (vec_perm_builder): Change element type
10892 from HOST_WIDE_INT to poly_int64.
10893 (vec_perm_indices::element_type): Update accordingly.
10894 (vec_perm_indices::clamp): Handle polynomial element_types.
10895 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10896 (vec_perm_indices::all_in_range_p): Likewise.
10897 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10899 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10900 polynomial vec_perm_indices element types.
10901 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10902 * fold-const.c (fold_vec_perm): Likewise.
10903 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10904 * tree-vect-generic.c (lower_vec_perm): Likewise.
10905 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10906 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10907 element type to HOST_WIDE_INT.
10909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10910 Alan Hayward <alan.hayward@arm.com>
10911 David Sherwood <david.sherwood@arm.com>
10913 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10914 rather than an int. Use plus_constant.
10915 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10916 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10919 Alan Hayward <alan.hayward@arm.com>
10920 David Sherwood <david.sherwood@arm.com>
10922 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10923 a HOST_WIDE_INT to a poly_int64.
10925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10926 Alan Hayward <alan.hayward@arm.com>
10927 David Sherwood <david.sherwood@arm.com>
10929 * calls.c (load_register_parameters): Cope with polynomial
10930 mode sizes. Require a constant size for BLKmode parameters
10931 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10932 forces a parameter to be padded at the lsb end in order to
10933 fill a complete number of words, require the parameter size
10934 to be ordered wrt UNITS_PER_WORD.
10936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10937 Alan Hayward <alan.hayward@arm.com>
10938 David Sherwood <david.sherwood@arm.com>
10940 * reload1.c (spill_stack_slot_width): Change element type
10941 from unsigned int to poly_uint64_pod.
10942 (alter_reg): Treat mode sizes as polynomial.
10944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10945 Alan Hayward <alan.hayward@arm.com>
10946 David Sherwood <david.sherwood@arm.com>
10948 * reload.c (complex_word_subreg_p): New function.
10949 (reload_inner_reg_of_subreg, push_reload): Use it.
10951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10952 Alan Hayward <alan.hayward@arm.com>
10953 David Sherwood <david.sherwood@arm.com>
10955 * lra-constraints.c (process_alt_operands): Reject matched
10956 operands whose sizes aren't ordered.
10957 (match_reload): Refer to this check here.
10959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10960 Alan Hayward <alan.hayward@arm.com>
10961 David Sherwood <david.sherwood@arm.com>
10963 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10964 that the mode size is in the set {1, 2, 4, 8, 16}.
10966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10967 Alan Hayward <alan.hayward@arm.com>
10968 David Sherwood <david.sherwood@arm.com>
10970 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10971 Use plus_constant instead of gen_rtx_PLUS.
10973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10974 Alan Hayward <alan.hayward@arm.com>
10975 David Sherwood <david.sherwood@arm.com>
10977 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10978 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10979 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10980 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10981 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10982 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10983 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10984 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10985 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10986 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10988 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10989 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10990 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10991 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10992 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10993 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10994 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10995 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10996 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10997 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10999 * expr.c (emit_move_resolve_push): Treat the input and result
11000 of PUSH_ROUNDING as a poly_int64.
11001 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11002 (emit_push_insn): Likewise.
11003 * lra-eliminations.c (mark_not_eliminable): Likewise.
11004 * recog.c (push_operand): Likewise.
11005 * reload1.c (elimination_effects): Likewise.
11006 * rtlanal.c (nonzero_bits1): Likewise.
11007 * calls.c (store_one_arg): Likewise. Require the padding to be
11008 known at compile time.
11010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11011 Alan Hayward <alan.hayward@arm.com>
11012 David Sherwood <david.sherwood@arm.com>
11014 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11015 Use plus_constant instead of gen_rtx_PLUS.
11017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11018 Alan Hayward <alan.hayward@arm.com>
11019 David Sherwood <david.sherwood@arm.com>
11021 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11022 rather than an int.
11024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11025 Alan Hayward <alan.hayward@arm.com>
11026 David Sherwood <david.sherwood@arm.com>
11028 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11029 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11030 via stack temporaries. Treat the mode size as polynomial too.
11032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11033 Alan Hayward <alan.hayward@arm.com>
11034 David Sherwood <david.sherwood@arm.com>
11036 * expr.c (expand_expr_real_2): When handling conversions involving
11037 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11038 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11039 as a poly_uint64 too.
11041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11042 Alan Hayward <alan.hayward@arm.com>
11043 David Sherwood <david.sherwood@arm.com>
11045 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
11052 REGMODE_NATURAL_SIZE.
11053 * expmed.c (store_bit_field_1): Likewise.
11054 * expr.c (store_constructor): Likewise.
11055 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11056 and polynomial REGMODE_NATURAL_SIZE.
11057 (gen_lowpart_common): Likewise.
11058 * reginfo.c (record_subregs_of_mode): Likewise.
11059 * rtlanal.c (read_modify_subreg_p): Likewise.
11061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11062 Alan Hayward <alan.hayward@arm.com>
11063 David Sherwood <david.sherwood@arm.com>
11065 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11066 numbers of elements.
11068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11069 Alan Hayward <alan.hayward@arm.com>
11070 David Sherwood <david.sherwood@arm.com>
11072 * match.pd: Cope with polynomial numbers of vector 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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11079 in a POINTER_PLUS_EXPR.
11081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11082 Alan Hayward <alan.hayward@arm.com>
11083 David Sherwood <david.sherwood@arm.com>
11085 * omp-simd-clone.c (simd_clone_subparts): New function.
11086 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11087 (ipa_simd_modify_function_body): Likewise.
11089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11090 Alan Hayward <alan.hayward@arm.com>
11091 David Sherwood <david.sherwood@arm.com>
11093 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11094 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11095 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11096 (expand_vector_condition, vector_element): Likewise.
11097 (subparts_gt): New function.
11098 (get_compute_type): Use subparts_gt.
11099 (count_type_subparts): Delete.
11100 (expand_vector_operations_1): Use subparts_gt instead of
11101 count_type_subparts.
11103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11104 Alan Hayward <alan.hayward@arm.com>
11105 David Sherwood <david.sherwood@arm.com>
11107 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11108 (vect_compile_time_alias): ...this new function. Do the calculation
11109 on poly_ints rather than trees.
11110 (vect_prune_runtime_alias_test_list): Update call accordingly.
11112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11113 Alan Hayward <alan.hayward@arm.com>
11114 David Sherwood <david.sherwood@arm.com>
11116 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11118 (vect_schedule_slp_instance): Likewise.
11120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11121 Alan Hayward <alan.hayward@arm.com>
11122 David Sherwood <david.sherwood@arm.com>
11124 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11125 constant and extern definitions for variable-length vectors.
11126 (vect_get_constant_vectors): Note that the number of units
11127 is known to be constant.
11129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11130 Alan Hayward <alan.hayward@arm.com>
11131 David Sherwood <david.sherwood@arm.com>
11133 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11134 of units as polynomial. Choose between WIDE and NARROW based
11137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11138 Alan Hayward <alan.hayward@arm.com>
11139 David Sherwood <david.sherwood@arm.com>
11141 * tree-vect-stmts.c (simd_clone_subparts): New function.
11142 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11145 Alan Hayward <alan.hayward@arm.com>
11146 David Sherwood <david.sherwood@arm.com>
11148 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11149 vectors as polynomial. Use build_index_vector for
11150 IFN_GOMP_SIMD_LANE.
11152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11153 Alan Hayward <alan.hayward@arm.com>
11154 David Sherwood <david.sherwood@arm.com>
11156 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11157 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11158 for variable-length vectors.
11159 (vectorizable_mask_load_store): Treat the number of units as
11160 polynomial, asserting that it is constant if the condition has
11161 already been enforced.
11162 (vectorizable_store, vectorizable_load): Likewise.
11164 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11165 Alan Hayward <alan.hayward@arm.com>
11166 David Sherwood <david.sherwood@arm.com>
11168 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11169 of units as polynomial. Punt if we can't tell at compile time
11170 which vector contains the final result.
11172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11173 Alan Hayward <alan.hayward@arm.com>
11174 David Sherwood <david.sherwood@arm.com>
11176 * tree-vect-loop.c (vectorizable_induction): Treat the number
11177 of units as polynomial. Punt on SLP inductions. Use an integer
11178 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11179 cast of such a series for variable-length floating-point
11182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11183 Alan Hayward <alan.hayward@arm.com>
11184 David Sherwood <david.sherwood@arm.com>
11186 * tree.h (build_index_vector): Declare.
11187 * tree.c (build_index_vector): New function.
11188 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11189 of units as polynomial, forcibly converting it to a constant if
11190 vectorizable_reduction has already enforced the condition.
11191 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11192 to create a {1,2,3,...} vector.
11193 (vectorizable_reduction): Treat the number of units as polynomial.
11194 Choose vectype_in based on the largest scalar element size rather
11195 than the smallest number of units. Enforce the restrictions
11198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11199 Alan Hayward <alan.hayward@arm.com>
11200 David Sherwood <david.sherwood@arm.com>
11202 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11203 number of units as polynomial.
11205 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11206 Alan Hayward <alan.hayward@arm.com>
11207 David Sherwood <david.sherwood@arm.com>
11209 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11210 * target.def (autovectorize_vector_sizes): Return the vector sizes
11211 by pointer, using vector_sizes rather than a bitmask.
11212 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11213 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11214 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11216 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11217 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11218 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11219 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11220 * omp-general.c (omp_max_vf): Likewise.
11221 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11222 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11223 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11224 * tree-vect-slp.c (vect_slp_bb): Likewise.
11225 * doc/tm.texi: Regenerate.
11226 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11228 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11229 the vector size as a poly_uint64 rather than an unsigned int.
11230 (current_vector_size): Change from an unsigned int to a poly_uint64.
11231 (get_vectype_for_scalar_type): Update accordingly.
11232 * tree.h (build_truth_vector_type): Take the size and number of
11233 units as a poly_uint64 rather than an unsigned int.
11234 (build_vector_type): Add a temporary overload that takes
11235 the number of units as a poly_uint64 rather than an unsigned int.
11236 * tree.c (make_vector_type): Likewise.
11237 (build_truth_vector_type): Take the number of units as a poly_uint64
11238 rather than an unsigned int.
11240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11241 Alan Hayward <alan.hayward@arm.com>
11242 David Sherwood <david.sherwood@arm.com>
11244 * target.def (get_mask_mode): Take the number of units and length
11245 as poly_uint64s rather than unsigned ints.
11246 * targhooks.h (default_get_mask_mode): Update accordingly.
11247 * targhooks.c (default_get_mask_mode): Likewise.
11248 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11249 * doc/tm.texi: Regenerate.
11251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11252 Alan Hayward <alan.hayward@arm.com>
11253 David Sherwood <david.sherwood@arm.com>
11255 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11256 * omp-general.c (omp_max_vf): Likewise.
11257 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11258 (expand_omp_simd): Handle polynomial safelen.
11259 * omp-low.c (omplow_simd_context): Add a default constructor.
11260 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11261 (lower_rec_simd_input_clauses): Update accordingly.
11262 (lower_rec_input_clauses): Likewise.
11264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11265 Alan Hayward <alan.hayward@arm.com>
11266 David Sherwood <david.sherwood@arm.com>
11268 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11269 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11270 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11271 (vect_analyze_slp_cost): Likewise.
11272 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11273 (vect_model_load_cost): Likewise.
11275 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11276 Alan Hayward <alan.hayward@arm.com>
11277 David Sherwood <david.sherwood@arm.com>
11279 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11280 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11281 from an unsigned int * to a poly_uint64_pod *.
11282 (calculate_unrolling_factor): New function.
11283 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11286 Alan Hayward <alan.hayward@arm.com>
11287 David Sherwood <david.sherwood@arm.com>
11289 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11290 from an unsigned int to a poly_uint64.
11291 (_loop_vec_info::slp_unrolling_factor): Likewise.
11292 (_loop_vec_info::vectorization_factor): Change from an int
11294 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11295 (vect_get_num_vectors): New function.
11296 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11297 (vect_get_num_copies): Use vect_get_num_vectors.
11298 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11299 to an unsigned int *.
11300 (vect_analyze_data_refs): Change min_vf from an int * to a
11302 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11303 than an unsigned HOST_WIDE_INT.
11304 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11305 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11306 to an unsigned int *.
11307 (vect_analyze_data_ref_dependences): Likewise.
11308 (vect_compute_data_ref_alignment): Handle polynomial vf.
11309 (vect_enhance_data_refs_alignment): Likewise.
11310 (vect_prune_runtime_alias_test_list): Likewise.
11311 (vect_shift_permute_load_chain): Likewise.
11312 (vect_supportable_dr_alignment): Likewise.
11313 (dependence_distance_ge_vf): Take the vectorization factor as a
11314 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11315 (vect_analyze_data_refs): Change min_vf from an int * to a
11317 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11318 vfm1 as a poly_uint64 rather than an int. Make the same change
11319 for the returned bound_scalar.
11320 (vect_gen_vector_loop_niters): Handle polynomial vf.
11321 (vect_do_peeling): Likewise. Update call to
11322 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11323 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11325 * tree-vect-loop.c (vect_determine_vectorization_factor)
11326 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11327 (vect_get_known_peeling_cost): Likewise.
11328 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11329 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11330 (vect_transform_loop): Likewise. Use the lowest possible VF when
11331 updating the upper bounds of the loop.
11332 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11333 rather than an int.
11334 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11335 polynomial unroll factors.
11336 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11337 (vect_make_slp_decision): Likewise.
11338 (vect_supported_load_permutation_p): Likewise, and polynomial
11340 (vect_analyze_slp_cost): Handle polynomial vf.
11341 (vect_slp_analyze_node_operations): Likewise.
11342 (vect_slp_analyze_bb_1): Likewise.
11343 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11344 than an unsigned HOST_WIDE_INT.
11345 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11346 (vectorizable_load): Handle polynomial vf.
11347 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11349 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11352 Alan Hayward <alan.hayward@arm.com>
11353 David Sherwood <david.sherwood@arm.com>
11355 * match.pd: Handle bit operations involving three constants
11356 and try to fold one pair.
11358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11360 * tree-vect-loop-manip.c: Include gimple-fold.h.
11361 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11362 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11363 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11364 Add a path that uses a step of VF instead of 1, but disable it
11366 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11367 and niters_no_overflow parameters. Update calls to
11368 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11369 Create a new SSA name if the latter choses to use a ste other
11370 than zero, and return it via niters_vector_mult_vf_var.
11371 * tree-vect-loop.c (vect_transform_loop): Update calls to
11372 vect_do_peeling, vect_gen_vector_loop_niters and
11373 slpeel_make_loop_iterate_ntimes.
11374 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11375 (vect_gen_vector_loop_niters): Update declarations after above changes.
11377 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11379 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11380 128-bit round to integer instructions.
11381 (ceil<mode>2): Likewise.
11382 (btrunc<mode>2): Likewise.
11383 (round<mode>2): Likewise.
11385 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11387 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11388 unaligned VSX load/store on P8/P9.
11389 (expand_block_clear): Allow the use of unaligned VSX
11390 load/store on P8/P9.
11392 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11394 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11396 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11397 swap associated with both a load and a store.
11399 2018-01-02 Andrew Waterman <andrew@sifive.com>
11401 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11402 * config/riscv/riscv.md (clear_cache): Use it.
11404 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11406 * web.c: Remove out-of-date comment.
11408 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11410 * expr.c (fixup_args_size_notes): Check that any existing
11411 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11412 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11413 (emit_single_push_insn): ...here.
11415 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11417 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11418 (const_vector_encoded_nelts): New function.
11419 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11420 (const_vector_int_elt, const_vector_elt): Declare.
11421 * emit-rtl.c (const_vector_int_elt_1): New function.
11422 (const_vector_elt): Likewise.
11423 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11424 of CONST_VECTOR_ELT.
11426 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11428 * expr.c: Include rtx-vector-builder.h.
11429 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11430 directly on the tree encoding.
11431 (const_vector_from_tree): Likewise.
11432 * optabs.c: Include rtx-vector-builder.h.
11433 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11434 sequence of "u" values.
11435 * vec-perm-indices.c: Include rtx-vector-builder.h.
11436 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11437 directly on the vec_perm_indices encoding.
11439 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11441 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11442 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11443 * rtx-vector-builder.h: New file.
11444 * rtx-vector-builder.c: Likewise.
11445 * rtl.h (rtx_def::u2): Add a const_vector field.
11446 (CONST_VECTOR_NPATTERNS): New macro.
11447 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11448 (CONST_VECTOR_DUPLICATE_P): Likewise.
11449 (CONST_VECTOR_STEPPED_P): Likewise.
11450 (CONST_VECTOR_ENCODED_ELT): Likewise.
11451 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11452 (unwrap_const_vec_duplicate): Likewise.
11453 (const_vec_series_p): Check for a non-duplicated vector encoding.
11454 Say that the function only returns true for integer vectors.
11455 * emit-rtl.c: Include rtx-vector-builder.h.
11456 (gen_const_vec_duplicate_1): Delete.
11457 (gen_const_vector): Call gen_const_vec_duplicate instead of
11458 gen_const_vec_duplicate_1.
11459 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11460 (gen_const_vec_duplicate): Use rtx_vector_builder.
11461 (gen_const_vec_series): Likewise.
11462 (gen_rtx_CONST_VECTOR): Likewise.
11463 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11464 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11465 Build a new vector rather than modifying a CONST_VECTOR in-place.
11466 (handle_special_swappables): Update call accordingly.
11467 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11468 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11469 Build a new vector rather than modifying a CONST_VECTOR in-place.
11470 (handle_special_swappables): Update call accordingly.
11472 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11474 * simplify-rtx.c (simplify_const_binary_operation): Use
11475 CONST_VECTOR_ELT instead of XVECEXP.
11477 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11479 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11480 the selector elements to be different from the data elements
11481 if the selector is a VECTOR_CST.
11482 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11483 ssizetype for the selector.
11485 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11487 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11488 before testing each element individually.
11489 * tree-vect-generic.c (lower_vec_perm): Likewise.
11491 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11493 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11494 * selftest-run-tests.c (selftest::run_tests): Call it.
11495 * vector-builder.h (vector_builder::operator ==): New function.
11496 (vector_builder::operator !=): Likewise.
11497 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11498 (vec_perm_indices::all_from_input_p): New function.
11499 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11500 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11501 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11502 instead of reading the VECTOR_CST directly. Detect whether both
11503 vector inputs are the same before constructing the vec_perm_indices,
11504 and update the number of inputs argument accordingly. Use the
11505 utility functions added above. Only construct sel2 if we need to.
11507 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11509 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11510 the broadcast of the low byte.
11511 (expand_mult_highpart): Use an explicit encoding for the permutes.
11512 * optabs-query.c (can_mult_highpart_p): Likewise.
11513 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11514 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11515 (vectorizable_bswap): Likewise.
11516 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11517 explicit encoding for the power-of-2 permutes.
11518 (vect_permute_store_chain): Likewise.
11519 (vect_grouped_load_supported): Likewise.
11520 (vect_permute_load_chain): Likewise.
11522 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11524 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11525 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11526 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11527 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11528 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11529 (vect_gen_perm_mask_any): Likewise.
11531 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11533 * int-vector-builder.h: New file.
11534 * vec-perm-indices.h: Include int-vector-builder.h.
11535 (vec_perm_indices): Redefine as an int_vector_builder.
11536 (auto_vec_perm_indices): Delete.
11537 (vec_perm_builder): Redefine as a stand-alone class.
11538 (vec_perm_indices::vec_perm_indices): New function.
11539 (vec_perm_indices::clamp): Likewise.
11540 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11541 (vec_perm_indices::new_vector): New function.
11542 (vec_perm_indices::new_expanded_vector): Update for new
11543 vec_perm_indices class.
11544 (vec_perm_indices::rotate_inputs): New function.
11545 (vec_perm_indices::all_in_range_p): Operate directly on the
11546 encoded form, without computing elided elements.
11547 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11548 encoding. Update for new vec_perm_indices class.
11549 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11550 the given vec_perm_builder.
11551 (expand_vec_perm_var): Update vec_perm_builder constructor.
11552 (expand_mult_highpart): Use vec_perm_builder instead of
11553 auto_vec_perm_indices.
11554 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11555 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11556 or double series encoding as appropriate.
11557 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11558 vec_perm_indices instead of auto_vec_perm_indices.
11559 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11560 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11561 (vect_permute_store_chain): Likewise.
11562 (vect_grouped_load_supported): Likewise.
11563 (vect_permute_load_chain): Likewise.
11564 (vect_shift_permute_load_chain): Likewise.
11565 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11566 (vect_transform_slp_perm_load): Likewise.
11567 (vect_schedule_slp_instance): Likewise.
11568 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11569 (vectorizable_mask_load_store): Likewise.
11570 (vectorizable_bswap): Likewise.
11571 (vectorizable_store): Likewise.
11572 (vectorizable_load): Likewise.
11573 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11574 vec_perm_indices instead of auto_vec_perm_indices. Use
11575 tree_to_vec_perm_builder to read the vector from a tree.
11576 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11577 vec_perm_builder instead of a vec_perm_indices.
11578 (have_whole_vector_shift): Use vec_perm_builder and
11579 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11580 truncation to calc_vec_perm_mask_for_shift.
11581 (vect_create_epilog_for_reduction): Likewise.
11582 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11583 from auto_vec_perm_indices to vec_perm_indices.
11584 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11585 instead of changing individual elements.
11586 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11587 the vector in d.perm.
11588 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11589 from auto_vec_perm_indices to vec_perm_indices.
11590 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11591 instead of changing individual elements.
11592 (arm_vectorize_vec_perm_const): Use new_vector to install
11593 the vector in d.perm.
11594 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11595 Update vec_perm_builder constructor.
11596 (rs6000_expand_interleave): Likewise.
11597 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11598 (rs6000_expand_interleave): Likewise.
11600 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11602 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11603 to qimode could truncate the indices.
11604 * optabs.c (expand_vec_perm_var): Likewise.
11606 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11608 * Makefile.in (OBJS): Add vec-perm-indices.o.
11609 * vec-perm-indices.h: New file.
11610 * vec-perm-indices.c: Likewise.
11611 * target.h (vec_perm_indices): Replace with a forward class
11613 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11614 * optabs.h: Include vec-perm-indices.h.
11615 (expand_vec_perm): Delete.
11616 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11617 (expand_vec_perm_const): Declare.
11618 * target.def (vec_perm_const_ok): Replace with...
11619 (vec_perm_const): ...this new hook.
11620 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11621 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11622 * doc/tm.texi: Regenerate.
11623 * optabs.def (vec_perm_const): Delete.
11624 * doc/md.texi (vec_perm_const): Likewise.
11625 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11626 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11627 expand_vec_perm for constant permutation vectors. Assert that
11628 the mode of variable permutation vectors is the integer equivalent
11629 of the mode that is being permuted.
11630 * optabs-query.h (selector_fits_mode_p): Declare.
11631 * optabs-query.c: Include vec-perm-indices.h.
11632 (selector_fits_mode_p): New function.
11633 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11634 is defined, instead of checking whether the vec_perm_const_optab
11635 exists. Use targetm.vectorize.vec_perm_const instead of
11636 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11637 fit in the vector mode before using a variable permute.
11638 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11639 vec_perm_indices instead of an rtx.
11640 (expand_vec_perm): Replace with...
11641 (expand_vec_perm_const): ...this new function. Take the selector
11642 as a vec_perm_indices rather than an rtx. Also take the mode of
11643 the selector. Update call to shift_amt_for_vec_perm_mask.
11644 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11645 Use vec_perm_indices::new_expanded_vector to expand the original
11646 selector into bytes. Check whether the indices fit in the vector
11647 mode before using a variable permute.
11648 (expand_vec_perm_var): Make global.
11649 (expand_mult_highpart): Use expand_vec_perm_const.
11650 * fold-const.c: Includes vec-perm-indices.h.
11651 * tree-ssa-forwprop.c: Likewise.
11652 * tree-vect-data-refs.c: Likewise.
11653 * tree-vect-generic.c: Likewise.
11654 * tree-vect-loop.c: Likewise.
11655 * tree-vect-slp.c: Likewise.
11656 * tree-vect-stmts.c: Likewise.
11657 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11659 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11660 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11661 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11662 (aarch64_vectorize_vec_perm_const): ...this new function.
11663 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11664 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11665 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11666 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11667 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11668 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11669 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11671 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11672 check for NEON modes.
11673 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11674 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11675 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11676 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11678 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11679 the old VEC_PERM_CONST conditions.
11680 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11681 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11682 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11683 (ia64_vectorize_vec_perm_const_ok): Merge into...
11684 (ia64_vectorize_vec_perm_const): ...this new function.
11685 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11686 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11687 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11688 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11689 * config/mips/mips.c (mips_expand_vec_perm_const)
11690 (mips_vectorize_vec_perm_const_ok): Merge into...
11691 (mips_vectorize_vec_perm_const): ...this new function.
11692 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11693 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11694 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11695 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11696 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11697 (rs6000_expand_vec_perm_const): Delete.
11698 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11700 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11701 (altivec_expand_vec_perm_const_le): Take each operand individually.
11702 Operate on constant selectors rather than rtxes.
11703 (altivec_expand_vec_perm_const): Likewise. Update call to
11704 altivec_expand_vec_perm_const_le.
11705 (rs6000_expand_vec_perm_const): Delete.
11706 (rs6000_vectorize_vec_perm_const_ok): Delete.
11707 (rs6000_vectorize_vec_perm_const): New function.
11708 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11709 an element count and rtx array.
11710 (rs6000_expand_extract_even): Update call accordingly.
11711 (rs6000_expand_interleave): Likewise.
11712 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11713 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11714 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11715 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11716 (rs6000_expand_vec_perm_const): Delete.
11717 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11718 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11719 (altivec_expand_vec_perm_const_le): Take each operand individually.
11720 Operate on constant selectors rather than rtxes.
11721 (altivec_expand_vec_perm_const): Likewise. Update call to
11722 altivec_expand_vec_perm_const_le.
11723 (rs6000_expand_vec_perm_const): Delete.
11724 (rs6000_vectorize_vec_perm_const_ok): Delete.
11725 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11726 reference to the SPE evmerge intructions.
11727 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11728 an element count and rtx array.
11729 (rs6000_expand_extract_even): Update call accordingly.
11730 (rs6000_expand_interleave): Likewise.
11731 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11732 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11734 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11736 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11738 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11739 vector mode and that that mode matches the mode of the data
11741 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11742 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11743 directly using expand_vec_perm_1 when forcing selectors into
11745 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11747 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11749 * optabs-query.h (can_vec_perm_p): Delete.
11750 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11751 * optabs-query.c (can_vec_perm_p): Split into...
11752 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11753 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11754 particular selector is valid.
11755 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11756 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11757 (vect_grouped_load_supported): Likewise.
11758 (vect_shift_permute_load_chain): Likewise.
11759 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11760 (vect_transform_slp_perm_load): Likewise.
11761 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11762 (vectorizable_bswap): Likewise.
11763 (vect_gen_perm_mask_checked): Likewise.
11764 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11765 implementations of variable permutation vectors into account
11766 when deciding which selector to use.
11767 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11768 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11769 with a false third argument.
11770 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11771 to test whether the constant selector is valid and can_vec_perm_var_p
11772 to test whether a variable selector is valid.
11774 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11776 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11777 * optabs-query.c (can_vec_perm_p): Likewise.
11778 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11779 instead of vec_perm_indices.
11780 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11781 (vect_gen_perm_mask_checked): Likewise,
11782 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11783 (vect_gen_perm_mask_checked): Likewise,
11785 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11787 * optabs-query.h (qimode_for_vec_perm): Declare.
11788 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11789 (qimode_for_vec_perm): ...this new function.
11790 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11792 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11794 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11795 does not have a conditional at the top.
11797 2018-01-02 Richard Biener <rguenther@suse.de>
11799 * ipa-inline.c (big_speedup_p): Fix expression.
11801 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11804 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11807 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11811 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11812 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11813 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11814 cond_taken_branch_cost 3->4.
11816 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11818 PR tree-optimization/83581
11819 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11820 TODO_cleanup_cfg if any changes have been made.
11822 PR middle-end/83608
11823 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11824 convert_modes if target mode has the right side, but different mode
11827 PR middle-end/83609
11828 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11829 last argument when extracting from CONCAT. If either from_real or
11830 from_imag is NULL, use expansion through memory. If result is not
11831 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11832 the parts directly to inner mode, if even that fails, use expansion
11835 PR middle-end/83623
11836 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11837 check for bswap in mode rather than HImode and use that in expand_unop
11840 Copyright (C) 2018 Free Software Foundation, Inc.
11842 Copying and distribution of this file, with or without modification,
11843 are permitted in any medium without royalty provided the copyright
11844 notice and this notice are preserved.