1 2018-05-01 Tom de Vries <tom@codesourcery.com>
4 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5 * vec.c (test_ordered_remove_if): New function.
6 (vec_c_tests): Call test_ordered_remove_if.
7 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
8 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
9 * tree-vect-patterns.c (vect_pattern_recog_1): Use
10 VEC_ORDERED_REMOVE_IF.
12 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14 PR tree-optimization/82665
15 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
16 pointer subtraction where arguments come from a memchr call.
18 2018-05-01 Jakub Jelinek <jakub@redhat.com>
20 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
21 --push-state --as-needed and --pop-state instead of --as-needed and
22 --no-as-needed if ld supports it.
23 * configure: Regenerated.
26 * doc/install.texi2html: Replace _002d with - and _002a with * in
27 generated html files using sed.
29 2018-04-30 David Malcolm <dmalcolm@redhat.com>
32 * gcc-rich-location.c (blank_line_before_p): New function.
33 (use_new_line): New function.
34 (gcc_rich_location::add_fixit_insert_formatted): New function.
36 (gcc_rich_location::add_fixit_insert_formatted): New function.
38 2018-04-30 David Malcolm <dmalcolm@redhat.com>
40 * selftest.c (assert_streq): Rename "expected" and "actual" to
41 "val1" and "val2". Extend NULL-handling to cover both inputs
42 symmetrically, while still requiring both to be non-NULL for a pass.
43 * selftest.h (assert_streq): Rename "expected" and "actual" to
45 (ASSERT_EQ): Likewise.
46 (ASSERT_EQ_AT): Likewise.
47 (ASSERT_KNOWN_EQ): Likewise.
48 (ASSERT_KNOWN_EQ_AT): Likewise.
49 (ASSERT_NE): Likewise.
50 (ASSERT_MAYBE_NE): Likewise.
51 (ASSERT_MAYBE_NE_AT): Likewise.
52 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
53 the assertion to pass.
54 (ASSERT_STREQ_AT): Likewise.
56 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
58 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
59 interaction with -pie.
61 2018-04-30 David Malcolm <dmalcolm@redhat.com>
63 * selftest.h: Fix alphabetization of per-source-file selftest
66 2018-04-30 Jason Merrill <jason@redhat.com>
68 PR c++/61982 - dead stores to destroyed objects.
69 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
72 2018-04-30 Jason Merrill <jason@redhat.com>
74 * tree.c (build_clobber): New.
76 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
78 2018-04-30 David Malcolm <dmalcolm@redhat.com>
80 * diagnostic-show-locus.c (layout::layout): Update for
81 location_get_source_line returning a char_span.
82 (struct char_span): Move to input.h.
83 (struct correction): Update for fields in char_span becoming
85 (struct source_line): Update for location_get_source_line
86 returning a char_span.
87 (layout::print_line): Likewise.
88 * edit-context.c (edited_file::print_content): Likewise.
89 (edited_file::print_diff_hunk): Likewise.
90 (edited_file::print_run_of_changed_lines): Likewise.
91 (edited_file::get_num_lines): Likewise.
92 (edited_line::edited_line): Likewise.
93 * final.c (asm_show_source): Likewise.
94 * input.c (location_get_source_line): Convert return type
95 from const char * to char_span, losing the final "line_len"
97 (dump_location_info): Update for the above.
98 (get_substring_ranges_for_loc): Likewise. Use a char_span
99 when handling the literal within the line.
100 (test_reading_source_line): Update for location_get_source_line
101 returning a char_span.
102 * input.h (class char_span): Move here from
103 diagnostic-show-locus.c, converting from a struct to a class.
104 Make data members private.
105 (char_span::operator bool): New.
106 (char_span::length): New.
107 (char_span::get_buffer): New.
108 (char_span::operator[]): New.
109 (char_span::subspan): Make const.
110 (char_span::xstrdup): New.
111 (location_get_source_line): Convert return type from const char *
112 to char_span, losing the final "line_size" param.
114 2018-04-30 Jan Hubicka <jh@suse.cz>
116 * lto-wrapper.c (ltrans_priorities): New static var.
118 (run_gcc): Read priorities and if doing parallel build order
119 the Makefile by them.
121 2018-04-30 David Malcolm <dmalcolm@redhat.com>
123 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
125 2018-04-30 Richard Biener <rguenther@suse.de>
127 * tree-cfg.c (verify_address): Remove base argument, add
128 flag whether to check TREE_ADDRESSABLE and do that.
129 (verify_expr): Remove.
130 (verify_types_in_gimple_reference): Add pieces from verify_expr.
131 (verify_gimple_assign_single): Likewise.
132 (verify_gimple_switch): Likewise.
133 (verify_expr_location_1): Dereference tp once. Add (disabled)
134 piece from verify_expr.
135 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
137 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
139 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
141 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
143 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
144 (small_data_pattern): Likewise.
145 (arc_rewrite_small_data): Likewise.
146 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
147 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
148 (get_symbol_alignment): New function.
149 (legitimate_small_data_address_p): Likewise.
150 (legitimate_scaled_address): Update, call
151 legitimate_small_data_address_p.
152 (output_sdata): New static variable.
153 (arc_print_operand): Update how we handle small data operands.
154 (arc_print_operand_address): Likewise.
155 (arc_legitimate_address_p): Update, use
156 legitimate_small_data_address_p.
157 (arc_rewrite_small_data_p): Remove.
158 (arc_rewrite_small_data_1): Likewise.
159 (arc_rewrite_small_data): Likewise.
160 (small_data_pattern): Likewise.
161 (compact_sda_memory_operand): Update to use
162 legitimate_small_data_address_p and get_symbol_alignment.
163 (prepare_move_operands): Don't rewite sdata pattern.
164 (prepare_extend_operands): Remove.
165 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
167 (zero_extendqisi2): Likewise.
168 (zero_extendhisi2): Likewise.
169 (extendqihi2): Likewise.
170 (extendqisi2): Likewise.
171 (extendhisi2): Likewise.
175 * config/arc/constraints.md (Usd): Change it to memory constraint.
177 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
179 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
180 as source of std instructions.
181 * config/arc/arc.md (movsi_insn): Update pattern predicate to
182 allow 6-bit constants as source for store instructions.
183 (movdi_insn): Update instruction pattern to allow 6-bit constants
184 as source for store instructions.
186 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
188 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
190 2018-04-30 Nathan Sidwell <nathan@acm.org>
191 Sandra Loosemore <sandra@codesourcery.com>
193 * dumpfile.c (dump_open): Allow '-' for stdout.
194 * doc/invoke.texi (Developer Options): Document dump filename
195 determination early. Document stdin/stdout selection.
197 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
199 Microblaze Target: PIC data text relative
201 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
202 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
204 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
205 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
206 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
207 New addressing mode for data-text relative position indepenedent code.
208 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
209 'ADDRESS_SYMBOLIC_TXT_REL'.
210 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
211 (microblaze_legitimate_pic_operand): Exclude function calls from
212 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
213 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
215 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
216 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
217 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
218 for 'address + offset'.
219 (microblaze_expand_prologue): Add new function prologue call for
221 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
222 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
223 table in case of TARGET_PIC_DATA_TEXT_REL.
224 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
225 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
226 Add new macros 'UNSPEC_TEXT',
227 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
228 + exclude function calls from 'UNSPEC_PLT' in case of data text
230 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
231 new target hook for generating address diff vector tables in case of
233 * doc/tm.texi : Regenerate.
234 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
235 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
236 of addr diff vector generation.
237 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
238 target hook definition.
239 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
240 Add default function for generate_pic_addr_diff_vec -> flag_pic.
241 * doc/invoke.texi (Add new pic option): Add new microblaze pic
242 option for data text relative.
244 2018-04-30 Richard Biener <rguenther@suse.de>
246 * tree-chrec.h (evolution_function_is_constant_p): Remove
248 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
250 2018-04-30 Richard Biener <rguenther@suse.de>
253 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
255 2018-04-30 Richard Biener <rguenther@suse.de>
257 PR tree-optimization/28364
258 PR tree-optimization/85275
259 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
260 copying first exit test.
262 2018-04-28 Mark Wielaard <mark@klomp.org>
264 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
266 (dwarf_AT): Handle DW_AT_addr_base.
267 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
269 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
272 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
273 (*ashl<dwi>3_doubleword_mask_1): Ditto.
274 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
275 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
277 2018-04-28 Richard Biener <rguenther@suse.de>
279 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
280 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
281 to reflect use. Only add interesting stmts.
283 2018-04-27 Martin Jambor <mjambor@suse.cz>
286 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
287 the jump function allows for passing through aggregate values.
289 2018-04-27 David Malcolm <dmalcolm@redhat.com>
291 * input.h (in_system_header_at): Convert from macro to inline
293 (from_macro_expansion_at): Likewise.
294 (from_macro_definition_at): Likewise.
296 2018-04-27 Jeff Law <law@redhat.com>
298 * config.gcc: Mark tile* targets as deprecated/obsolete.
300 2018-04-27 Richard Biener <rguenther@suse.de>
302 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
305 2018-04-27 Richard Biener <rguenther@suse.de>
307 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
309 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
311 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
312 with Yd constraint. Set "preferred_for_speed" attribute from
313 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
315 (*movdi_internal): Ditto.
316 (movti_interunit splitters): Remove
317 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
318 (movdi_interunit splitters): Ditto.
319 * config/i386/constraints.md (Ye): Remove.
320 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
322 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
325 * config/aarch64/constraints.md (Usg): Limit to 31.
328 2018-04-27 Jakub Jelinek <jakub@redhat.com>
330 PR tree-optimization/85529
331 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
332 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
333 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
334 zero extension or masking of the MSB bit.
335 (optimize_range_tests): Add FIRST_BB argument, pass it through
336 to optimize_range_tests_var_bound.
337 (maybe_optimize_range_tests, reassociate_bb): Adjust
338 optimize_range_tests callers.
340 2018-04-26 Richard Biener <rguenther@suse.de>
341 Jakub Jelinek <jakub@redhat.com>
343 * cgraph.h (symbol_table): Just declare debug method here.
344 * symtab.c (symbol_table::debug): Define.
346 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
348 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
350 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
352 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
353 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
354 (*movdi_internal): Substitute Yi and Yj constraint with x
355 and Ym and Yn constraint with y constraint. Update "isa"
356 attribute and set "preferred_for_speed" attribute from
357 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
358 (*movsi_internal): Ditto.
359 (*movdf_internal): Ditto.
360 (*movsf_internal): Ditto.
361 (*zero_extendsidi2): Ditto.
362 * config/i386/sse.md (vec_set<mode>_0): Ditto.
363 (sse2_loadld): Ditto.
364 (*vec_extract<ssevecmodelower>_0): Ditto.
365 (*vec_extractv4si_0_zext_sse4): Ditto.
366 (vec_concatv2di): Ditto.
367 (*vec_dup<mode>): Ditto.
368 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
369 * config/i386/constraints.md (Yi): Remove.
374 2018-04-26 Nathan Sidwell <nathan@acm.org>
376 * dumpfile.c (dump_open): New.
377 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
378 (dump_finish): Detect stdio/stderr by value not name.
380 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
382 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
384 2018-04-26 Tom de Vries <tom@codesourcery.com>
387 * config/nvptx/nvptx.c (verify_neutering_jumps)
388 (verify_neutering_labels): New function
389 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
391 2018-04-26 Tom de Vries <tom@codesourcery.com>
394 * config/nvptx/nvptx.c (needs_neutering_p): New function.
395 (nvptx_single): Use needs_neutering_p to skip over insns that do not
398 2018-04-26 Richard Biener <rguenther@suse.de>
399 Tom de Vries <tom@codesourcery.com>
402 * lto-streamer-out.c (output_function): Fixup loops if required to match
403 discovery done in the reader.
405 2018-04-26 Richard Biener <rguenther@suse.de>
407 PR tree-optimization/85116
408 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
409 have a loop exit from the single latch predecessor. Remove
410 case of header with just condition.
411 (ch_base::copy_headers): Exclude infinite loops from any
413 (pass_ch::execute): Record exits.
415 2018-04-26 Richard Biener <rguenther@suse.de>
417 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
418 prologue cost vector and pass it to vect_get_load_cost.
419 (vect_get_peeling_costs_all_drs): Likewise.
420 (vect_peeling_hash_get_lowest_cost): Likewise.
421 (vect_enhance_data_refs_alignment): Likewise.
423 2018-04-26 Richard Biener <rguenther@suse.de>
426 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
427 checking of integer<->pointer conversions.
428 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
429 sign-/zero-extending pointer types.
430 (expand_omp_for_static_chunk): Likewise.
432 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
433 Jean Lee <xiaoyur347@gmail.com>
435 * config/mips/mips.c (mips_asan_shadow_offset): New function.
436 (TARGET_ASAN_SHADOW_OFFSET): Define.
437 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
438 true for -fsanitize=address.
440 2018-04-25 Mark Wielaard <mark@klomp.org>
442 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
445 2018-04-25 Jakub Jelinek <jakub@redhat.com>
447 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
448 than "alu", remove explicit "memory" and "imm_disp" attributes.
449 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
452 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
453 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
456 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
459 * config/i386/i386.c (ix86_expand_builtin): Change memory
460 operand to XI, extend p0 to Pmode.
461 * config/i386/i386.md: Change unspec volatile and operand
462 1 mode to XI, change operand 0 mode to P.
464 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
466 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
467 GET_MODE_MASK before any checking.
468 (nds32_can_use_bset_p): Likewise.
469 (nds32_can_use_btgl_p): Likewise.
471 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
473 * config/nds32/nds32-doubleword.md: New define_split pattern for
474 illegal register number.
476 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
478 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
480 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
482 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
484 2018-04-25 Richard Biener <rguenther@suse.de>
486 * lto-streamer.h (LTO_major_version): Bump to 8.
488 2018-04-25 Jakub Jelinek <jakub@redhat.com>
490 * BASE-VER: Set to 9.0.0.
492 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
494 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
495 in __abskf2 and __powikf2.
497 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
500 * config/aarch64/constraints.md (Usg, Usj): New constraints.
501 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
502 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
503 Use the above on operand 2. Reindent.
504 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
506 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
509 * common/config/i386/i386-common.c (ix86_handle_option): Don't
511 * config/i386/i386.opt (mcet): Removed.
512 * doc/install.texi: Remove -mcet documentation.
513 * doc/invoke.texi: Likewise.
515 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
518 * doc/install.texi: Remove -mcet from bootstrap-cet.
520 2018-04-24 Jakub Jelinek <jakub@redhat.com>
523 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
524 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
528 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
529 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
530 containing a CONST_VECTOR.
532 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
534 * doc/install.texi: Update newlib dependency for nvptx.
536 2018-04-24 Jakub Jelinek <jakub@redhat.com>
539 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
540 instead of INTVAL when shifting x left.
542 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
544 PR tree-optimization/85478
545 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
546 vect_grouped_store_supported for single element vectors.
548 2018-04-24 Richard Biener <rguenther@suse.de>
551 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
552 load cost increase to the case of non-constant step.
554 2018-04-24 Jakub Jelinek <jakub@redhat.com>
557 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
558 destination if any_malformed_asm.
560 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
563 * expr.c (store_field): In the bitfield case, if the value comes from
564 a function call and is returned in registers by means of a PARALLEL,
565 do not change the mode of the temporary unless BLKmode and VOIDmode.
567 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
569 PR rtl-optimization/85423
570 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
571 dependencies to debug insns when the previous insn is non-debug.
573 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
575 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
576 enums into a single definition.
577 (fls): Fix predicates and printing.
580 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
582 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
583 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
584 and short u6 immediate.
585 (check_if_valid_sleep_operand): Remove.
586 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
588 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
590 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
591 flag_always_save_lp condition.
592 * config/nds32/nds32.opt (malways-save-lp): New option.
594 2018-04-22 Shiva Chen <shiva0217@gmail.com>
596 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
597 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
598 * config/nds32/nds32.h
599 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
600 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
602 2018-04-22 Shiva Chen <shiva0217@gmail.com>
604 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
605 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
607 2018-04-22 Shiva Chen <shiva0217@gmail.com>
608 Chung-Ju Wu <jasonwucj@gmail.com>
610 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
612 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
613 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
615 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
617 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
619 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
621 * config/nds32/nds32-protos.h (nds32_data_alignment,
622 nds32_local_alignment): Declare.
623 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
624 nds32_local_alignment): New functions.
625 (TARGET_CONSTANT_ALIGNMENT): Define.
626 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
628 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
630 * config/nds32/nds32.c
631 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
632 (TARGET_MODES_TIEABLE_P): Likewise.
634 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
636 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
639 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
640 Chung-Ju Wu <jasonwucj@gmail.com>
642 * config/nds32/constants.md (unspec_volatile_element): Add enum values
643 for unaligned access.
644 * config/nds32/nds32-intrinsic.c: Implementation of expanding
646 * config/nds32/nds32-intrinsic.md: Likewise.
647 * config/nds32/nds32_intrinsic.h: Likewise.
648 * config/nds32/nds32.h (nds32_builtins): Likewise.
649 * config/nds32/nds32.opt (munaligned-access): New option.
650 * config/nds32/nds32.c (nds32_asm_file_start): Display
651 flag_unaligned_access status.
653 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
655 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
656 -mno-relax is present.
657 * config/riscv/linux.h (LINK_SPEC): Ditto.
659 2018-04-20 Martin Sebor <msebor@redhat.com>
662 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
664 (gimple_fold_builtin_stxcpy_chk): Same.
665 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
667 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
670 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
671 __powikf2 when long double is IEEE 128-bit.
673 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
675 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
676 step to make sure stack always aligned.
678 2018-04-20 Carl Love <cel@us.ibm.com>
681 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
684 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
685 Tom de Vries <tom@codesourcery.com>
688 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
689 Emit insns for calls too.
690 (nvptx_find_par): Always look for worker-level predecessor insn.
691 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
693 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
694 (nvptx_process_pars): Propagate frames for calls.
696 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
699 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
701 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
702 (ix86_handle_option): Don't handle OPT_mibt.
703 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
705 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
707 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
708 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
709 (ix86_target_macros): Define __CET__ with flag_cf_protection
711 * config/i386/i386.c (isa2_opts): Remove -mibt.
712 * config/i386/i386.h (TARGET_IBT): Removed.
713 (TARGET_IBT_P): Likewise.
714 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
715 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
716 * config/i386/i386.opt (mcet): Update help message.
719 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
720 -mcet as an alias for -mshstk.
722 2018-04-20 Richard Biener <rguenther@suse.de>
725 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
726 complexity by forcing a single use of the multiply operand.
728 2018-04-20 Martin Jambor <mjambor@suse.cz>
731 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
732 recursion dependency to only apply to non-clones.
734 2018-04-20 Martin Jambor <mjambor@suse.cz>
737 * ipa-cp.c (create_specialized_node): Check that clones of
738 self-recursive edges exist during IPA-CP.
740 2018-04-19 Toon Moene <toon@moene.org>
742 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
745 2018-04-19 Jakub Jelinek <jakub@redhat.com>
747 PR tree-optimization/85467
748 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
749 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
750 VECTOR_CST element to type.
752 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
755 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
756 * config/i386/i386.md (builtin_setjmp_setup): Removed.
757 (builtin_longjmp): Likewise.
758 (save_stack_nonlocal): New pattern.
759 (restore_stack_nonlocal): Likewise.
761 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
764 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
765 Replace ASM_OUTPUT_LABEL with fprintf.
767 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
770 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
771 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
772 * config/i386/i386-c.c (ix86_target_macros_internal): Also
773 define __IBT__ and __SHSTK__ for -fcf-protection.
774 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
776 (ix86_trampoline_init): Likewise.
777 (x86_output_mi_thunk): Likewise.
778 (ix86_notrack_prefixed_insn_p): Likewise.
779 (ix86_option_override_internal): Don't disallow -fcf-protection.
780 * config/i386/i386.md (rdssp<mode>): Also enable for
782 (incssp<mode>): Likewise.
783 (nop_endbr): Likewise.
784 * config/i386/i386.opt (mcet): Change help message to built-in
788 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
789 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
790 enable CET built-in functions.
792 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
794 * common/config/i386/i386-common.c
795 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
796 OPTION_MASK_ISA_MOVDIRI_UNSET,
797 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
798 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
799 * config.gcc (movdirintrin.h): New header.
800 * config/i386/cpuid.h (bit_MOVDIRI,
801 bit_MOVDIR64B): New bits.
802 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
804 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
805 (VOID, PVOID, PCVOID)): New function types.
806 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
807 __builtin_ia32_directstoreu_u64,
808 __builtin_ia32_movdir64b): New builtins.
809 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
810 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
812 (ix86_valid_target_attribute_inner_p): Ditto.
813 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
814 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
815 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
816 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
817 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
818 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
819 (movdiri<mode>, movdir64b_<mode>): New.
820 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
821 * config/i386/immintrin.h: Include movdirintrin.h.
822 * config/i386/movdirintrin.h: New file.
823 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
825 2018-04-19 Richard Biener <rguenther@suse.de>
828 * cfg.c (clear_bb_flags): When loop state says we have
829 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
831 2018-04-19 Richard Biener <rguenther@suse.de>
833 PR tree-optimization/84737
834 * tree-vect-data-refs.c (vect_copy_ref_info): New function
835 copying restrict info.
836 (vect_setup_realignment): Use it.
837 * tree-vectorizer.h (vect_copy_ref_info): Declare.
838 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
839 the first DR to all generated stores.
840 (vectorizable_load): Likewise for loads.
842 2018-04-19 Jakub Jelinek <jakub@redhat.com>
844 PR tree-optimization/85446
845 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
846 the integral and pointer types to have the same precision.
848 * doc/install.texi: Document --disable-cet being the default and
851 2018-04-18 Martin Liska <mliska@suse.cz>
853 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
856 2018-04-18 Martin Liska <mliska@suse.cz>
859 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
862 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
863 arguments if they are comparable.
865 2018-04-18 Martin Liska <mliska@suse.cz>
868 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
871 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
874 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
877 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
878 ENDBR after calling __morestack.
880 2018-04-18 David Malcolm <dmalcolm@redhat.com>
883 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
884 by using gcc_base_ver to generate a gcc_driver_version, and use
885 it when generating GCC_DRIVER_NAME.
886 * configure: Regenerate.
888 2018-04-18 Jakub Jelinek <jakub@redhat.com>
891 * config.gcc: Obsolete powerpc*-*-*spe*.
893 2018-04-17 Jakub Jelinek <jakub@redhat.com>
896 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
897 (stabstr_D): Change type of unum from unsigned int to
898 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
901 2018-04-17 Jim Wilson <jimw@sifive.com>
904 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
905 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
906 Set arg_pointer_offset after using pretend_args_size.
908 2018-04-17 Jakub Jelinek <jakub@redhat.com>
910 PR rtl-optimization/85431
911 * dse.c (record_store): Ignore zero width stores.
914 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
915 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
916 __builtin_stack_restore rather than after it.
917 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
918 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
919 argument instead of virtual_dynamic_stack_rtx.
921 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
923 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
925 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
926 Add note to error message to explain internal mapping of overloaded
927 built-in function name to non-overloaded built-in function name.
928 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
931 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
934 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
935 where the inputs overlap with the output.
937 2018-04-17 Jakub Jelinek <jakub@redhat.com>
939 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
940 (=v, v) alternative and explicit "memory" attribute.
941 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
942 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
944 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
945 "sselog1" type instead of "sselog".
946 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
947 "sselog". Remove explicit "memory" attribute.
948 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
949 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
951 (vec_extract_hi_v32hi): Merge all alternatives into one, use
952 "sselog1" type instead of "sselog". Remove explicit "memory"
954 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
955 use "sselog1" type instead of "sselog". Remove explicit "memory"
957 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
958 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
960 (vec_extract_hi_v64qi): Merge all alternatives into one, use
961 "sselog1" type instead of "sselog". Remove explicit "memory"
963 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
964 use "sselog1" type instead of "sselog". Remove explicit "memory"
968 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
971 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
974 2018-04-17 Martin Jambor <mjambor@suse.cz>
977 * ipa-cp.c (create_specialized_node): Call
978 expand_all_artificial_thunks if necessary.
980 2018-04-17 Martin Liska <mliska@suse.cz>
983 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
984 in message, remote space in between '_G' and '('.
986 2018-04-17 Jakub Jelinek <jakub@redhat.com>
989 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
990 avx512f_vmcmp<mode>3<round_saeonly_name>,
991 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
992 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
993 avx512f_rndscale<mode><round_saeonly_name>,
994 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
995 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
996 Use %<iptr>2 instead of %2 for -masm=intel.
997 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
998 avx512f_vcvttss2usi<round_saeonly_name>,
999 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1001 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1002 avx512f_vcvttsd2usi<round_saeonly_name>,
1003 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1004 Use %q1 instead of %1 for -masm=intel.
1005 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1006 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1007 of %3 for -masm=intel.
1008 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1010 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1012 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1014 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1015 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1017 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1019 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1020 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1021 %g1 and one with %0 and %1.
1022 (avx512er_vmrcp28<mode><round_saeonly_name>,
1023 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1025 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1026 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1027 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1028 of %0 and %{%4%} for -masm=intel.
1029 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1030 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1031 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1032 order of %0 and %{%5%}%{z%} for -masm=intel.
1034 2018-04-17 Jan Hubicka <jh@suse.cz>
1037 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1039 2018-04-17 Martin Liska <mliska@suse.cz>
1042 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1043 for target_clone error message. Make default implementation
1044 clone to be a local declaration.
1045 (separate_attrs): Add new argument and check for an empty
1047 (expand_target_clones): Handle it.
1048 (ipa_target_clone): Make redirection just for target_clones
1051 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1052 Tom de Vries <tom@codesourcery.com>
1055 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1056 tiled basic blocks without omp continue statements.
1058 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1061 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1062 vec_extract expression as having side effects to make sure it gets
1065 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1068 * config/i386/i386.c (get_builtin_code_for_version): Check
1071 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1074 * gcc/config.gcc: Support "skylake".
1075 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1077 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1078 (processor_target_table): Add "skylake".
1079 (ix86_option_override_internal): Add "skylake".
1080 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1081 PROCESSOR_CANNONLAKE.
1082 (get_builtin_code_for_version): Fix priority for
1083 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1084 PROCESSOR_SKYLAKE-AVX512.
1085 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1086 (processor_type): Add PROCESSOR_SKYLAKE.
1088 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1089 Jason Merrill <jason@redhat.com>
1092 * convert.c (convert_to_integer_1): Use direct recursion for
1093 enumeral types and types with a precision less than the number
1094 of bits in their mode.
1096 2018-04-16 Julia Koval <julia.koval@intel.com>
1099 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1100 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1102 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1105 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1106 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1107 and -mno-direct-move.
1109 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1112 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1113 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1114 Ensure negative shifts result in {0}.
1116 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1118 PR rtl-optimization/79916
1119 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1120 regs (if any) to define how to gnerate SD moves when LRA is in
1123 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1125 PR rtl-optimization/85393
1126 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1127 * except.c (expand_dw2_landing_pad_for_region): Make static.
1128 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1129 a label and unconditional jump to old_bb, rather than
1130 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1133 PR rtl-optimization/85376
1134 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1135 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1136 instead of a specific value.
1138 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1139 Bin Cheng <bin.cheng@arm.com>
1141 PR tree-optimization/82965
1142 PR tree-optimization/83991
1143 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1144 by_profile_only parameter.
1145 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1146 information if the loop was predicted to iterate too many times.
1147 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1149 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1152 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1155 2018-04-13 Martin Liska <mliska@suse.cz>
1156 Jakub Jelinek <jakub@redhat.com>
1159 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1160 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1161 * builtins.c (expand_builtin_memory_copy_args): Use
1162 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1163 handle dest_addr == pc_rtx.
1165 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1168 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1169 asked to not generate direct moves.
1170 (fix_trunc<mode>si2_stfiwx): Similar.
1171 (fix_trunc<mode>si2_internal): Similar.
1173 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1176 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1177 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1178 lookup if dest in some wider mode is known to be const0_rtx and
1179 if so, record permanent equivalence for it to be ZERO_EXTEND of
1180 the narrower mode destination.
1182 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1184 * lto-streamer-out.c (output_function): Revert 259346.
1185 * omp-expand.c (expand_oacc_for): Likewise.
1187 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1189 PR rtl-optimization/85354
1190 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1191 * sel-sched.c (sel_global_init): ... here.
1193 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1196 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1197 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1198 mode for PE-COFF targets.
1199 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1200 (i386_pe_asm_lto_end): Likewise.
1201 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1202 (TARGET_ASM_LTO_END): Likewise.
1203 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1204 (i386_pe_asm_lto_start): New function.
1205 (i386_pe_asm_lto_end): Likewise.
1207 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1208 Richard Biener <rguenther@suse.de>
1211 * lto-streamer-out.c (output_function): Fix CFG loop state before
1213 * omp-expand.c (expand_oacc_for): Handle calls to internal
1214 functions like regular functions.
1216 2018-04-12 Richard Biener <rguenther@suse.de>
1219 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1220 for the early LTO debug to properly generate references to it
1221 during DIE emission. Do not re-use that for the skeleton for
1223 (dwarf2out_early_finish): Likewise.
1225 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1228 * config/i386/sse.md
1229 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1230 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1231 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1232 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1233 and output is a reg, avoid creating invalid lowpart subreg, but
1234 instead split into a 512-bit move. Don't split if not AVX512VL,
1235 input is xmm16+ reg and output is a mem.
1236 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1237 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1238 xmm16+ reg and output is a mem.
1240 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1242 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1243 also for flag_dwarf2_cfi_asm.
1245 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1247 PR rtl-optimization/85342
1248 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1249 a bool scalar var inside of the loop instead. Don't try to update
1250 recog_data.operand after failed apply_change_group.
1252 2018-04-12 Tom de Vries <tom@codesourcery.com>
1255 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1256 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1257 array with flexible array member as array without given dimension.
1258 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1259 argument for undefined param to true.
1261 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1264 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1265 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1266 from PowerPC section.
1267 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1268 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1269 help text that is too long.
1270 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1271 help text that is too long.
1272 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1273 help text that is too long.
1275 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1277 * config/alpha/alpha.md (stack_probe_internal): Rename
1278 from "probe_stack". Update all callers.
1280 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1282 PR rtl-optimization/84566
1283 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1284 sched_macro_fuse_insns.
1286 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1289 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1290 (compute_block_dependences): ... from here.
1292 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1294 PR tree-optimization/85331
1295 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1296 from int to HOST_WIDE_INT.
1298 2018-04-11 Martin Jambor <mjambor@suse.cz>
1301 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1302 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1303 not the same as the source val.
1304 (cgraph_edge_brings_value_p): New parameter.
1305 (gather_edges_for_value): Pass destination value to
1306 cgraph_edge_brings_value_p.
1307 (perhaps_add_new_callers): Likewise.
1308 (get_info_about_necessary_edges): Likewise and exclude values brought
1309 only by self-recursive edges.
1310 (create_specialized_node): Redirect only clones of self-calling edges.
1311 (+self_recursive_pass_through_p): New function.
1312 (find_more_scalar_values_for_callers_subset): Use it.
1313 (find_aggregate_values_for_callers_subset): Likewise.
1314 (known_aggs_to_agg_replacement_list): Removed.
1315 (decide_whether_version_node): Re-calculate known constants for all
1316 remaining context clones.
1318 2018-04-11 Richard Biener <rguenther@suse.de>
1321 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1322 from early DWARF output.
1323 (dwarf2out_early_finish): Output line info unconditionally into
1324 early DWARF and add reference to it.
1326 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1329 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1330 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1331 other than V2DFmode using iptr mode attribute.
1332 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1334 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1336 PR rtl-optimization/84659
1337 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1339 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1342 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1344 (output_loc_list): Pass address of a dummy size variable even in the
1345 locview handling loop.
1346 (index_location_lists): Add comment on why skip_loc_list_entry can't
1349 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1352 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1355 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1358 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1359 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1360 and -mstring-compare-inline-limit.
1362 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1365 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1366 for stack clash protection in a register whenever we need it to be in
1369 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1371 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1372 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1374 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1377 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1379 (mlong-double-): Ditto.
1380 * config/rs6000/sysv4.opt (msdata=): Ditto.
1381 (mtls-size=): Ditto.
1383 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1385 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1386 erroneous entries for
1387 "vector int vec_ldl (int, long int *)", and
1388 "vector unsigned int vec_ldl (int, unsigned long int *)".
1389 Add comments and entries for
1390 "vector bool char vec_ldl (int, bool char *)",
1391 "vector bool short vec_ldl (int, bool short *)",
1392 "vector bool int vec_ldl (int, bool int *)",
1393 "vector bool long long vec_ldl (int, bool long long *)",
1394 "vector pixel vec_ldl (int, pixel *)",
1395 "vector long long vec_ldl (int, long long *)",
1396 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1397 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1398 type tree bool_long_long_type_node and correct definition of
1399 bool_V2DI_type_node to make reference to this new type tree.
1400 (rs6000_mangle_type): Replace erroneous reference to
1401 bool_long_type_node with bool_long_long_type_node.
1402 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1403 comments to emphasize sign distinctions for char and int types and
1404 replace RS6000_BTI_bool_long constant with
1405 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1406 use of RS6000_BTI_pixel.
1407 (bool_long_type_node): Remove this macro definition.
1408 (bool_long_long_type_node): New macro definition
1410 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1412 PR rtl-optimization/85300
1413 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1414 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1415 simplify_unary_operation fails.
1417 2018-04-10 Martin Liska <mliska@suse.cz>
1419 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1420 cgraph_edge and ipa_ref.
1422 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1426 * config/i386/sse.md
1427 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1428 computation of the VEC_MERGE selector from mask.
1429 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1430 Fix decoding of the VEC_MERGE selector into mask.
1432 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1434 PR tree-optimization/85286
1435 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1437 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1439 * final.c (final_1): Set insn_last_address as well as
1440 insn_current_address.
1442 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1445 * explow.c (emit_stack_probe): Call validize_mem on memory location
1446 before passing it to gen_probe_stack. Create address operand and
1447 legitimize it for the probe_stack_address case.
1449 2018-04-09 Jan Hubicka <jh@suse.cz>
1452 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1453 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1454 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1456 (free_lang_data): Rebuild type inheritance graph.
1458 2018-04-09 Martin Sebor <msebor@redhat.com>
1460 * invoke.texi (-finline-small-functions): Mention other optimization
1462 (-findirect-inlining, -fpartial-inlining): Same.
1463 (-finline-functions-called-once): Same.
1464 (-freorder-blocks-and-partition): Same.
1466 2018-04-09 Jan Hubicka <jh@suse.cz>
1469 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1470 jumps; choose last target that matches the criteria (i.e.
1471 no partition changes for non-crossing jumps).
1472 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1473 support for redirecting crossing jumps to non-crossing.
1475 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1477 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1478 also for naked functions.
1480 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1482 * config/arc/arc.md (add_shift): New pattern.
1483 (add_shift2): Likewise.
1484 (sub_shift): Likewise.
1485 (sub_shift_cmp0_noout): Likewise.
1486 (compare_si_ashiftsi): Likewise.
1487 (xbfu_cmp0_noout): New combine pattern.
1488 (xbfu_cmp0"): Likewise.
1489 (movsi_set_cc_insn): Place the predicable variant first.
1490 (commutative_binary_cmp0_noout): Remove clobber.
1491 (commutative_binary_cmp0): New pattern.
1492 (noncommutative_binary_cmp0): Likewise.
1493 (noncommutative_binary_cmp0_noout): Likewise.
1494 (noncommutative_binary_comparison_result_used): Removed.
1495 (rsub_cmp0): New pattern.
1496 (rsub_cmp0_noout): Likewise.
1497 (extzvsi): Changed, keep only meaningful variants.
1498 (SQH, SEZ): New iterators.
1499 (SQH_postfix): New mode attribute.
1500 (SEZ_prefix): New code attribute.
1501 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1502 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1503 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1505 (noncommutative_operator): Check the availability of barrel
1508 2018-04-09 Richard Biener <rguenther@suse.de>
1510 PR tree-optimization/85284
1511 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1512 Only use the niter constraining form of simple_iv when the exit
1515 2018-04-09 Tom de Vries <tom@codesourcery.com>
1518 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1519 (define_expand "*memory_barrier"): New define_expand.
1520 (define_insn "memory_barrier"): New insn.
1522 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1524 PR rtl-optimization/80463
1525 PR rtl-optimization/83972
1526 PR rtl-optimization/83480
1528 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1529 correct producer for the insn.
1530 (tidy_control_flow): Fixup seqnos in case of debug insns.
1532 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1534 PR rtl-optimization/83913
1536 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1537 different sched-times when merging exprs.
1539 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1541 PR rtl-optimization/83962
1543 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1544 tidy_fallthru_edge and tidy_control_flow.
1546 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1548 PR rtl-optimization/83530
1550 * sel-sched.c (force_next_insn): New global variable.
1551 (remove_insn_for_debug): When force_next_insn is true, also leave only
1552 next insn in the ready list.
1553 (sel_sched_region): When the region wasn't scheduled, make another pass
1554 over it with force_next_insn set to 1.
1556 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1558 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1560 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1561 for interrupt control.
1562 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1563 functions for interrupt control.
1564 * config/nds32/nds32-intrinsic.md: Likewise.
1565 * config/nds32/nds32_intrinsic.h: Likewise.
1566 * config/nds32/nds32.h (nds32_builtins): Likewise.
1568 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1570 * config/nds32/nds32.c (nds32_init_machine_status,
1571 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1572 strict_aligned_p field.
1573 (nds32_expand_to_rtl_hook): New function.
1574 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1575 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1577 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1578 Chung-Ju Wu <jasonwucj@gmail.com>
1580 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1581 * config/nds32/nds32-n7.md: New file.
1582 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1583 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1585 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1586 * config/nds32/nds32.md (pipeline_model): Add n7.
1587 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1588 * config/nds32/pipelines.md: Include n7 settings.
1590 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1591 Chung-Ju Wu <jasonwucj@gmail.com>
1593 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1594 * config/nds32/nds32-e8.md: New file.
1595 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1596 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1598 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1599 * config/nds32/nds32.md (pipeline_model): Add e8.
1600 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1601 * config/nds32/pipelines.md: Include e8 settings.
1603 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1604 Chung-Ju Wu <jasonwucj@gmail.com>
1606 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1607 * config/nds32/nds32-n8.md: New file.
1608 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1609 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1611 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1612 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1613 * config/nds32/nds32.md (pipeline_model): Add n8.
1614 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1615 * config/nds32/pipelines.md: Include n8 settings.
1617 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1618 Chung-Ju Wu <jasonwucj@gmail.com>
1620 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1621 * config/nds32/nds32-n9-2r1w.md: New file.
1622 * config/nds32/nds32-n9-3r2w.md: New file.
1623 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1624 nds32_register_ports): New or modify for cpu n9.
1625 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1627 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1628 * config/nds32/nds32-utils.c: New file.
1629 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1630 TARGET_MUL_SLOW): Define.
1631 * config/nds32/nds32.md (pipeline_model): New attribute.
1632 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1633 New options that support cpu n9.
1634 * config/nds32/pipelines.md: Include n9 settings.
1635 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1637 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1639 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1640 information if necessary.
1641 (output_cond_branch_compare_zero): Likewise.
1642 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1643 (nds32_target_alignment): Refine for alignment.
1644 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1645 (FUNCTION_BOUNDARY): Modify.
1646 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1648 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1650 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1652 * config/nds32/constants.md (unspec_volatile_element): Add values for
1653 TLB operation and data prefetch.
1654 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1655 functions for TLB operation and data prefetch.
1656 * config/nds32/nds32-intrinsic.md: Likewise.
1657 * config/nds32/nds32_intrinsic.h: Likewise.
1658 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1659 (nds32_print_operand): Likewise.
1660 * config/nds32/nds32.h (nds32_builtins): Likewise.
1662 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1663 Andrew Pinski <pinsika@gcc.gnu.org>
1666 * match.pd: Use constant_boolean_node of correct type instead of
1667 boolean_true_node or boolean_false_node for simplifying
1668 pointer comparisons to zero.
1670 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1672 PR tree-optimization/80021
1673 * tree.c (verify_type_variant): Make error call in verify_variant_match
1674 translatable and remove final full stop.
1676 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1678 * config/nds32/constants.md (unspec_volatile_element): Add
1679 UNSPEC_VOLATILE_EH_RETURN.
1680 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1681 nds32_output_stack_pop): Support dwarf exception handling process.
1682 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1683 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1684 exception handling process.
1685 (nds32_compute_stack_frame): Likewise.
1686 (nds32_return_addr_rtx): Likewise.
1687 (nds32_initial_elimination_offset): Likewise.
1688 (nds32_expand_prologue): Likewise.
1689 (nds32_expand_epilogue): Likewise.
1690 (nds32_dynamic_chain_address): New function.
1691 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1693 (DYNAMIC_CHAIN_ADDRESS): Define.
1694 (EH_RETURN_DATA_REGNO): Define.
1695 (EH_RETURN_STACKADJ_RTX): Define.
1696 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1697 patterns for dwarf exception handling.
1699 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1701 * config/nds32/nds32.h: Clean up obsolete macros.
1703 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1705 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1706 Add enum values for particular instructions.
1707 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1708 particular intrinsic functions.
1709 * config/nds32/nds32-intrinsic.md: Likewise.
1710 * config/nds32/nds32_intrinsic.h: Likewise.
1711 * config/nds32/nds32.h (nds32_builtins): Likewise.
1712 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1713 (btst, ave): New patterns for particular instructions.
1715 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1717 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1718 Add enum values for atomic load/store and memory sync.
1719 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1721 * config/nds32/nds32-intrinsic.md: Likewise.
1722 * config/nds32/nds32_intrinsic.h: Likewise.
1723 * config/nds32/nds32.h (nds32_builtins): Likewise.
1725 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1727 PR tree-optimization/85257
1728 * fold-const.c (native_encode_vector): If not all elts could fit
1729 and off is -1, return 0 rather than offset.
1730 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1731 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1732 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1733 adjust buffer in native_interpret_expr call.
1735 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1737 * config/nds32/constants.md (unspec_volatile_element): Add cache
1738 control enum values.
1739 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1740 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1741 * config/nds32/nds32.c (nds32_cctl_names): New.
1742 (nds32_print_operand): Handle cache control register names.
1743 * config/nds32/nds32.h (nds32_builtins): New enum values.
1744 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1746 * config/nds32/nds32.md (type): Add mmu.
1747 * config/nds32/pipelines.md (simple_insn): Add mmu.
1749 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1751 * config/nds32/nds32.md (type): Remove call.
1752 * config/nds32/pipelines.md (simple_insn): Likewise.
1754 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1756 * config/nds32/constants.md (unspec_volatile_element): Add
1757 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1758 UNSPEC_VOLATILE_FMFCFG.
1759 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1760 description for fmfcfg and fmfcsr.
1761 (bdesc_1arg): Add fmtcsr.
1762 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1763 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1764 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1765 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1766 unspec_fmfcfg): New patterns.
1767 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1768 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1769 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1770 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1771 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1772 __nds32__fmfcfg): Define.
1774 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1776 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1777 intrinsic register names.
1778 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1779 intrinsic register enum values and macros.
1781 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1783 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1784 for load/store addressing form.
1785 (nds32_print_operand_address): Likewise.
1787 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1790 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1791 based on LABEL_REF. Remove useless assertion.
1792 (pic_address_needs_scratch): Fix formatting.
1793 (sparc_legitimize_pic_address): Minor tweaks.
1794 (sparc_delegitimize_address): Adjust assertion accordingly.
1795 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1796 into symbolic_operand.
1797 (movsi_high_pic_label_ref): Likewise.
1798 (movsi_lo_sum_pic_label_ref): Likewise.
1799 (movdi_pic_label_ref): Likewise.
1800 (movdi_high_pic_label_ref): Likewise.
1801 (movdi_lo_sum_pic_label_ref): Likewise.
1803 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1805 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1806 custom LIB_SPEC setup.
1808 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1809 Kito Cheng <kito.cheng@gmail.com>
1811 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1812 * config/riscv/freebsd.h: New.
1814 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1816 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1817 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1820 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1821 Kito Cheng <kito.cheng@gmail.com>
1823 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1824 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1825 * config/nds32/nds32-protos.h (nds32_output_call,
1826 nds32_output_return): Declare.
1827 * config/nds32/nds32.md: Refine all the call and return patterns.
1829 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1832 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1833 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1835 PR rtl-optimization/84872
1836 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1837 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1840 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1842 * expr.c (copy_blkmode_to_reg): Revert 254862.
1843 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1845 2018-04-06 Richard Biener <rguenther@suse.de>
1848 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1849 after seeing a component reference with an adjacent field. Treat
1850 refs to arrays at struct end of external decls similar to
1851 refs to unconstrained commons.
1853 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1856 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1857 look through SAVE_EXPRs with non-side-effects argument. Adjust
1859 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1862 2018-04-06 Richard Biener <rguenther@suse.de>
1865 * alias.c (find_base_term): New wrapper around find_base_term
1866 unwinding CSELIB_VAL_PTR changes.
1867 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1870 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1872 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1874 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1875 constant definitions.
1876 ("nop"): lr 0,0 -> nopr r0
1877 ("nop_lr0", "nop_lr1"): New insn definitions.
1879 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1881 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1882 NDS32_V3PUSH_AVAILABLE_P macro.
1884 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1885 Chung-Ju Wu <jasonwucj@gmail.com>
1887 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1888 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1889 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1890 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1891 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1892 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1893 * config/nds32/constraints.md: New constraints and checking for hard
1894 float configuration.
1895 * config/nds32/iterators.md: New mode iterator and attribute for hard
1896 float configuration.
1897 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1899 * config/nds32/nds32-fpu.md: New file.
1900 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1901 deal with hard float code generation.
1902 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1904 (abi_type, float_reg_number): New enum type.
1905 * config/nds32/nds32-predicates.c: New predicates for hard float.
1906 * config/nds32/nds32-protos.h: Declare functions for hard float.
1907 * config/nds32/nds32.c: Implementation for hard float configuration.
1908 * config/nds32/nds32.h: Definitions for hard float configuration.
1909 * config/nds32/nds32.md: Include hard float machine description and
1910 modify patterns for hard float configuration.
1911 * config/nds32/nds32.opt: New options for hard float configuration.
1912 * config/nds32/predicates.md: New predicates for hard float
1915 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1917 * common/config/nds32/nds32-common.c
1918 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1920 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1923 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1924 CONSTRUCTOR_ELT (ctor, ...)->value.
1926 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1929 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1931 2018-04-05 Tom de Vries <tom@codesourcery.com>
1934 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1937 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1938 Kito Cheng <kito.cheng@gmail.com>
1940 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1941 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1942 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1943 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1945 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1946 Kito Cheng <kito.cheng@gmail.com>
1948 * config/nds32/constraints.md (Ufe): New memory constraint.
1949 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1950 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1951 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1953 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1954 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1956 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1958 * config/nds32/nds32.md: Use optimize_size in the condition for
1959 alu-shift instructions.
1961 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1963 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1965 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1967 * config/nds32/nds32.md (negsi2): Refine pattern.
1969 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1970 Chung-Ju Wu <jasonwucj@gmail.com>
1972 * config/nds32/iterators.md (shift_rotate): New code iterator.
1973 (shift): New code attribute.
1974 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1975 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1976 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1977 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1978 bit-wise operations.
1979 (andsi3, *andsi3): Ditto.
1980 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1981 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1982 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1983 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1984 nds32_ior_operand, nds32_xor_operand): New predicates.
1986 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1988 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1989 (addsi3, subsi3): ... this.
1991 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1993 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1995 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1997 * config/nds32/nds32.md: Adjust indention.
1999 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2001 * config/nds32/nds32.md (feature): New attribute.
2003 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2005 * config/nds32/nds32.md (subtype): New attribute.
2007 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2010 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2011 expansion to perform a bitwise AND of the argument followed by a
2012 boolean negation of the result.
2014 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2016 PR rtl-optimization/84878
2017 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2018 the basic block. Assert the use reference is not artificial and that
2019 it has an associated insn.
2021 2018-04-04 Michael Matz <matz@suse.de>
2023 * builtins.c (compute_objsize): Pass correct operand
2024 to array_at_struct_end_p.
2026 2018-04-04 Richard Biener <rguenther@suse.de>
2029 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2030 from contexts for DINFO_LEVEL_TERSE and below.
2032 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2034 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2035 resiter_operand condition.
2036 * config/nds32/nds32.md (*move<mode>): Ditto.
2038 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2039 Monk Chiang <sh.chiang04@gmail.com>
2041 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2043 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2045 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2047 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2048 Kito Cheng <kito.cheng@gmail.com>
2050 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2051 nds32_cond_code_str, output_cond_branch,
2052 output_cond_branch_compare_zero, nds32_expand_cbranch,
2053 nds32_expand_cstore, nds32_expand_movcc,
2054 nds32_output_cbranchsi4_equality_zero,
2055 nds32_output_cbranchsi4_equality_reg,
2056 nds32_output_cbranchsi4_equality_reg_or_const_int,
2057 nds32_output_cbranchsi4_greater_less_zero: New functions.
2058 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2059 nds32_expand_cstore, nds32_expand_movcc,
2060 nds32_output_cbranchsi4_equality_zero,
2061 nds32_output_cbranchsi4_equality_reg,
2062 nds32_output_cbranchsi4_equality_reg_or_const_int,
2063 nds32_output_cbranchsi4_greater_less_zero): Declare.
2064 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2065 nds32_rimm11s_operand): New predicates.
2066 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2067 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2070 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2072 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2073 * config/nds32/nds32.md: Ditto.
2074 * config/nds32/pipelines.md: Ditto.
2076 2018-04-04 Richard Biener <rguenther@suse.de>
2078 PR tree-optimization/85168
2079 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2080 propagating abnormals.
2082 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2084 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2086 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2087 Kito Cheng <kito.cheng@gmail.com>
2089 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2090 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2091 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2092 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2093 * config/nds32/nds32.md (sibcall_internal): New.
2094 (sibcall_register): Remove.
2095 (sibcall_immediate): Remove.
2096 (sibcall_value_internal): New.
2097 (sibcall_value_register): Remove.
2098 (sibcall_value_immediate): Remove.
2099 * config/nds32/predicates.md (nds32_general_register_operand): New.
2100 (nds32_call_address_operand): New.
2102 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2104 PR rtl-optimization/85167
2105 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2106 bb_defs if *split_p, instead preinitialize it to NULL.
2108 PR tree-optimization/85156
2109 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2110 evaluating the argument multiple times.
2112 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2114 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2116 (_mm_cvtpd_ps): Likewise.
2117 (_mm_cvttpd_epi32): Likewise.
2118 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2119 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2120 vector, pixel, and bool following altivec.h include.
2122 2018-04-03 Martin Sebor <msebor@redhat.com>
2124 * doc/extend.texi (Common Function Attributes): Clarify.
2125 (const attribute): Likewise.
2126 (pure attribute): Likewise.
2128 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2131 * config/i386/i386.c (ix86_expand_vector_set): Use
2132 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2134 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2136 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2137 instructions when changing rounding bits to preserve precision bits
2138 in the x87 control word.
2140 2018-04-03 Martin Liska <mliska@suse.cz>
2142 PR tree-optimization/82491
2143 * rtl.h (strip_offset_and_add): Replace += suboffset with
2144 poly_uint64 () + suboffset.
2146 2018-03-29 Martin Liska <mliska@suse.cz>
2147 Martin Jambor <mjambor@suse.cz>
2150 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2151 param_type is not an integral or pointer type.
2153 2018-04-03 Richard Biener <rguenther@suse.de>
2155 * sese.h (recompute_all_dominators): Remove.
2157 2018-04-02 Martin Sebor <msebor@redhat.com>
2159 * doc/invoke.texi (-Wrestrict): Fix typos.
2161 2018-04-02 Jim Wilson <jimw@sifive.com>
2163 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2164 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2165 (<optab>di3, <optab>si3_extend): Likewise.
2166 (<optab>si3_mask, <optab>si3_mask_1): New.
2167 (<optab>di3_mask, <optab>di3_mask_1): New.
2168 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2169 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2170 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2172 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2174 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2177 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2179 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2180 (nds32_canonicalize_comparison): New function.
2182 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2183 Kito Cheng <kito.cheng@gmail.com>
2184 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2186 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2187 * config/nds32/constants.md (unspec_volatile_element): Add
2188 UNSPEC_VOLATILE_RELAX_GROUP.
2189 * config/nds32/nds32-relax-opt.c: New file.
2190 * config/nds32/nds32-predicates.c
2191 (nds32_symbol_load_store_p): New function.
2192 * config/nds32/nds32-protos.h
2193 (nds32_symbol_load_store_p): Declare function.
2194 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2195 * config/nds32/nds32.c
2196 (nds32_register_pass): New function to register pass.
2197 (nds32_register_passes): New function to register passes.
2198 * config/nds32/nds32.md (relax_group): New pattern.
2199 * config/nds32/nds32.opt (mrelax-hint): New option.
2200 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2202 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2204 * config/nds32/t-nds32: Modify files dependency.
2206 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2208 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2209 (PROFILE_HOOK): Define its implementation.
2211 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2213 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2214 type and 32-bit size.
2216 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2219 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2220 (V_128_256): New mode iterator.
2221 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2222 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2223 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2225 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2226 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2228 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2231 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2232 NaN inputs correctly.
2234 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2237 * config/rs6000/vsx.md (??r): New mode attribute.
2238 (*vsx_mov<mode>_64bit): Use it.
2239 (*vsx_mov<mode>_32bit): Likewise.
2241 2018-03-30 Martin Sebor <msebor@redhat.com>
2243 PR tree-optimization/84818
2244 * builtins.c (check_access): Use warning_n.
2246 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2249 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2251 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2254 2018-03-30 Julia Koval <julia.koval@intel.com>
2257 * x86-tune.def (movx, partial_reg_dependency): Enable for
2260 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2263 * lra-constraints.c (process_alt_operands): Move setting
2264 this_alternative_matches below.
2266 2018-03-29 Martin Liska <mliska@suse.cz>
2269 * doc/invoke.texi: Document how LTO works with debug info.
2270 Describe auto-load support of binutils. Mention 'x86-64'
2271 as valid option value of -march option.
2273 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2275 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2278 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2279 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2280 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2283 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2286 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2287 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2288 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2289 for RS6000_BTM_POWERPC64.
2290 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2291 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2292 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2297 2018-03-28 Carl Love <cel@us.ibm.com>
2300 2017-09-27 Carl Love <cel@us.ibm.com>
2302 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2303 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2304 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2307 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2309 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2310 instead of __vector bool.
2311 (_mm_max_pu8): Likewise.
2312 (_mm_min_pi16): Likewise.
2314 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2317 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2318 (DIVWEUO): Likewise.
2320 (DIVDEUO): Likewise.
2321 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2322 DIVWEUO and DIVDEUO.
2323 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2324 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2325 (div_extend): Likewise.
2326 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2328 (__builtin_divweuo): Likewise.
2329 (__builtin_divdeo): Likewise.
2330 (__builtin_divdeuo): Likewise.
2332 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2335 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2336 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2338 PR tree-optimization/82004
2339 * gimple-match-head.c (optimize_pow_to_exp): New function.
2340 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2341 Don't fold to exp if optimize_pow_to_exp is false.
2343 2018-03-28 Martin Liska <mliska@suse.cz>
2346 * calls.c (initialize_argument_information): Fix trailing space.
2347 * common.opt: Fix typo and provide better explanation for
2348 -fsanitize-coverage option.
2349 * config/i386/i386.opt: Fix typo.
2351 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2352 Martin Liska <mliska@suse.cz>
2355 * gimplify.c (asan_poison_variable): Don't do the check for
2356 gimplify_omp_ctxp here.
2357 (gimplify_decl_expr): Do it here.
2358 (gimplify_target_expr): Likewise.
2360 2018-03-28 Martin Liska <mliska@suse.cz>
2363 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2364 chkp_type_bounds_count if MPX is not enabled.
2366 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2368 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2370 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2373 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2374 function to create the function decl for complex long double
2375 multiply and divide for -mabi=ieeelongdouble.
2376 (init_float128_ieee): Call it.
2378 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2381 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2382 -fcf-protection=branch -mibt.
2383 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2385 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2388 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2390 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2393 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2394 extern array declarations.
2396 2018-03-27 Richard Biener <rguenther@suse.de>
2399 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2400 explicit single_use checks.
2402 2018-03-27 Richard Biener <rguenther@suse.de>
2404 PR tree-optimization/85082
2405 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2408 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2410 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2411 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2412 Turn on fasynchronous-unwind-tables and funwind-tables.
2414 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2417 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2418 (*bmi_blsr_<mode>_ccz): Ditto.
2420 2018-03-26 Tom de Vries <tom@codesourcery.com>
2422 PR tree-optimization/85063
2423 * omp-general.c (offloading_function_p): New function. Factor out
2425 * omp-offload.c (pass_omp_target_link::gate): ... here.
2426 * omp-general.h (offloading_function_p): Declare.
2427 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2428 with attribute omp declare target for offloading functions.
2430 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2432 PR tree-optimization/84005
2433 * tree-data-ref.h (get_base_for_alignment): Declare.
2434 * tree-data-ref.c (get_base_for_alignment_1): New function.
2435 (get_base_for_alignment): Likewise.
2436 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2437 get_base_for_alignment to find a suitable base object, instead
2438 of always using drb->base_address.
2440 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2443 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2444 known size by default.
2446 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2449 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2450 and non BLKmode operands.
2452 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2455 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2456 Clean up attributes.
2458 2018-03-23 Richard Biener <rguenther@suse.de>
2461 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2462 we are going to emit early debug for LTO.
2464 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2467 * function.c (match_asm_constraints_1): Don't optimize if input
2468 doesn't satisfy general_operand predicate for output's mode.
2471 * alias.c (write_dependence_p): Don't require for x_canonicalized
2472 non-VOIDmode if x has VOIDmode.
2475 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2476 just don't try to optimize it rather than assert it never happens.
2478 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2480 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2481 macro expansions for definition of ST_INTERNAL_<mode> and
2482 LD_INTERNAL_<mode> builtins.
2483 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2485 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2487 (altivec_expand_st_builtin): Likewise.
2488 (altivec_expand_builtin): Remove calls to deleted functions.
2489 (rs6000_address_for_altivec): Delete this function.
2490 * config/rs6000/vector.md: Remove expands for
2491 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2493 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2496 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2497 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2498 re-computing once computed.
2499 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2500 (arm_init_machine_status): Initialize
2501 machine->static_chain_stack_bytes.
2503 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2506 * doc/extend.texi: Add four new prototypes for vec_ld.
2507 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2508 definitions for more logical presentation.
2509 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2510 entries for V1TI variants of __builtin_altivec_ld builtin.
2511 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2512 handling of V1TI variant of LVX icode pattern.
2513 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2514 (rs6000_gimple_fold_builtin): Likewise.
2515 (altivec_init_builtins): Add code to define
2516 __builtin_altivec_lvx_v1ti function.
2518 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2521 * function.c (match_asm_constraints_1): Don't do the optimization
2522 if input isn't a REG, SUBREG, MEM or constant.
2524 2018-03-22 Tom de Vries <tom@codesourcery.com>
2526 PR tree-optimization/84956
2527 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2528 bb_has_abnormal_pred.
2530 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2533 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2534 DECL_INITIAL (decl) to decl at the end.
2535 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2538 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2540 * doc/extend.texi (__builtin_tgmath): Document when complex
2541 integer types are treated as _Complex _Float64.
2543 2018-03-21 Tom de Vries <tom@codesourcery.com>
2545 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2547 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2549 PR tree-optimization/84960
2550 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2551 if it is ENTRY block, move them into single succ of ENTRY in that case.
2553 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2555 PR tree-optimization/84811
2556 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2557 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2558 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2559 (known_subrange_p): Update accordingly. Cast each value involved
2560 in the size comparison, rather than casting the result of the
2563 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2565 PR tree-optimization/84982
2566 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2567 by flipping the least significant bit rather than all bits from
2568 bitpos to bitpos + bitsize - 1.
2570 2018-03-21 Nathan Sidwell <nathan@acm.org>
2572 * doc/extend.texi (Deprecated Features): Remove mention of
2573 long-deleted deprecations.
2575 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2578 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2579 * configure: Regenerate.
2581 2018-03-21 Tom de Vries <tom@codesourcery.com>
2583 PR tree-optimization/83126
2584 * tree-parloops.c (num_phis): New function.
2585 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2587 2018-03-21 Nathan Sidwell <nathan@acm.org>
2589 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2590 mention anon-struct/union members and trailing attributes.
2592 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2594 PR tree-optimization/84969
2595 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2596 builtin memset partitions if they set different rhs values.
2598 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2600 PR rtl-optimization/84989
2601 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2602 VEC_DUPLICATE with scalar result mode.
2604 2018-03-21 Martin Liska <mliska@suse.cz>
2607 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2608 not intended return statement.
2610 2018-03-21 Martin Liska <mliska@suse.cz>
2613 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2614 (chkp_find_bound_slots_1): Limit number of iterations.
2616 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2619 * Minor grammar fixes for x86 options.
2621 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2624 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2625 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2627 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2630 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2631 (altivec_lvx_<mode>_1op): Likewise.
2632 (altivec_stvx_<mode>_2op): Likewise.
2633 (altivec_stvx_<mode>_1op): Likewise.
2634 (altivec_lvx_<VM2:mode>): New define_expand.
2635 (altivec_stvx_<VM2:mode>): Likewise.
2636 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2637 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2638 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2639 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2640 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2641 (rs6000_gen_lvx): Likewise.
2642 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2643 (altivec_expand_stv_builtin): Likewise.
2644 (altivec_expand_builtin): Likewise.
2645 * config/rs6000/vector.md: Likewise.
2647 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2650 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2653 2018-03-20 Richard Biener <rguenther@suse.de>
2656 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2657 sign-conversions as zero, fall back to standard scalar_stmt
2660 2018-03-20 Martin Liska <mliska@suse.cz>
2663 * predict.c (rebuild_frequencies): Handle case when we have
2664 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2666 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2669 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2670 flag_section_anchors.
2671 * varasm.c (use_blocks_for_decl_p): Remove hack for
2672 dw2_force_const_mem.
2675 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2677 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2678 be created, use lowpart_subreg of operands[0] rather than operands[0]
2680 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2681 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2682 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2683 and n constraint instead of aarch64_shift_imm_di and Usd.
2684 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2685 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2687 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2690 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2691 to favor GPR over NEON registers.
2692 (<shift>di3_neon): Likewise.
2694 2018-03-20 Tom de Vries <tom@codesourcery.com>
2697 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2698 (nvptx_process_pars): Emit bar.sync asap and alap.
2700 2018-03-20 Tom de Vries <tom@codesourcery.com>
2703 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2704 seen_label if seen_label is already set.
2706 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2709 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2710 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2711 Use 1U instead of 1. Formatting fixes.
2714 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2715 instead of TREE_TYPE (s1) for the return value.
2717 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2719 PR tree-optimization/84946
2720 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2721 bitsize + bitsize in poly_uint64 rather than poly_int64.
2724 * dwarf2asm.c: Include fold-const.c.
2725 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2726 of decl rather than decl itself.
2728 PR rtl-optimization/84643
2729 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2731 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2734 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2735 calling assemble_variable.
2737 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2740 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2741 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2743 2018-03-19 Jim Wilson <jimw@sifive.com>
2746 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2747 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2748 (riscv_first_stack_step): Likewise.
2749 (riscv_option_override): Use STACK_BOUNDARY instead of
2751 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2753 (BIGGEST_ALIGNMENT): Set to 128.
2754 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2755 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2758 2018-03-19 Richard Biener <rguenther@suse.de>
2760 PR tree-optimization/84933
2761 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2762 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2764 2018-03-19 Richard Biener <rguenther@suse.de>
2766 PR tree-optimization/84859
2767 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2768 (cond_if_else_store_replacement): Perform sinking operation on
2769 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2770 Generalize what a BB with a single eligible store is.
2772 2018-03-19 Richard Biener <rguenther@suse.de>
2774 PR tree-optimization/84929
2775 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2776 chrec_is_positive against non-chrec arg.
2778 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2781 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2783 2018-03-18 Martin Liska <mliska@suse.cz>
2785 PR rtl-optimization/84635
2786 * regrename.c (build_def_use): Use matches_mode only when
2789 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2791 PR tree-optimization/84913
2792 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2793 vectorize chains of COND_EXPRs.
2795 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2797 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2799 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2801 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2803 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2805 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2807 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2808 Kito Cheng <kito.cheng@gmail.com>
2810 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2811 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2812 (nds32_adjust_reg_alloc_order): New function.
2813 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2815 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2817 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2818 nds32_print_operand, nds32_print_operand_address): Use
2819 HOST_WIDE_INT_PRINT_DEC instead.
2821 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2823 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2825 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2828 * config/i386/i386.c (initial_ix86_tune_features,
2829 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2831 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2832 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2833 rather than 1u << ix86_tune. Formatting fix.
2834 (ix86_option_override_internal): Change ix86_arch_mask from
2835 unsigned int to unsigned HOST_WIDE_INT, initialize to
2836 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2837 (ix86_function_specific_restore): Likewise.
2839 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2842 * postreload.c (reload_combine_recognize_pattern): Perform
2843 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2844 truncate_int_for_mode the result for the destination's mode.
2847 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2848 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2850 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2853 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2856 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2857 regno_allocno_class_array and sorted_pseudos.
2858 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2859 insns where regno is used.
2861 2018-03-16 Martin Liska <mliska@suse.cz>
2864 * multiple_target.c (create_dispatcher_calls): Redirect
2865 reference in the symbol table.
2867 2018-03-16 Martin Liska <mliska@suse.cz>
2870 * multiple_target.c (create_dispatcher_calls): Redirect also
2873 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2877 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2878 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2879 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2881 2018-03-16 Julia Koval <julia.koval@intel.com>
2883 * doc/invoke.texi (Skylake Server): Add CLWB.
2884 Cannonlake): Remove CLWB.
2886 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2888 PR tree-optimization/84841
2889 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2891 (FLOAT_ONE_CONST_TYPE): Define.
2892 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2893 (sort_by_operand_rank): Put entries with higher constant_type last
2894 rather than first to match comments.
2896 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2898 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2901 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2904 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2905 check and use address of target always.
2907 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2910 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2911 (indirect_thunk_bnd_needed): Likewise.
2912 (indirect_thunks_used): Likewise.
2913 (indirect_thunks_bnd_used): Likewise.
2914 (indirect_return_needed): New.
2915 (indirect_return_bnd_needed): Likewise.
2916 (output_indirect_thunk_function): Add a bool argument for
2918 (output_indirect_thunk_function): Don't generate alias for
2919 function return thunk.
2920 (ix86_code_end): Call output_indirect_thunk_function to generate
2921 function return thunks.
2922 (ix86_output_function_return): Set indirect_return_bnd_needed
2923 and indirect_return_needed instead of indirect_thunk_bnd_needed
2924 and indirect_thunk_needed.
2926 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2928 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2929 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2930 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2932 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2933 Paul Hua <paul.hua.gm@gmail.com>
2936 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2938 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2940 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2941 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2944 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2947 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2948 instead of GET_MODE_SIZE when comparing Units.
2950 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2953 * varasm.c (hash_section): Return an unchangeble hash value
2954 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2955 Return !aarch64_can_use_per_function_literal_pools_p ().
2957 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2960 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2961 rather than address of cmode as last argument to prepare_cmp_insn.
2963 2018-03-15 Julia Koval <julia.koval@intel.com>
2965 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2966 F_AVX512VNNI, F_AVX512BITALG): New.
2968 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2971 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2972 insn for floating-point loads and stores.
2974 2018-03-14 Carl Love <cel@us.ibm.com>
2976 * config/rs6000/rs6000-c.c: Add macro definitions for
2977 ALTIVEC_BUILTIN_VEC_PERMXOR.
2978 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2979 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2980 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2981 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2983 * config/doc/extend.texi: Add prototypes for vec_permxor.
2985 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2988 * diagnostic-show-locus.c (class layout_point): Convert m_line
2989 from int to linenum_type.
2990 (line_span::comparator): Use linenum "compare" function when
2991 comparing line numbers.
2992 (test_line_span): New function.
2993 (layout_range::contains_point): Convert param "row" from int to
2995 (layout_range::intersects_line_p): Likewise.
2996 (layout::will_show_line_p): Likewise.
2997 (layout::print_source_line): Likewise.
2998 (layout::should_print_annotation_line_p): Likewise.
2999 (layout::print_annotation_line): Likewise.
3000 (layout::print_leading_fixits): Likewise.
3001 (layout::annotation_line_showed_range_p): Likewise.
3002 (struct line_corrections): Likewise for field m_row.
3003 (line_corrections::line_corrections): Likewise for param "row".
3004 (layout::print_trailing_fixits): Likewise.
3005 (layout::get_state_at_point): Likewise.
3006 (layout::get_x_bound_for_row): Likewise.
3007 (layout::print_line): Likewise.
3008 (diagnostic_show_locus): Likewise for locals "last_line" and
3010 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3011 * input.c (selftest::test_linenum_comparisons): New function.
3012 (selftest::input_c_tests): Call it.
3013 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3014 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3015 * selftest.h (ASSERT_GT): New macro.
3016 (ASSERT_GT_AT): New macro.
3017 (ASSERT_LT): New macro.
3018 (ASSERT_LT_AT): New macro.
3020 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3022 PR rtl-optimization/84780
3023 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3025 2018-03-14 Martin Liska <mliska@suse.cz>
3027 * tree.c (record_node_allocation_statistics): Use
3028 get_stats_node_kind.
3029 (get_stats_node_kind): New function extracted from
3030 record_node_allocation_statistics.
3031 (free_node): Use get_stats_node_kind.
3033 2018-03-14 Richard Biener <rguenther@suse.de>
3035 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3036 that the value-set of ANTIC_IN doesn't grow.
3039 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3041 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3042 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3044 2018-03-14 Julia Koval <julia.koval@intel.com>
3046 * config.gcc (icelake-client, icelake-server): New.
3048 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3049 (initial_ix86_arch_features): Ditto.
3050 (PTA_SKYLAKE): Add SGX.
3051 (PTA_ICELAKE): Remove.
3052 (PTA_ICELAKE_CLIENT): New.
3053 (PTA_ICELAKE_SERVER): New.
3054 (ix86_option_override_internal): Split up icelake on icelake client and
3056 (get_builtin_code_for_version): Ditto.
3057 (fold_builtin_cpu): Ditto.
3058 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3059 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3060 * config/i386/i386.h (processor_type): Ditto.
3061 * doc/invoke.texi: Ditto.
3063 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3066 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3067 INTEGER_CST offset, add it together with bitpos / 8 and
3068 sign extend based on POINTER_SIZE.
3072 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3075 * config/i386/constraints.md (Yc): New register constraint.
3076 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3077 Use Yc constraint for alternative 2 of operand 0. Remove
3078 preferred_for_speed attribute.
3080 2018-03-14 Richard Biener <rguenther@suse.de>
3082 PR tree-optimization/84830
3083 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3084 with the old one to avoid oscillations.
3086 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3089 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3091 (assign_by_spills): Return a flag of reload assignment failure.
3092 Do not process the reload assignment failures. Do not spill other
3093 reload pseudos if they has the same reg class. Update n if
3095 (lra_assign): Add a return arg. Set up from the result of
3096 assign_by_spills call.
3097 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3098 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3099 usage_insns if it is not NULL.
3100 (spill_hard_reg_in_range): New function.
3101 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3102 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3103 function prototypes.
3104 (lra_assign): Change prototype.
3105 * lra.c (lra): Add code to deal with fails by splitting hard reg
3108 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3110 * config/riscv/riscv.opt (mrelax): New option.
3111 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3112 "norelax" when riscv_mrelax is disabled.
3113 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3115 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3118 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3119 reassociation for int modes.
3121 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3123 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3124 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3126 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3127 * config/aarch64/aarch64-sve.md
3128 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3129 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3130 (*extend<mode><Vwide>2): Rename to...
3131 (aarch64_sve_extend<mode><Vwide>2): ...this.
3132 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3133 renaming the old pattern to...
3134 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3136 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3137 define_expand, renaming the old pattern to...
3138 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3139 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3140 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3141 account when deciding which SVE instruction the optab should use.
3142 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3144 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3146 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3147 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3148 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3149 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3150 (tlsdesc_small_<mode>): Turn a define_expand and use
3151 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3152 (tlsdesc_small_advsimd_<mode>): ...this.
3153 (tlsdesc_small_sve_<mode>): New pattern.
3155 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3157 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3158 (UNSPEC_UMUL_HIGHPART): New constants.
3159 (MUL_HIGHPART): New int iteraor.
3160 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3161 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3163 (*<su>mul<mode>3_highpart): New define_insn.
3165 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3168 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3171 2018-03-13 Martin Liska <mliska@suse.cz>
3174 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3176 (sem_item_optimizer::~sem_item_optimizer): Release it.
3177 (sem_item_optimizer::merge_classes): Register variable aliases.
3178 (sem_item_optimizer::fixup_pt_set): New function.
3179 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3180 * ipa-icf.h: Declare new variables and functions.
3182 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3185 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3186 integer_pow2p@2 and test integer_pow2p in condition.
3187 (A < 0 ? C : 0): Similarly for @1.
3190 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3191 characters starting at p contain '\0' character, don't look beyond
3195 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3196 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3199 * reg-stack.c (change_stack): Change update_end var from int to
3200 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3201 also call set_block_for_insn on the newly added insns and rescan.
3204 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3205 on the last operand.
3208 * tree.c (stabilize_reference_1): Return save_expr (e) for
3209 STATEMENT_LIST even if it doesn't have side-effects.
3211 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3213 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3215 2018-03-12 Renlin Li <renlin.li@arm.com>
3217 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3218 aarch64_output_scalar_simd_mov_immediate.
3220 2018-03-12 Martin Sebor <msebor@redhat.com>
3222 PR tree-optimization/83456
3223 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3224 for perfectly overlapping calls to memcpy.
3225 (gimple_fold_builtin_memory_chk): Same.
3226 (gimple_fold_builtin_strcpy): Handle no-warning.
3227 (gimple_fold_builtin_stxcpy_chk): Same.
3228 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3230 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3232 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3233 parameter. Use it for SFmode.
3234 (rs6000_function_arg_advance_1): Adjust.
3235 (rs6000_function_arg): Adjust.
3236 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3238 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3240 PR rtl-optimization/84169
3241 PR rtl-optimization/84780
3242 * combine.c (can_combine_p): Check for a 2-insn combination whether
3243 the destination register is used between the two insns, too.
3245 2018-03-12 Richard Biener <rguenther@suse.de>
3247 PR tree-optimization/84803
3248 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3249 for refs DR analysis didn't process.
3251 2018-03-12 Richard Biener <rguenther@suse.de>
3253 PR tree-optimization/84777
3254 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3255 force-vectorize loops ignore whether we are optimizing for size.
3257 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3259 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3260 (TARGET_MD_ASM_ADJUST): Define.
3262 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3263 Kito Cheng <kito.cheng@gmail.com>
3264 Chung-Ju Wu <jasonwucj@gmail.com>
3266 * config/nds32/nds32.c (nds32_compute_stack_frame,
3267 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3268 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3269 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3270 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3271 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3272 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3273 * config/nds32/nds32.md (prologue, epilogue): Use macro
3274 NDS32_V3PUSH_AVAILABLE_P to do checking.
3276 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3279 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3280 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3281 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3282 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3283 addition of most attributes on !orig_type_die or the attribute not
3284 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3286 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3287 Chung-Ju Wu <jasonwucj@gmail.com>
3289 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3291 * config/nds32/nds32.opt (mvh): New option.
3293 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3294 Chung-Ju Wu <jasonwucj@gmail.com>
3296 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3298 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3299 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3302 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3303 Chung-Ju Wu <jasonwucj@gmail.com>
3305 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3307 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3308 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3310 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3311 Kito Cheng <kito.cheng@gmail.com>
3312 Chung-Ju Wu <jasonwucj@gmail.com>
3314 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3315 UNSPEC_FFMISM and UNSPEC_FLMISM.
3316 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3317 for ffb, ffmism and flmism.
3318 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3319 (unspec_ffmism): Ditto.
3320 (unspec_flmism): Ditto.
3321 (nds32_expand_builtin_impl): Check if string extension is available.
3322 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3323 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3325 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3328 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3331 * lra-assigns.c (assign_by_spills): Return a flag of reload
3332 assignment failure. Do not process the reload assignment
3333 failures. Do not spill other reload pseudos if they has the same
3335 (lra_assign): Add a return arg. Set up from the result of
3336 assign_by_spills call.
3337 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3338 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3339 usage_insns if it is not NULL.
3340 (spill_hard_reg_in_range): New function.
3341 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3342 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3343 function prototypes.
3344 (lra_assign): Change prototype.
3345 * lra.c (lra): Add code to deal with fails by splitting hard reg
3348 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3351 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3353 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3356 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3357 (dw_val_node): Add val_symbolic_view.
3358 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3359 (symview_upper_bound): New.
3360 (new_line_info_table): Initialize symviews_since_reset.
3361 (dwarf2out_source_line): Count symviews_since_reset and set
3362 symview_upper_bound.
3363 (dw_val_equal_p): Handle symview.
3364 (add_AT_symview): New.
3365 (print_dw_val): Handle symview.
3366 (attr_checksum, attr_checksum_ordered): Likewise.
3367 (same_dw_val_p, size_of_die): Likewise.
3368 (value_format, output_die): Likewise.
3369 (add_high_low_attributes): Use add_AT_symview for entry_view.
3370 (dwarf2out_finish): Reset symview_upper_bound, clear
3373 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3376 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3377 Add strict argument and use it.
3378 (rs6000_split_multireg_move): Update for new strict argument.
3379 (mem_operand_gpr): Disallow all non-offsettable addresses.
3380 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3382 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3385 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3386 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3387 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3390 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3391 decl, use remap_type if we want to use the type.
3393 2018-03-09 Martin Sebor <msebor@redhat.com>
3395 PR tree-optimization/84526
3396 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3398 (builtin_access::generic_overlap): Be prepared to handle non-array
3401 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3403 PR rtl-optimization/84682
3404 * lra-constraints.c (process_address_1): Check is_address flag
3405 for address constraints.
3406 (process_alt_operands): Likewise.
3407 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3408 preprocess_constraints.
3409 * recog.h (preprocess_constraints): Add oploc parameter.
3411 * recog.c (preprocess_constraints): Test address_operand for
3412 CT_ADDRESS constraints.
3414 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3417 * lra-assigns.c (assign_by_spills): Return a flag of reload
3418 assignment failure. Do not process the reload assignment
3419 failures. Do not spill other reload pseudos if they has the same
3421 (lra_assign): Add a return arg. Set up from the result of
3422 assign_by_spills call.
3423 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3424 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3425 usage_insns if it is not NULL.
3426 (spill_hard_reg_in_range): New function.
3427 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3428 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3429 function prototypes.
3430 (lra_assign): Change prototype.
3431 * lra.c (lra): Add code to deal with fails by splitting hard reg
3434 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3437 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3438 Accept complain bool parameter. Only emit errors if it is true.
3439 (arm_parse_cpu_option_name): Likewise.
3440 (arm_target_thumb_only): Adjust callers of the above.
3441 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3442 prototype to take a default true bool parameter.
3443 (arm_parse_arch_option_name): Likewise.
3445 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3446 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3450 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3451 * configure: Regenerate.
3452 * configure.ac ("linker --version-script option"): New.
3453 ("linker soname option"): New.
3455 2018-03-09 Richard Biener <rguenther@suse.de>
3457 PR tree-optimization/84775
3458 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3459 immediate uses of predicate stmts and mark them modified.
3462 PR tree-optimization/84178
3463 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3465 (version_loop_for_if_conversion): Delay update_ssa call.
3466 (tree_if_conversion): Delay update_ssa until after predicate
3469 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3472 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3473 when the function accesses prior frames.
3475 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3478 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3479 gen_llsym, otherwise call maybe_gen_llsym.
3482 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3483 has ',' character inside of it.
3485 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3488 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3489 as clobbering CC_REGNUM.
3491 2018-03-08 Richard Biener <rguenther@suse.de>
3494 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3495 (follow_copies_to_constant): Do not follow SSA names registered
3498 2018-03-08 Richard Biener <rguenther@suse.de>
3500 PR tree-optimization/84178
3501 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3503 (version_loop_for_if_conversion): Delay update_ssa call.
3504 (tree_if_conversion): Delay update_ssa until after predicate
3507 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3509 PR tree-optimization/84178
3510 * tree-if-conv.c (release_bb_predicate): Remove the
3511 the assertion that the stmts have NULL use_ops.
3512 Discard the statements, asserting that they haven't
3513 yet been added to a BB.
3515 2018-03-08 Richard Biener <rguenther@suse.de>
3517 PR tree-optimization/84746
3518 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3519 (phi_translate): Pass in destination ANTIC_OUT set.
3520 (phi_translate_1): Likewise. For a simplified result lookup
3521 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3522 (phi_translate_set): Adjust.
3523 (do_pre_regular_insertion): Likewise.
3524 (do_pre_partial_partial_insertion): Likewise.
3526 2018-03-08 Martin Liska <mliska@suse.cz>
3528 PR gcov-profile/84735
3529 * doc/gcov.texi: Document usage of profile files.
3530 * gcov-io.h: Document changes in the format.
3532 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3536 * dwarf2out.c (struct dw_line_info_table): Update comments for
3538 (FORCE_RESET_NEXT_VIEW): New.
3539 (FORCE_RESETTING_VIEW_P): New.
3540 (RESETTING_VIEW_P): Check for -1 too.
3541 (ZERO_VIEW_P): Likewise.
3542 (new_line_info_table): Force-reset next view.
3543 (dwarf2out_begin_function): Likewise.
3544 (dwarf2out_source_line): Simplify zero_view_p initialization.
3545 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3546 view directly. Omit view when omitting .loc at line 0.
3548 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3550 PR tree-optimization/84740
3551 * tree-switch-conversion.c (process_switch): Call build_constructors
3552 only if info.phi_count is non-zero.
3554 PR tree-optimization/84739
3555 * tree-tailcall.c (find_tail_calls): Check call arguments against
3556 DECL_ARGUMENTS (current_function_decl) rather than
3557 DECL_ARGUMENTS (func) when checking for tail recursion.
3559 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3561 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3562 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3563 Volker Reichelt's entry and add entries for people that perform
3564 GCC fuzzy testing and report numerous bugs.
3566 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3569 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3570 readonly data in sdata, if that is disabled.
3571 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3572 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3573 -mreadonly-in-sdata option.
3575 2018-03-07 Martin Sebor <msebor@redhat.com>
3577 PR tree-optimization/84468
3578 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3579 basic block when looking for nul assignment.
3581 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3584 * except.h (output_function_exception_table): Adjust prototype.
3585 * except.c (output_function_exception_table): Remove FNNAME parameter
3586 and add SECTION parameter. Ouput one part of the table at a time.
3587 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3588 the first part of the exception table and emit unwind directives.
3589 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3590 (i386_pe_seh_cold_init): Likewise.
3591 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3592 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3593 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3594 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3595 * config/i386/winnt.c: Include except.h.
3596 (struct seh_frame_state): Add reg_offset, after_prologue and
3597 in_cold_section fields.
3598 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3599 (i386_pe_seh_cold_init): New function.
3600 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3601 to seh->in_cold_section.
3602 (seh_emit_push): Record the offset of the push.
3603 (seh_emit_save): Record the offet of the save.
3604 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3605 Test seh->after_prologue to disregard the epilogue.
3606 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3607 (i386_pe_end_cold_function): New function.
3609 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3612 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3613 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3616 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3618 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3619 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3622 * multiple_target.c: Include tree-inline.h and intl.h.
3623 (expand_target_clones): Diagnose and fail if node->definition and
3624 !tree_versionable_function_p (node->decl).
3626 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3628 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3630 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3631 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3632 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3634 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3637 * combine.c (try_combine): Use reg_or_subregno instead of handling
3638 just paradoxical SUBREGs and REGs.
3640 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3642 * config/arc/arc.c (arc_finalize_pic): Remove function.
3643 (arc_must_save_register): We use single base PIC register, remove
3644 checks to save/restore the PIC register.
3645 (arc_expand_prologue): Likewise.
3646 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3648 (arc_verify_short): Likewise.
3649 (arc_attr_type): Likewise.
3650 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3651 (walk_stores): Likewise.
3652 (arc_address_cost): Make it static.
3653 (arc_verify_short): Likewise.
3654 (branch_dest): Likewise.
3655 (arc_attr_type): Likewise.
3656 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3657 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3658 (arc_final_prescan_insn): Remove inserting the nops due to
3659 hardware hazards. It is done in reorg step.
3660 (insn_length_variant_t): Remove.
3661 (insn_length_parameters_t): Likewise.
3662 (arc_insn_length_parameters): Likewise.
3663 (arc_get_insn_variants): Likewise.
3664 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3666 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3669 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3672 PR tree-optimization/84687
3673 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3675 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3677 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3679 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3680 Rename to ppc_speculation_barrier.
3681 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3682 __builtin_ppc_speculation_barrier.
3684 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3687 * combine.c (combine_simplify_rtx): Don't try to simplify if
3688 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3691 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3693 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3694 to 32 bytes when compiling for POWER9.
3696 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3699 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3700 regparm >= 3 with no arg reg available also for calls with
3701 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3704 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3706 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3708 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3711 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3713 2018-03-05 Richard Biener <rguenther@suse.de>
3715 PR tree-optimization/84486
3716 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3717 When inserting a __builtin_assume_aligned call set the LHS
3718 SSA name alignment info accordingly.
3720 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3722 PR tree-optimization/84114
3723 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3724 Avoid reassociation of FLOAT_MODE addition.
3726 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3728 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3729 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3730 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3731 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3732 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3733 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3734 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3736 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3737 __builtin_ia32_wbinvd): New builtins.
3738 (SPECIAL_ARGS2): New.
3739 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3740 (SPECIAL_ARGS2): New.
3741 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3742 (ix86_valid_target_attribute_inner_p): Ditto.
3743 (ix86_init_mmx_sse_builtins): Add special_args2.
3744 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3745 TARGET_WBNOINVD_P): New.
3746 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3747 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3748 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3749 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3750 * config/i386/pconfigintrin.h: New file.
3751 * config/i386/wbnoinvdintrin.h: Ditto.
3752 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3753 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3755 2018-03-05 Richard Biener <rguenther@suse.de>
3757 PR tree-optimization/84670
3758 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3760 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3761 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3762 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3763 doesn't grow if all successors (recursively) were visited at least
3766 2018-03-05 Richard Biener <rguenther@suse.de>
3768 PR tree-optimization/84650
3769 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3770 if executed in the loop pipeline.
3772 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3774 * doc/configfiles.texi (Configuration Files): Move info about
3775 conditionalizing $target-protos.h to...
3776 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3777 differs from $target-protos.h.
3779 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3780 Chung-Ju Wu <jasonwucj@gmail.com>
3782 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3783 * config/nds32/nds32-multiple.md (setmemsi): Define.
3784 * config/nds32/nds32-memory-manipulation.c
3785 (nds32_gen_dup_4_byte_to_word_value): New.
3786 (emit_setmem_word_loop): New.
3787 (emit_setmem_byte_loop): New.
3788 (nds32_expand_setmem_loop): New.
3789 (nds32_expand_setmem_loop_v3m): New.
3790 (nds32_expand_setmem_unroll): New.
3791 (nds32_expand_setmem): New.
3793 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3794 Chung-Ju Wu <jasonwucj@gmail.com>
3796 * config/nds32/nds32-memory-manipulation.c
3797 (nds32_emit_load_store): New.
3798 (nds32_emit_post_inc_load_store): New.
3799 (nds32_emit_mem_move): New.
3800 (nds32_emit_mem_move_block): New.
3801 (nds32_expand_movmemsi_loop_unknown_size): New.
3802 (nds32_expand_movmemsi_loop_known_size): New.
3803 (nds32_expand_movmemsi_loop): New.
3804 (nds32_expand_movmemsi_unroll): New.
3805 (nds32_expand_movmemqi): Rename ...
3806 (nds32_expand_movmemsi): ... to this.
3807 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3808 (movmemsi): ... to this.
3809 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3810 (nds32_expand_movmemsi): ... to this.
3812 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3813 Monk Chiang <sh.chiang04@gmail.com>
3814 Chung-Ju Wu <jasonwucj@gmail.com>
3816 * config/nds32/nds32-protos.h
3817 (nds32_expand_load_multiple): New arguments.
3818 (nds32_expand_store_multiple): Ditto.
3819 (nds32_valid_multiple_load_store): Rename ...
3820 (nds32_valid_multiple_load_store_p): ... to this.
3821 * config/nds32/nds32-memory-manipulation.c
3822 (nds32_expand_load_multiple): Refine implementation.
3823 (nds32_expand_store_multiple): Ditto.
3824 * config/nds32/nds32-multiple.md
3825 (load_multiple): Update nds32_expand_load_multiple interface.
3826 (store_multiple): Update nds32_expand_store_multiple interface.
3827 * config/nds32/nds32-predicates.c
3828 (nds32_valid_multiple_load_store): Rename ...
3829 (nds32_valid_multiple_load_store_p): ... to this and refine
3831 * config/nds32/predicates.md
3832 (nds32_load_multiple_and_update_address_operation): New predicate.
3833 (nds32_store_multiple_and_update_address_operation): New predicate.
3835 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3836 Chung-Ju Wu <jasonwucj@gmail.com>
3838 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3839 (combo): New attribute.
3840 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3842 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3844 * config/nds32/nds32.opt: Change -mcmodel= default value.
3846 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3847 Monk Chiang <sh.chiang04@gmail.com>
3848 Chung-Ju Wu <jasonwucj@gmail.com>
3850 * config/nds32/constants.md (unspec_element): New enum.
3851 * config/nds32/constraints.md (Umw): New constraint.
3852 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3853 * config/nds32/nds32-intrinsic.md: Likewise.
3854 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3855 (nds32_valid_smw_lwm_base_p): New.
3856 (nds32_output_smw_single_word): New.
3857 (nds32_output_lmw_single_word): New.
3858 (nds32_expand_unaligned_load): New.
3859 (nds32_expand_unaligned_store): New.
3860 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3861 (nds32_output_smw_single_word): Declare.
3862 (nds32_output_lmw_single_word): Declare.
3863 (nds32_expand_unaligned_load): Declare.
3864 (nds32_expand_unaligned_store): Declare.
3865 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3866 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3867 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3868 NDS32_BUILTIN_UASTORE_DW.
3869 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3872 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3873 Kito Cheng <kito.cheng@gmail.com>
3874 Chung-Ju Wu <jasonwucj@gmail.com>
3876 * config/nds32/nds32-intrinsic.c
3877 (nds32_expand_builtin_null_ftype_reg): Delete.
3878 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3879 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3880 (nds32_read_argument): New.
3881 (nds32_legitimize_target): Ditto.
3882 (nds32_legitimize_argument): Ditto.
3883 (nds32_check_constant_argument): Ditto.
3884 (nds32_expand_unop_builtin): Ditto.
3885 (nds32_expand_unopimm_builtin): Ditto.
3886 (nds32_expand_binop_builtin): Ditto.
3887 (nds32_builtin_decl_impl): Ditto.
3888 (builtin_description): Ditto.
3889 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3890 (nds32_init_builtins_impl): Ditto.
3891 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3892 (nds32_builtin_decl): New.
3893 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3894 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3896 2018-03-02 Jeff Law <law@redhat.com>
3898 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3899 (redundant_insn, fill_simple_delay_slots): Likewise.
3900 (fill_slots_from_thread): Likewise.
3901 * resource.c (mark_referenced_resources): Likewise.
3902 (mark_set_resources, find_dead_or_set_registers): Likewise.
3904 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3906 * substring-locations.h (format_warning_va): Formatting fix for
3908 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3910 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3911 * substring-locations.c: Include intl.h.
3912 (format_warning_va): Turned into small wrapper around
3913 format_warning_n_va, renamed to ...
3914 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3915 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3917 (format_warning_at_substring_n): New function.
3918 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3919 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3920 format_warning_at_substring with just a shorter name instead of
3921 const function pointer.
3922 (fmtwarn_n): New function.
3923 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3924 appropriate, get rid of all the fmtstr temporaries, move conditionals
3925 with G_() wrapped string literals directly into fmtwarn arguments,
3926 cast dir.len to (int), formatting fixes.
3928 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3930 * doc/invoke.texi: Remove "Cilk Plus" references.
3932 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3933 Richard Biener <rguenther@suse.de>
3936 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3937 for error or warning attributes if CALL_FROM_THUNK_P is set.
3940 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3943 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3944 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3947 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3948 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3950 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3951 instead of -1U in last predictors element's probability member.
3953 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3956 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3957 arguments if they are comparable.
3959 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3961 PR tree-optimization/84634
3962 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3963 masks and masked_loop_p with a single loop_masks, making sure it's
3964 null for bb vectorization.
3966 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3968 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3969 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3970 loop->force_vectorize to check whether there is no alias.
3972 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3975 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3977 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3979 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3980 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3981 instead of a loop around prev_real_insn.
3982 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3986 * config/i386/i386.c (ix86_print_operand): Use conditional
3987 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3990 2018-03-02 Richard Biener <rguenther@suse.de>
3992 PR tree-optimization/84427
3993 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3994 (bitmap_set_subtract_values): Rewrite to handle multiple
3997 (prune_clobbered_mems): Likewise.
3998 (phi_translate): Take edge instead of pred/phiblock.
3999 (phi_translate_1): Likewise.
4000 (phi_translate_set): Likewise. Insert all translated
4001 exprs for a value into the set, keeping possibly multiple
4002 expressions per value.
4003 (compute_antic_aux): Adjust for phi_translate changes.
4004 When intersecting union the expressions and prune those
4005 not in the final value set, keeping possibly multiple
4006 expressions per value. Do not use value-insertion
4007 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4008 all expressions. Add verification that the value-sets
4009 only shrink during iteration.
4010 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4011 (do_pre_regular_insertion): Likewise.
4012 (do_pre_partial_partial_insertion): Likewise.
4014 2018-03-02 Richard Biener <rguenther@suse.de>
4017 * config/darwin.c (saved_debug_info_level): New static global.
4018 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4019 (darwin_asm_lto_end): Restore debug info generation settings.
4021 2018-03-01 Martin Liska <mliska@suse.cz>
4024 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4027 2018-03-01 Richard Biener <rguenther@suse.de>
4030 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4031 type annotation with LTO when debug was disabled at compile-time.
4033 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4035 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4037 (mips_final_postscan_insn): Likewise.
4039 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4041 PR rtl-optimization/84528
4042 * alias.c (init_alias_target): Add commentary.
4043 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4044 a unique base value if the frame pointer is not eliminated
4045 to the stack pointer.
4047 2018-03-01 Tom de Vries <tom@codesourcery.com>
4049 PR rtl-optimization/83327
4050 * lra-int.h (hard_regs_spilled_into): Declare.
4051 * lra.c (hard_regs_spilled_into): Define.
4052 (init_reg_info): Init hard_regs_spilled_into.
4053 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4054 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4055 (process_bb_lives): Handle hard_regs_spilled_into.
4056 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4057 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4059 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4061 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4062 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4063 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4064 * config/rs6000/aix72.h: New file.
4066 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4068 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4069 instead of warning_at with conditional singular and plural messages
4073 * stor-layout.c (update_alignment_for_field): For
4074 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4075 && !DECL_PACKED (field), do the alignment update, just use
4076 only desired_align instead of MAX (type_align, desired_align)
4078 (place_field): Don't do known_align < desired_align handling
4079 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4080 is non-NULL, instead do it after rli->prev_field handling and
4081 only if not within a bitfield word. For DECL_PACKED (field)
4082 use type_align of BITS_PER_UNIT.
4084 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4086 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4087 superfluous parentheses and trailing spaces.
4089 2018-02-28 Richard Biener <rguenther@suse.de>
4091 PR tree-optimization/84584
4092 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4093 SCoPs with fake exit edge.
4095 2018-02-28 Martin Liska <mliska@suse.cz>
4098 * timevar.c (timer::print): Fix format to properly print 100%
4101 2018-02-28 Richard Biener <rguenther@suse.de>
4104 * genmatch.c (capture_info::walk_match): Do not mark
4105 captured expressions without operands as expr_p given
4106 they act more like predicates and should be subject to
4107 "lost tail" side-effect preserving.
4109 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4111 PR rtl-optimization/81611
4112 * auto-inc-dec.c (attempt_change): Move dead note from
4113 mem_insn if it's the next use of regno
4114 (find_address): Take address use of reg holding
4115 non-incremented value. Add parm to limit search to the named
4117 (merge_in_block): Attempt to use a mem insn that is the next
4118 use of the original regno.
4120 2018-02-27 Martin Sebor <msebor@redhat.com>
4123 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4124 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4126 2018-02-27 Martin Sebor <msebor@redhat.com>
4128 PR translation/84207
4129 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4130 n argument to unsigned HOST_WIDE_INT.
4131 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4132 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4133 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4134 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4136 2018-02-27 Richard Biener <rguenther@suse.de>
4138 PR tree-optimization/84512
4139 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4140 Do not use the estimate returned from record_stmt_cost for
4141 the scalar iteration cost but sum properly using add_stmt_cost.
4143 2018-02-27 Richard Biener <rguenther@suse.de>
4145 PR tree-optimization/84466
4146 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4147 Adjust last change to less strictly validate use operands.
4149 2018-02-27 Martin Liska <mliska@suse.cz>
4151 PR gcov-profile/84548
4152 * gcov.c (process_file): Allow partial overlap and consider it
4153 also as group functions.
4154 (output_lines): Properly calculate range of lines for a group.
4156 2018-02-27 Martin Liska <mliska@suse.cz>
4158 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4159 'ggc' suffixes. Change first column width.
4160 (timer::print): Fix formatting of the column.
4162 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4164 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4165 preserve inline entry blocks for the sake of debug inline
4166 entry point markers alone.
4167 (remove_unused_locals): Suggest in comments a better place to
4168 force the preservation of inline entry blocks that are
4169 otherwise unused, but do not preserve them.
4171 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4173 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4175 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4178 * config/i386/constraints.md (Bs): Replace
4179 ix86_indirect_branch_register with
4180 TARGET_INDIRECT_BRANCH_REGISTER.
4182 * config/i386/i386.md (indirect_jump): Likewise.
4183 (tablejump): Likewise.
4184 (*sibcall_memory): Likewise.
4185 (*sibcall_value_memory): Likewise.
4186 Peepholes of indirect call and jump via memory: Likewise.
4187 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4188 (*sibcall_value_GOT_32): Likewise.
4189 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4190 (GOT_memory_operand): Likewise.
4191 (call_insn_operand): Likewise.
4192 (sibcall_insn_operand): Likewise.
4193 (GOT32_symbol_operand): Likewise.
4194 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4196 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4198 PR rtl-optimization/83496
4199 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4200 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4201 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4202 redundant insn, if any.
4203 (relax_delay_slots): Likewise.
4204 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4206 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4208 PR tree-optimization/83965
4209 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4210 that grouped statements are part of a reduction chain. Return
4211 true if the statement is not marked as a reduction itself but
4213 (vect_recog_dot_prod_pattern): Don't check whether the statement
4214 is part of a group here.
4215 (vect_recog_sad_pattern): Likewise.
4216 (vect_recog_widen_sum_pattern): Likewise.
4218 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4221 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4223 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4226 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4228 (ix86_output_indirect_function_return): New prototype.
4229 (ix86_split_simple_return_pop_internal): Likewise.
4230 * config/i386/i386.c (indirect_return_via_cx): New.
4231 (indirect_return_via_cx_bnd): Likewise.
4232 (indirect_thunk_name): Handle return va CX_REG.
4233 (output_indirect_thunk_function): Create alias for
4234 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4235 (ix86_output_indirect_jmp): Remove the bool argument.
4236 (ix86_output_indirect_function_return): New function.
4237 (ix86_split_simple_return_pop_internal): Likewise.
4238 * config/i386/i386.md (*indirect_jump): Don't pass false
4239 to ix86_output_indirect_jmp.
4240 (*tablejump_1): Likewise.
4241 (simple_return_pop_internal): Change it to define_insn_and_split.
4242 Call ix86_split_simple_return_pop_internal to split it for
4244 (simple_return_indirect_internal): Call
4245 ix86_output_indirect_function_return instead of
4246 ix86_output_indirect_jmp.
4248 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4251 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4252 memset and value initialization afterwards.
4254 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4256 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4258 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4261 * common/config/aarch64/aarch64-common.c
4262 (aarch_option_optimization_table[]): Switch
4263 off fomit-frame-pointer
4265 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4266 Chung-Ju Wu <jasonwucj@gmail.com>
4268 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4270 (store_multiple): Ditto.
4272 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4274 * config.gcc: Add --with-cpu support for nds32 target.
4275 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4276 * config/nds32/nds32.opt: Add -mcpu= option.
4278 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4280 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4281 isel=yes): Warn for these deprecated options.
4283 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4285 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4286 ISA_2_5_MASKS_EMBEDDED.
4288 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4290 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4291 p->max as pointers rather than using iterative_hash_expr.
4293 2018-02-23 Carl Love <cel@us.ibm.com>
4295 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4296 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4298 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4299 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4300 P8V_BUILTIN_VEC_VUNSIGNED2.
4302 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4305 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4306 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4308 * lra-constraints.c (curr_insn_transform): Set up
4309 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4311 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4312 * lra-eliminations.c (spill_pseudos): Ditto.
4313 (process_insn_for_elimination): Ditto.
4314 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4315 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4316 LRA_NON_CLOBBERED_ALT.
4318 2018-02-22 Martin Sebor <msebor@redhat.com>
4320 PR tree-optimization/84480
4321 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4322 to maybe_diag_stxncpy_trunc. Call it.
4323 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4324 from gimple_fold_builtin_strcpy. Print inlining stack.
4325 (handle_builtin_stxncpy): Print inlining stack.
4326 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4328 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4331 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4332 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4333 and -fcheck-pointer-bounds are used together.
4334 (indirect_thunk_prefix): New enum.
4335 (indirect_thunk_need_prefix): New function.
4336 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4337 "_nt" instead of "_bnd" for NOTRACK prefix.
4338 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4339 (output_indirect_thunk_function): Likewise.
4341 (ix86_code_end): Update output_indirect_thunk_function calls.
4342 (ix86_output_indirect_branch_via_reg): Replace
4343 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4344 (ix86_output_indirect_branch_via_push): Likewise.
4345 (ix86_output_function_return): Likewise.
4346 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4347 incompatible with -fcf-protection=branch and
4348 -fcheck-pointer-bounds.
4350 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4353 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4354 Change gcc_assert call to output_operand_lossage.
4356 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4358 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4360 2018-02-22 DJ Delorie <dj@redhat.com>
4361 Sebastian Perta <sebastian.perta@renesas.com>
4362 Oleg Endo <olegendo@gcc.gnu.org>
4364 * config/rx/rx.c (rx_rtx_costs): New function.
4365 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4367 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4369 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4371 2018-02-22 Martin Liska <mliska@suse.cz>
4374 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4375 Add "native" as a possible value.
4377 2018-02-22 Martin Liska <mliska@suse.cz>
4380 * config/i386/i386.c (ix86_option_override_internal):
4381 Add "native" as a possible value for -march and -mtune.
4383 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4386 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4387 to all type variants.
4389 PR tree-optimization/84503
4390 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4391 width as info->bitpos + info->bitsize - start.
4392 (merged_store_group::merge_overlapping): Simplify width computation.
4393 (check_no_overlap): New function.
4394 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4395 start + width and last_order of the group, fail if check_no_overlap
4397 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4398 to group if check_no_overlap fails.
4400 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4402 * config/rs6000/altivec.md: Delete contraint arguments to
4403 define_expand, define_split, and define_peephole2, and in
4404 define_insn_and_split if always unused.
4405 * config/rs6000/darwin.md: Ditto.
4406 * config/rs6000/dfp.md: Ditto.
4407 * config/rs6000/rs6000.md: Ditto.
4408 * config/rs6000/sync.md: Ditto.
4409 * config/rs6000/vector.md: Ditto.
4410 * config/rs6000/vsx.md: Ditto.
4412 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4414 * config/rs6000/altivec.md: Write output control strings as braced
4415 blocks instead of double-quoted strings.
4416 * config/rs6000/darwin.md: Ditto.
4417 * config/rs6000/rs6000.md: Ditto.
4418 * config/rs6000/vector.md: Ditto.
4419 * config/rs6000/vsx.md: Ditto.
4421 2018-02-21 Jason Merrill <jason@redhat.com>
4423 PR c++/84314 - ICE with templates and fastcall attribute.
4424 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4426 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4428 * ipa-cp.c (determine_versionability): Fix comment typos.
4430 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4433 * ipa-cp.c (determine_versionability): Do not version functions caling
4436 2018-02-21 Martin Liska <mliska@suse.cz>
4439 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4440 Add "native" as a possible value.
4441 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4442 the macro when native cpu detection is available.
4444 2018-02-21 Martin Liska <mliska@suse.cz>
4447 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4448 Add "native" as a possible value.
4449 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4450 when native cpu detection is available.
4452 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4453 Martin Sebor <msebor@redhat.com>
4455 PR tree-optimization/84478
4456 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4458 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4459 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4460 support which is conservatively correct, for 2 only stay conservative
4461 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4462 argument to the 2 argument get_range_strlen, adjust 6 arg
4463 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4465 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4466 (gimple_fold_builtin_strlen): Pass true as last argument to
4469 2018-02-20 Martin Sebor <msebor@redhat.com>
4472 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4473 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4474 (builtin_memref::builtin_memref): Factor out parts into
4475 set_base_and_offset and call it.
4477 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4480 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4481 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4482 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4483 search at the associated MODE_INT.
4485 2018-02-20 Jeff Law <law@redhat.com>
4488 PR tree-optimization/81592
4490 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4491 for range data rather than using global data.
4492 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4493 range data rather than using global data.
4494 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4495 pass it to children as needed.
4496 (struct directive::fmtresult): Similarly.
4497 (struct directive::set_width): Similarly.
4498 (struct directive::set_precision): Similarly.
4499 (format_integer, format_directive, parse_directive): Similarly.
4500 (format_none): Accept unnamed vr_values parameter.
4501 (format_percent, format_floating, format_character): Similarly.
4502 (format_string, format_plain): Similarly.
4503 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4504 the EVRP range analyzer for range data rather than using global data.
4505 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4506 gimple-ssa-evrp-analyze.h
4507 (class sprintf_dom_walker): Add after_dom_children member function.
4508 Add evrp_range_analyzer member.
4509 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4510 range analyzer as needed.
4511 (sprintf_dom_walker::after_dom_children): New member function.
4512 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4514 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4515 (evrp_range_analyzer::pop_to_marker): Likewise.
4517 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4519 PR tree-optimization/84419
4520 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4521 with the required type if its current type is compatible but
4524 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4527 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4528 after vectorization.
4530 2018-02-20 Martin Liska <mliska@suse.cz>
4533 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4534 possible values if we don't have a hint.
4536 2018-02-20 Martin Liska <mliska@suse.cz>
4540 * final.c (shorten_branches): Build align_tab array with one
4542 * opts.c (finish_options): Add alignment option limit check.
4543 (MAX_CODE_ALIGN): Likewise.
4544 (MAX_CODE_ALIGN_VALUE): Likewise.
4545 * doc/invoke.texi: Document maximum allowed option value for
4546 all -falign-* options.
4548 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4551 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4552 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4553 * var-tracking.c (emit_note_insn_var_location): Remove all references
4554 to NOTE_INSN_CALL_ARG_LOCATION.
4555 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4556 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4557 Use copy_rtx_if_shared.
4558 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4559 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4560 (dwarf2out_var_location): Remove handling of
4561 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4563 * final.c (final_scan_insn): Remove all references to
4564 NOTE_INSN_CALL_ARG_LOCATION.
4565 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4566 before dumping final insns.
4567 * except.c (emit_note_eh_region_end): Remove all references to
4568 NOTE_INSN_CALL_ARG_LOCATION.
4569 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4570 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4571 * config/arc/arc.c (hwloop_optimize): Likewise.
4572 * config/arm/arm.c (create_fix_barrier): Likewise.
4573 * config/s390/s390.c (s390_chunkify_start): Likewise.
4574 * config/sh/sh.c (find_barrier): Likewise.
4575 * config/i386/i386.c (rest_of_insert_endbranch,
4576 ix86_seh_fixup_eh_fallthru): Likewise.
4577 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4578 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4579 * config/frv/frv.c (frv_function_prologue): Likewise.
4580 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4582 (note_outside_basic_block_p): Remove all references to
4583 NOTE_INSN_CALL_ARG_LOCATION.
4584 * gengtype.c (adjust_field_rtx_def): Likewise.
4585 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4587 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4588 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4591 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4594 PR tree-optimization/84452
4595 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4596 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4599 2018-02-19 Martin Liska <mliska@suse.cz>
4602 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4604 2018-02-19 Martin Liska <mliska@suse.cz>
4605 Richard Sandiford <richard.sandiford@linaro.org>
4607 PR tree-optimization/82491
4608 * gimple-fold.c (get_base_constructor): Make earlier bail out
4611 2018-02-19 Carl Love <cel@us.ibm.com>
4613 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4614 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4616 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4617 P8V_BUILTIN_VEC_NEG.
4619 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4621 * config/rl78/rl78.md (movdf): New define expand.
4623 2018-02-19 Martin Liska <mliska@suse.cz>
4626 * doc/invoke.texi: Fix typo.
4627 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4629 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4631 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4632 handle rs6000_single_float and rs6000_double_float specially for
4635 2018-02-16 Jeff Law <law@redhat.com>
4637 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4640 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4643 * ipa-inline.c (inline_small_functions): Fix a typo.
4645 2018-02-16 Nathan Sidwell <nathan@acm.org>
4647 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4649 2018-02-16 Carl Love <cel@us.ibm.com>
4651 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4652 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4653 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4654 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4655 expansion to P8V_BUILTIN_VEC_FLOAT2.
4657 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4659 PR rtl-optimization/70023
4660 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4661 src_regno into account.
4663 2018-02-16 Carl Love <cel@us.ibm.com>
4665 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4666 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4667 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4668 * config/rs6000/rs6000.c: Remove case statements for
4669 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4670 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4671 and P9V_BUILTIN_VEC_VINSERT4B.
4672 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4673 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4674 * config/rs6000/vsx.md:
4675 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4678 2018-02-16 Carl Love <cel@us.ibm.com>
4680 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4682 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4684 * config/rs6000/rs6000-c.c: Add the definitions for
4685 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4686 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4687 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4688 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4689 definition for insert4b and define insn *insert3b_internal.
4690 * doc/extend.texi: Add documentation for vec_extract4b.
4692 2018-02-16 Nathan Sidwell <nathan@acm.org>
4694 * doc/extend.texi (Backwards Compatibility): Mention friend
4695 injection. Note for-scope is deprecated.
4696 * doc/invoke.texi (-ffriend-injection): Deprecate.
4698 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4700 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4701 that moved to I2, also allow destinations that are a paradoxical
4702 subreg (instead of a normal reg).
4704 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4707 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4710 2018-02-16 Richard Biener <rguenther@suse.de>
4712 PR tree-optimization/84037
4713 PR tree-optimization/84016
4715 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4716 Adjust vec_construct for the fact we need additional higher latency
4717 128bit inserts for AVX256 and AVX512 vector builds.
4718 (ix86_add_stmt_cost): Scale vector construction cost for
4721 2018-02-16 Richard Biener <rguenther@suse.de>
4723 PR tree-optimization/84417
4724 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4725 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4726 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4728 2018-02-16 Martin Liska <mliska@suse.cz>
4731 * internal-fn.def (ASAN_CHECK): Set proper flags.
4732 (ASAN_MARK): Likewise.
4734 2018-02-16 Julia Koval <julia.koval@intel.com>
4736 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4737 from PTA_CANNONLAKE.
4739 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4742 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4743 Use ++iter rather than iter++ for std::list iterators.
4744 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4745 defer deleting them until all nodes in the forest are processed. Do
4746 free even leaf nodes. Change to_process into auto_vec.
4749 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4750 * vec.h (vec_default_construct): Use memset instead of placement new
4751 if BROKEN_VALUE_INITIALIZATION is defined.
4752 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4753 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4756 PR rtl-optimization/83723
4757 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4758 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4759 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4761 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4763 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4765 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4767 PR rtl-optimization/81443
4768 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4769 from inner REGs to paradoxical SUBREGs.
4771 2018-02-16 Richard Biener <rguenther@suse.de>
4773 PR tree-optimization/84399
4774 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4775 For operands we can analyze at their definition make sure we can
4776 analyze them at each use as well.
4778 2018-02-16 Richard Biener <rguenther@suse.de>
4780 PR tree-optimization/84190
4781 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4782 volatile accesses if the decl isn't volatile.
4784 2018-02-15 Jason Merrill <jason@redhat.com>
4786 PR c++/84314 - ICE with templates and fastcall attribute.
4787 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4788 TYPE_CANONICAL on an existing type.
4790 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4792 PR tree-optimization/84383
4793 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4794 dstoff nor call operand_equal_p if dstbase is NULL.
4796 PR tree-optimization/84334
4797 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4798 also a CONSTANT_CLASS_P, punt.
4800 2018-02-14 Jim Wilson <jimw@sifive.com>
4802 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4803 first SMALL_OPERAND check. New local min_second_step. Move assert
4804 to where locals are set. Add TARGET_RVC support.
4805 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4807 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4809 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4811 2018-02-14 Martin Sebor <msebor@redhat.com>
4813 PR tree-optimization/83698
4814 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4815 arrays constrain the offset range to their bounds.
4816 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4817 (builtin_access::overlap): Avoid setting the size of overlap if it's
4819 (maybe_diag_overlap): Also consider arrays when deciding what values
4820 of offsets to include in diagnostics.
4822 2018-02-14 Martin Sebor <msebor@redhat.com>
4825 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4826 that correspond to the kind of a declaration.
4828 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4831 * config/pa/pa.md: Load address of PIC label using the linkage table
4832 if the label is nonlocal.
4834 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4836 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4837 warning message if user requests -maltivec=be.
4838 * doc/invoke.texi: Document deprecation of -maltivec=be.
4840 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4843 * config/rs6000/rs6000-c.c: Update definitions for
4844 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4845 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4847 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4850 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4851 add _get_ssp intrinsics. Remove argument from
4852 __builtin_ia32_rdssp[d|q].
4853 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4854 * config/i386/i386-builtin.def: Remove argument from
4855 __builtin_ia32_rdssp[d|q].
4856 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4857 ix86_expand_special_args_builtin for _rdssp[d|q].
4858 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4859 Clear register before usage.
4860 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4861 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4863 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4865 PR tree-optimization/84357
4866 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4867 operand 1 of an ARRAY_REF too.
4869 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4872 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4873 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4875 (set_of_reg): New struct.
4876 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4877 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4878 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4880 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4881 Split into bitclr, bitset, bitinvert patterns if appropriate.
4882 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4883 use rx_fuse_in_memory_bitop.
4884 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4885 to named insn, correct maximum insn length.
4887 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4890 * machmode.def: Define a complex mode for PARTIAL_INT.
4891 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4893 * doc/rtl.texi: Document CSPImode.
4894 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4896 (msp430_hard_regno_nregs_with_padding): Likewise.
4898 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4901 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4903 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4905 PR rtl-optimization/84169
4906 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4907 we generated a parallel as new i3 and we split that to new i2 and i3
4908 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4909 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4910 those to i2, not i1. Partially rewrite this scan code.
4912 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4915 * stor-layout.c (place_field): For variable length fields, adjust
4916 offset_align afterwards not just based on the field's alignment,
4917 but also on the size.
4920 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4921 of exps and logs in the use_exp2 case.
4923 2018-02-13 Jeff Law <law@redhat.com>
4925 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4928 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4931 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4935 * common.opt (gas-loc-support, gas-locview-support): New.
4936 (ginline-points, ginternal-reset-location-views): New.
4937 * doc/invoke.texi: Document them. Use @itemx where intended.
4938 (gvariable-location-views): Adjust.
4939 * target.def (reset_location_view): New.
4940 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4941 (TARGET_RESET_LOCATION_VIEW): New.
4942 * doc/tm.texi: Rebuilt.
4943 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4944 (dwarf2out_default_as_locview_support): New.
4945 (output_asm_line_debug_info): Use option variables.
4946 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4947 (output_loc_list): Likewise.
4948 (add_high_low_attributes): Check option variables.
4949 Don't output entry view attribute in strict mode.
4950 (gen_inlined_subroutine_die): Check option variables.
4951 (dwarf2out_inline_entry): Likewise.
4952 (init_sections_and_labels): Likewise.
4953 (dwarf2out_early_finish): Likewise.
4954 (maybe_reset_location_view): New, from...
4955 (dwarf2out_var_location): ... here. Call it.
4956 * debug.h (dwarf2out_default_as_loc_support): Declare.
4957 (dwarf2out_default_as_locview_support): Declare.
4958 * hooks.c (hook_int_rtx_insn_0): New.
4959 * hooks.h (hook_int_rtx_insn_0): Declare.
4960 * toplev.c (process_options): Take -gas-loc-support and
4961 -gas-locview-support from dwarf2out. Enable
4962 -gvariable-location-views by default only with locview
4963 assembler support. Enable -ginternal-reset-location-views by
4964 default only if the target defines the corresponding hook.
4965 Enable -ginline-points by default if location views are
4966 enabled; force it disabled if statement frontiers are
4968 * tree-inline.c (expand_call_inline): Check option variables.
4969 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4971 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4973 PR tree-optimization/84321
4974 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4975 handling. Also check whether the anti-range contains any values
4976 that satisfy the mask; switch to a VR_RANGE if not.
4978 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4981 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4983 2018-02-13 Martin Jambor <mjambor@suse.cz>
4986 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4987 of call statements, also set location of a load to a temporary.
4989 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4991 * config/rl78/rl78.c (add_vector_labels): New function.
4992 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4993 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4994 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4995 which checks that no arguments are passed.
4996 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4997 * doc/extend.texi: Documentation for the new attribute.
4999 2018-02-13 Andreas Schwab <schwab@suse.de>
5001 * config/riscv/linux.h (CPP_SPEC): Define.
5003 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5006 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5007 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5008 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5009 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5010 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5011 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5012 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5013 temporarily for AES and PCLMUL builtins.
5015 PR tree-optimization/84339
5016 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5017 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5021 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5022 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5023 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5025 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5027 * omp-simd-clone.h: New file.
5028 * omp-simd-clone.c: Include omp-simd-clone.h.
5029 (expand_simd_clones): No longer static.
5030 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5031 cgraph.h and omp-simd-clone.h.
5032 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5033 (vect_recog_widen_shift_pattern): Formatting fix.
5034 (vect_pattern_recog_1): Don't check optab for calls.
5037 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5038 operands[2] into a REG before using gen_lowpart on it.
5040 2018-02-12 Jeff Law <law@redhat.com>
5043 * config/sh/sh.c (find_barrier): Consider a sibling call
5046 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5047 successfully back substituting a reg.
5049 2018-02-12 Richard Biener <rguenther@suse.de>
5051 PR tree-optimization/84037
5052 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5053 parameter, move visited init to caller.
5054 (vect_slp_analyze_operations): Separate cost from validity
5055 check, initialize visited once for all instances.
5056 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5058 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5060 (vect_model_promotion_demotion_cost): Likewise.
5061 (vectorizable_bswap): Guard cost modeling with !slp_node
5062 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5064 (vectorizable_call): Likewise.
5065 (vectorizable_conversion): Likewise.
5066 (vectorizable_assignment): Likewise.
5067 (vectorizable_shift): Likewise.
5068 (vectorizable_operation): Likewise.
5069 (vectorizable_store): Likewise.
5070 (vectorizable_load): Likewise.
5071 (vectorizable_condition): Likewise.
5072 (vectorizable_comparison): Likewise.
5074 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5077 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5078 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5079 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5081 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5084 * params.def (inline-min-speedup): Increase from 8 to 15.
5085 (max-inline-insns-auto): Decrease from 40 to 30.
5086 * ipa-split.c (consider_split): Add some buffer for function to
5087 be considered inlining candidate.
5088 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5091 2018-02-12 Richard Biener <rguenther@suse.de>
5093 PR tree-optimization/84037
5094 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5095 matched stmts if we cannot swap the non-matched ones.
5097 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5099 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5100 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5101 _mm_maskz_scalef_round_ss): New intrinsics.
5102 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5103 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5104 __builtin_ia32_scalefss_round): Remove.
5105 (__builtin_ia32_scalefsd_mask_round,
5106 __builtin_ia32_scalefss_mask_round): New intrinsics.
5107 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5108 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5109 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5110 "<round_constraint>")): Changed to ...
5111 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5112 "<round_scalar_constraint>")): ... this.
5113 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5114 %0, %1, %2<round_op3>}"): Changed to ...
5115 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5116 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5117 %2<round_scalar_mask_op3>}"): ... this.
5118 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5120 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5122 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5123 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5124 (_mm_maskz_sqrt_round_ss): New intrinsics.
5125 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5126 (__builtin_ia32_sqrtsd_mask_round)
5127 (__builtin_ia32_sqrtss_mask_round): New builtins.
5128 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5129 (__builtin_ia32_sqrtss_round): Remove.
5130 (__builtin_ia32_sqrtsd_mask_round)
5131 (__builtin_ia32_sqrtss_mask_round): New builtins.
5132 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5133 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5134 ((match_operand:VF_128 1 "vector_operand"
5135 "xBm,<round_constraint>")): Changed to ...
5136 ((match_operand:VF_128 1 "vector_operand"
5137 "xBm,<round_scalar_constraint>")): ... this.
5138 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5139 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5140 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5141 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5142 %<iptr>1<round_scalar_mask_op3>}): ... this.
5143 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5144 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5146 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5149 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5150 Cast vec_cmpeq result to correct type.
5151 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5152 Cast vec_cmpgt result to correct type.
5154 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5156 * final.c (final_scan_insn_1): Renamed from...
5157 (final_scan_insn): ... this. New wrapper, to recover
5158 seen from the outermost call in recursive ones.
5159 * config/sparc/sparc.c (output_return): Drop seen from call.
5160 (output_sibcall): Likewise.
5161 * config/visium/visium.c (output_branch): Likewise.
5163 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5165 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5168 2018-02-10 Alan Modra <amodra@gmail.com>
5171 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5172 Specify LR as an input.
5174 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5177 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5178 remove_member_access_dummy_vars): New functions.
5179 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5180 lower_omp_1, execute_lower_omp): Use them.
5182 PR rtl-optimization/84308
5183 * shrink-wrap.c (spread_components): Release todo vector.
5185 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5187 PR rtl-optimization/57193
5188 * ira-color.c (struct allocno_color_data): Add member
5189 conflict_allocno_hard_prefs.
5190 (update_conflict_allocno_hard_prefs): New.
5191 (bucket_allocno_compare_func): Add a preference based on
5192 conflict_allocno_hard_prefs.
5193 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5194 (color_allocnos): Remove a dead code. Initiate
5195 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5197 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5200 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5201 constraint from =wa to wa. Avoid a subreg on the output operand,
5202 instead use a pseudo and subreg it in a move.
5203 (p9_xxbrd_<mode>): Changed to ...
5204 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5205 (p9_xxbrd_v2df): New expander.
5206 (p9_xxbrw_<mode>): Changed to ...
5207 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5208 (p9_xxbrw_v4sf): New expander.
5210 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5212 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5214 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5217 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5218 multiply in 32-bit mode.
5219 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5220 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5223 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5225 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5226 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5227 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5228 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5230 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5233 * dwarf2out.c (is_trivial_indirect_ref): New function.
5234 (dwarf2out_late_global_decl): Do not generate a location
5235 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5236 and that are not defined in the current unit.
5238 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5240 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5241 instead of a libcall for UNORDERED.
5243 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5246 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5247 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5249 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5252 * config/s390/s390.c (s390_set_current_function): Invoke
5253 s390_indirect_branch_settings also if fndecl didn't change.
5255 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5257 * config/rs6000/rs6000.md (blockage): Set length to zero.
5259 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5261 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5263 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5266 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5267 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5271 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5272 PARALLEL incoming that failed vt_get_decl_and_offset check.
5275 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5276 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5278 (get_variable_section): For decls in named .bss* sections pass true as
5279 second argument to bss_initializer_p.
5281 2018-02-09 Marek Polacek <polacek@redhat.com>
5282 Jakub Jelinek <jakub@redhat.com>
5285 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5286 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5287 Sync some changes from cxx_fold_indirect_ref.
5289 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5291 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5293 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5294 (BLOCK_INLINE_ENTRY_LABEL): New.
5295 (dwarf2out_var_location): Disregard inline entry markers.
5296 (inline_entry_data): New struct.
5297 (inline_entry_data_hasher): New hashtable type.
5298 (inline_entry_data_hasher::hash): New.
5299 (inline_entry_data_hasher::equal): New.
5300 (inline_entry_data_table): New variable.
5301 (add_high_low_attributes): Add DW_AT_entry_pc and
5302 DW_AT_GNU_entry_view attributes if a pending entry is found
5303 in inline_entry_data_table. Add old entry_pc attribute only
5304 if debug nonbinding markers are disabled.
5305 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5306 markers are enabled.
5307 (block_within_block_p, dwarf2out_inline_entry): New.
5308 (dwarf2out_finish): Check that no entries remained in
5309 inline_entry_data_table.
5310 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5311 (final_scan_insn, notice_source_line): Likewise.
5312 (rest_of_clean_state): Skip inline entry markers.
5313 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5315 * gimple.c (gimple_build_debug_inline_entry): New.
5316 * gimple.h (enum gimple_debug_subcode): Add
5317 GIMPLE_DEBUG_INLINE_ENTRY.
5318 (gimple_build_debug_inline_entry): Declare.
5319 (gimple_debug_inline_entry_p): New.
5320 (gimple_debug_nonbind_marker_p): Adjust.
5321 * insn-notes.def (INLINE_ENTRY): New.
5322 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5323 inline entry marker notes.
5324 (print_insn): Likewise.
5325 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5326 (INSN_DEBUG_MARKER_KIND): Likewise.
5327 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5328 * tree-inline.c (expand_call_inline): Build and insert
5329 debug_inline_entry stmt.
5330 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5331 inline entry blocks early, if nonbind markers are enabled.
5332 (dump_scope_block): Dump fragment info.
5333 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5334 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5335 (gimple_build_debug_inline_entry): New.
5336 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5337 Enable/disable inline entry points too.
5338 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5339 (DEBUG_INSN): Describe inline entry markers.
5341 * common.opt (gvariable-location-views): New.
5342 (gvariable-location-views=incompat5): New.
5343 * config.in: Rebuilt.
5344 * configure: Rebuilt.
5345 * configure.ac: Test assembler for view support.
5346 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5347 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5348 * dwarf2out.c (var_loc_view): New typedef.
5349 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5350 (dwarf2out_locviews_in_attribute): New.
5351 (dwarf2out_locviews_in_loclist): New.
5352 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5353 (enum dw_line_info_opcode): Add LI_adv_address.
5354 (struct dw_line_info_table): Add view.
5355 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5356 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5357 (zero_view_p): New variable.
5358 (ZERO_VIEW_P): New macro.
5359 (output_asm_line_debug_info): New.
5360 (struct var_loc_node): Add view.
5361 (add_AT_view_list, AT_loc_list): New.
5362 (add_var_loc_to_decl): Add view param. Test it against last.
5363 (new_loc_list): Add view params. Record them.
5364 (AT_loc_list_ptr): Handle loc and view lists.
5365 (view_list_to_loc_list_val_node): New.
5366 (print_dw_val): Handle dw_val_class_view_list.
5367 (size_of_die): Likewise.
5368 (value_format): Likewise.
5369 (loc_list_has_views): New.
5370 (gen_llsym): Set vl_symbol too.
5371 (maybe_gen_llsym, skip_loc_list_entry): New.
5372 (dwarf2out_maybe_output_loclist_view_pair): New.
5373 (output_loc_list): Output view list or entries too.
5374 (output_view_list_offset): New.
5375 (output_die): Handle dw_val_class_view_list.
5376 (output_dwarf_version): New.
5377 (output_compilation_unit_header): Use it.
5378 (output_skeleton_debug_sections): Likewise.
5379 (output_rnglists, output_line_info): Likewise.
5380 (output_pubnames, output_aranges): Update version comments.
5381 (output_one_line_info_table): Output view numbers in asm comments.
5382 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5383 Call maybe_gen_llsym.
5384 (loc_list_from_tree_1): Adjust.
5385 (add_AT_location_description): Create view list attribute if
5386 needed, check it's absent otherwise.
5387 (convert_cfa_to_fb_loc_list): Adjust.
5388 (maybe_emit_file): Call output_asm_line_debug_info for test.
5389 (dwarf2out_var_location): Reset views as needed. Precompute
5390 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5391 attribute. Set view.
5392 (new_line_info_table): Reset next view.
5393 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5394 (dwarf2out_source_line): Likewise. Output view resets and labels to
5395 the assembler, or select appropriate line info opcodes.
5396 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5397 (optimize_string_length): Catch it. Adjust.
5398 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5399 dw_val_class_view_list, and remove it if no longer needed.
5400 (hash_loc_list): Hash view numbers.
5401 (loc_list_hasher::equal): Compare them.
5402 (optimize_location_lists): Check whether a view list symbol is
5403 needed, and whether the locview attribute is present, and
5404 whether they match. Remove the locview attribute if no longer
5406 (index_location_lists): Call skip_loc_list_entry for test.
5407 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5408 Use output_dwarf_version.
5409 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5410 (struct dw_val_node): Add val_view_list.
5411 * final.c (SEEN_NEXT_VIEW): New.
5412 (set_next_view_needed): New.
5413 (clear_next_view_needed): New.
5414 (maybe_output_next_view): New.
5415 (final_start_function): Rename to...
5416 (final_start_function_1): ... this. Take pointer to FIRST,
5417 add SEEN parameter. Emit param bindings in the initial view.
5418 (final_start_function): Reintroduce SEEN-less interface.
5419 (final): Rename to...
5420 (final_1): ... this. Take SEEN parameter. Output final pending
5421 next view at the end.
5422 (final): Reintroduce seen-less interface.
5423 (final_scan_insn): Output pending next view before switching
5424 sections or ending a block. Mark the next view as needed when
5425 outputting variable locations. Notify debug backend of section
5426 changes, and of location view changes.
5427 (rest_of_handle_final): Adjust.
5428 * toplev.c (process_options): Autodetect value for debug variable
5429 location views option. Warn on incompat5 without -gdwarf-5.
5430 * doc/invoke.texi (gvariable-location-views): New.
5431 (gvariable-location-views=incompat5): New.
5432 (gno-variable-location-views): New.
5434 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5436 PR tree-optimization/84136
5437 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5438 that the result of find_edge is non-NULL.
5440 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5443 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5444 storing integer register in SImode. Fix cost of 256 and 512
5445 byte aligned SSE register store.
5447 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5449 * config/i386/i386.c (ix86_multiplication_cost): Fix
5450 multiplication cost for TARGET_AVX512DQ.
5452 2018-02-08 Marek Polacek <polacek@redhat.com>
5454 PR tree-optimization/84238
5455 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5458 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5460 PR tree-optimization/84265
5461 * tree-vect-stmts.c (vectorizable_store): Don't treat
5462 VMAT_CONTIGUOUS accesses as grouped.
5463 (vectorizable_load): Likewise.
5465 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5467 PR tree-optimization/81635
5468 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5469 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5470 (test_round_for_mask): New functions.
5471 (wide_int_cc_tests): Call test_round_for_mask.
5472 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5473 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5474 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5475 range returned by get_range_info.
5477 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5480 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5481 * symtab.c: Include builtins.h
5482 (symtab_node::output_to_lto_symbol_table_p): Move here
5483 from lto-streamer-out.c:output_symbol_p.
5484 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5485 (output_symbol_p): Move all logic to symtab.c
5486 (produce_symtab): Update.
5488 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5490 * config/s390/s390-opts.h (enum indirect_branch): Define.
5491 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5492 (s390_indirect_branch_via_thunk)
5493 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5494 (enum s390_indirect_branch_type): Define.
5495 * config/s390/s390.c (struct s390_frame_layout, struct
5496 machine_function): Remove.
5497 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5498 (indirect_branch_table_label_no, indirect_branch_table_name):
5500 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5501 (enum s390_indirect_branch_option): Define.
5502 (s390_return_addr_from_memory): New function.
5503 (s390_handle_string_attribute): New function.
5504 (s390_attribute_table): Add new attribute handler.
5505 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5506 (s390_indirect_branch_via_thunk): New function.
5507 (s390_indirect_branch_via_inline_thunk): New function.
5508 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5509 sibling call optimization for non z10 compiles.
5510 (s390_emit_call): Force indirect branch target to be a single
5511 register. Add r1 clobber for non-z10 compiles.
5512 (s390_emit_epilogue): Emit return jump via return_use expander.
5513 (s390_reorg): Handle JUMP_INSNs as execute targets.
5514 (s390_option_override_internal): Perform validity checks for the
5515 new command line options.
5516 (s390_indirect_branch_attrvalue): New function.
5517 (s390_indirect_branch_settings): New function.
5518 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5519 (s390_output_indirect_thunk_function): New function.
5520 (s390_code_end): Implement target hook.
5521 (s390_case_values_threshold): Implement target hook.
5522 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5524 * config/s390/s390.h (struct s390_frame_layout)
5525 (struct machine_function): Move here from s390.c.
5526 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5527 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5528 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5529 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5530 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5531 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5532 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5533 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5534 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5535 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5536 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5537 (mnemonic attribute): Add values which aren't recognized
5539 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5540 pattern for branch conversion. Fix mnemonic attribute.
5541 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5542 indirect branch via thunk if requested.
5543 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5544 ("*indirect_jump"): Disable for branch conversion using out of
5546 ("indirect_jump_via_thunk<mode>_z10")
5547 ("indirect_jump_via_thunk<mode>")
5548 ("indirect_jump_via_inlinethunk<mode>_z10")
5549 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5550 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5551 ("casesi_jump_via_inlinethunk<mode>_z10")
5552 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5553 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5554 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5555 ("*indirect2_jump"): Disable for branch conversion.
5556 ("casesi_jump"): Turn into expander and expand patterns for branch
5558 ("return_use"): New expander.
5559 ("*return"): Emit return via thunk and rename it to ...
5560 ("*return<mode>"): ... this one.
5561 * config/s390/s390.opt: Add new options and and enum for the
5564 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5566 * lra-constraints.c (match_reload): Unconditionally use
5567 gen_lowpart_SUBREG, rather than selecting between that
5568 and equivalent gen_rtx_SUBREG code.
5570 2018-02-08 Richard Biener <rguenther@suse.de>
5572 PR tree-optimization/84233
5573 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5574 changed flag instead of boguously re-using phi_inserted.
5576 2018-02-08 Martin Jambor <mjambor@suse.cz>
5578 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5579 static local variables.
5581 2018-02-08 Richard Biener <rguenther@suse.de>
5583 PR tree-optimization/84278
5584 * tree-vect-stmts.c (vectorizable_store): When looking for
5585 smaller vector types to perform grouped strided loads/stores
5586 make sure the mode is supported by the target.
5587 (vectorizable_load): Likewise.
5589 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5591 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5592 Increase LDP/STP opportunities by adding adjacent callee-saves.
5594 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5596 PR rtl-optimization/84068
5597 PR rtl-optimization/83459
5598 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5600 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5602 PR tree-optimization/84224
5603 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5604 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5607 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5610 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5611 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5612 count, remove one USE.
5614 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5616 * doc/install.texi (Configuration): Document the
5617 --with-long-double-format={ibm,ieee} PowerPC configuration
5621 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5622 Convert from define_expand to be define_insn_and_split. Rework
5623 float/double/_Float128 conversions to QI/HI/SImode to work with
5624 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5625 conversions to QI/HImode types did a store and then a load to
5626 truncate the value. For conversions to VSX registers, don't split
5627 the insn, instead emit the code directly. Use the code iterator
5628 any_fix to combine signed and unsigned conversions.
5629 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5630 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5631 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5632 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5633 (fix_<mode>di2_hw): Likewise.
5634 (fixuns_<mode>di2_hw): Likewise.
5635 (fix_<mode>si2_hw): Likewise.
5636 (fixuns_<mode>si2_hw): Likewise.
5637 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5638 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5639 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5640 fix<uns>_trunc<SFDF:mode>si2_p8.
5641 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5643 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5644 (fix<uns>_<mode>_mem): Likewise.
5645 (fctiw<u>z_<mode>_mem): Likewise.
5646 (fix<uns>_<mode>_mem): Likewise.
5647 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5648 the register allocator from doing a direct move to the GPRs to do
5649 a store, and instead use the ISA 3.0 store byte/half-word from
5650 vector register instruction. For IEEE 128-bit floating point,
5651 also optimize stores of 32-bit ints.
5652 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5654 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5656 * genextract.c (push_pathstr_operand): New function to support
5658 (walk_rtx): Call push_pathstr_operand.
5659 (print_path): Support [a-zA-Z].
5661 2018-02-07 Richard Biener <rguenther@suse.de>
5663 PR tree-optimization/84037
5664 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5665 (cse_and_gimplify_to_preheader): Declare.
5666 (vect_get_place_in_interleaving_chain): Likewise.
5667 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5669 (_loop_vec_info::~_loop_vec_info): Delete it.
5670 (cse_and_gimplify_to_preheader): New function.
5671 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5672 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5673 (vectorizable_load): Likewise. For grouped stores always base
5674 the IV on the first element.
5675 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5676 condition before gimplifying.
5678 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5680 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5681 *DIV_EXPR and *MOD_EXPR.
5683 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5686 * config/i386/i386.c (ix86_option_override_internal): Mask out
5687 the CF_SET bit when checking -fcf-protection.
5689 2018-02-07 Tom de Vries <tom@codesourcery.com>
5692 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5695 2018-02-07 Richard Biener <rguenther@suse.de>
5697 PR tree-optimization/84204
5698 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5701 PR tree-optimization/84205
5702 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5703 special-case isl_ast_op_zdiv_r.
5705 PR tree-optimization/84223
5706 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5707 Only add conditions from within the region.
5708 (gather_bbs::after_dom_children): Adjust.
5710 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5713 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5714 * config/avr/avr.md: Only post-reload split REG-REG moves if
5715 either register is GENERAL_REG_P.
5717 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5719 PR tree-optimization/84235
5720 * tree-ssa-scopedtables.c
5721 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5722 if the subtraction is performed in floating point type where NaNs are
5723 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5724 build 1. Formatting fix.
5726 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5729 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5730 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5731 and skip it regardless of bb boundaries. Use CALL_P macro,
5732 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5733 unnecessarily, formatting fix.
5735 2018-02-06 Michael Collison <michael.collison@arm.com>
5737 * config/arm/thumb2.md:
5738 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5739 (*thumb_mov_notscc): Ditto.
5741 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5744 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5745 unsigned_fix, not "s".
5747 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5749 * configure.ac (gcc_fn_eh_frame_ro): New function.
5750 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5751 correct .eh_frame permissions.
5752 * configure: Regenerate.
5754 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5756 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5759 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5761 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5762 Display warning message for -mno-speculate-indirect-jumps.
5764 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5766 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5768 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5770 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5772 PR tree-optimization/84225
5773 * tree-eh.c (find_trapping_overflow): Only call
5774 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5776 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5779 * config/i386/i386.c: Reimplement the check of possible options
5780 -mibt/-mshstk conbination. Change error messages.
5781 * doc/invoke.texi: Fix a typo: remove extra '='.
5783 2018-02-06 Marek Polacek <polacek@redhat.com>
5785 PR tree-optimization/84228
5786 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5788 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5791 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5792 emitted arch directives.
5793 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5794 __ARM_FEATURE_COPROC before changing architectures.
5796 2018-02-06 Richard Biener <rguenther@suse.de>
5798 * config/i386/i386.c (print_reg): Fix typo.
5799 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5801 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5803 * configure: Regenerate.
5805 2018-02-05 Martin Sebor <msebor@redhat.com>
5807 PR tree-optimization/83369
5808 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5811 2018-02-05 Martin Liska <mliska@suse.cz>
5813 * doc/invoke.texi: Cherry-pick upstream r323995.
5815 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5817 * ira.c (ira_init_register_move_cost): Adjust comment.
5819 2018-02-05 Martin Liska <mliska@suse.cz>
5821 PR gcov-profile/84137
5822 * doc/gcov.texi: Fix typo in documentation.
5824 2018-02-05 Martin Liska <mliska@suse.cz>
5826 PR gcov-profile/83879
5827 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5828 using dlopen functionality.
5830 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5832 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5833 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5834 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5835 _mm_maskz_range_round_ss): New intrinsics.
5836 (__builtin_ia32_rangesd128_round)
5837 (__builtin_ia32_rangess128_round): Remove.
5838 (__builtin_ia32_rangesd128_mask_round,
5839 __builtin_ia32_rangess128_mask_round): New builtins.
5840 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5841 __builtin_ia32_rangess128_round): Remove.
5842 (__builtin_ia32_rangesd128_mask_round,
5843 __builtin_ia32_rangess128_mask_round): New builtins.
5844 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5845 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5846 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5847 "<round_saeonly_constraint>")): Changed to ...
5848 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5849 "<round_saeonly_scalar_constraint>")): ... this.
5850 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5851 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5852 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5853 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5854 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5856 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5858 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5860 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5861 Remove all values except native, 8540 and 8548.
5863 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5865 * config/i386/i386.c (ix86_output_function_return): Pass
5866 INVALID_REGNUM, instead of -1, as invalid register number to
5867 indirect_thunk_name and output_indirect_thunk.
5869 2018-02-02 Julia Koval <julia.koval@intel.com>
5871 * config.gcc: Add -march=icelake.
5872 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5873 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5874 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5875 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5876 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5877 (processor_target_table): Add icelake.
5878 (ix86_option_override_internal): Handle new PTAs.
5879 (get_builtin_code_for_version): Handle icelake.
5880 (M_INTEL_COREI7_ICELAKE): New.
5881 (fold_builtin_cpu): Handle icelake.
5882 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5883 * doc/invoke.texi: Add -march=icelake.
5885 2018-02-02 Julia Koval <julia.koval@intel.com>
5887 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5888 to wide_int_bitmask.
5889 * wide-int-bitmask.h: New.
5891 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5894 * config/i386/i386.md: Replace Pmode with word_mode in
5895 builtin_setjmp_setup and builtin_longjmp to support x32.
5897 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5901 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5903 (rs6000_supported_cpu_names): New static variable.
5904 (linux_cpu_translation_table): Likewise.
5905 (elf_platform) <cpu>: Define new static variable and use it.
5906 Translate kernel AT_PLATFORM name to canonical name if needed.
5907 Error if platform name is unknown.
5909 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5912 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5914 2018-02-01 Jeff Law <law@redhat.com>
5917 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5918 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5919 the scratch if RELEASE_VIA_POP is false.
5920 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5921 If we have to save a temporary register, decrement SIZE appropriately.
5922 Pass new arguments to release_scratch_register_on_entry.
5923 (ix86_adjust_stack_and_probe): Likewise.
5924 (ix86_emit_probe_stack_range): Pass new arguments to
5925 release_scratch_register_on_entry.
5927 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5929 PR rtl-optimization/84157
5930 * combine.c (change_zero_ext): Use REG_P predicate in
5931 front of HARD_REGISTER_P predicate.
5933 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5935 * config/avr/avr.c (avr_option_override): Move disabling of
5936 -fdelete-null-pointer-checks to...
5937 * common/config/avr/avr-common.c (avr_option_optimization_table):
5940 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5942 PR tree-optimization/81635
5943 * tree-data-ref.c (split_constant_offset_1): For types that
5944 wrap on overflow, try to use range info to prove that wrapping
5947 2018-02-01 Renlin Li <renlin.li@arm.com>
5950 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5952 (aarch64_register_move_cost): Likewise.
5953 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5955 (REG_CLASS_NAMES): Likewise.
5956 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5957 TAILCALL_ADDR_REGS. Remove IP registers.
5958 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5960 2018-02-01 Richard Biener <rguenther@suse.de>
5962 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5963 for specifying RPO order and allow NULL for that.
5964 * domwalk.c (dom_walker::dom_walker): Likewise.
5965 (dom_walker::walk): Handle NULL RPO order.
5966 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5968 (rewrite_update_dom_walker): Likewise.
5969 (mark_def_dom_walker): Likewise.
5971 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5973 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5974 (aarch64_maybe_expand_sve_subreg_move): Declare.
5975 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5976 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5978 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5979 that are semantically a reverse operation.
5980 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5981 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5982 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5984 (aarch64_can_change_mode_class): For big-endian, forbid changes
5985 between two SVE modes if they have different element sizes.
5987 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5989 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5990 the TImode handling for big-endian targets.
5992 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5994 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5995 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5997 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5998 Remove BSWAP handing for big-endian targets and use the form of
5999 LD1RQ appropariate for the mode.
6001 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6003 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6004 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6007 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6010 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6011 check for operands that need to go through aarch64_sve_reload_be.
6013 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6015 PR tree-optimization/81661
6016 PR tree-optimization/84117
6017 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6018 * tree-eh.c: Include gimplify.h.
6019 (find_trapping_overflow, replace_trapping_overflow,
6020 rewrite_to_non_trapping_overflow): New functions.
6021 * tree-vect-loop.c: Include tree-eh.h.
6022 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6023 * tree-data-ref.c: Include tree-eh.h.
6024 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6026 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6028 PR rtl-optimization/84123
6029 * combine.c (change_zero_ext): Check if hard register satisfies
6030 can_change_dest_mode before calling gen_lowpart_SUBREG.
6032 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6035 * ira.c (ira_init_register_move_cost): Remove assert.
6037 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6039 PR rtl-optimization/84071
6040 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6041 * doc/tm.texi: Regenerate.
6043 2018-01-31 Richard Biener <rguenther@suse.de>
6045 PR tree-optimization/84132
6046 * tree-data-ref.c (analyze_miv_subscript): Properly
6047 check whether evolution_function_is_affine_multivariate_p
6048 before calling gcd_of_steps_may_divide_p.
6050 2018-01-31 Julia Koval <julia.koval@intel.com>
6053 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6054 * config/i386/i386.md (rdpid_rex64) New.
6055 (rdpid): Make 32bit only.
6057 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6060 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6061 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6063 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6066 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6068 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6070 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6072 PR rtl-optimization/84071
6073 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6074 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6076 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6078 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6079 (arc_attribute_table): Add 'aux' attribute.
6080 (arc_in_small_data_p): Consider aux like variables.
6081 (arc_is_aux_reg_p): New function.
6082 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6083 (arc_get_aux_arg): New function.
6084 (prepare_move_operands): Handle aux-register access.
6085 (arc_handle_aux_attribute): New function.
6086 * doc/extend.texi (ARC Variable attributes): Add subsection.
6088 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6090 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6091 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6092 (arc_attribute_table): Add 'uncached' attribute.
6093 (arc_print_operand): Print '.di' flag for uncached memory
6095 (arc_in_small_data_p): Do not consider for small data the uncached
6097 (arc_is_uncached_mem_p): New function.
6098 * config/arc/predicates.md (compact_store_memory_operand): Check
6099 for uncached memory accesses.
6100 (nonvol_nonimm_operand): Likewise.
6101 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6103 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6106 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6107 falign-loops=): Add Optimization flag.
6109 2018-01-30 Jeff Law <law@redhat.com>
6112 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6113 INT_REGISTERS_SAVED. Check it prior to calling
6114 get_scratch_register_on_entry.
6115 (ix86_adjust_stack_and_probe): Similarly.
6116 (ix86_emit_probe_stack_range): Similarly.
6117 (ix86_expand_prologue): Corresponding changes.
6119 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6122 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6123 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6125 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6128 * lra-constraints.c (curr_insn_transform): Process AND in the
6131 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6133 PR rtl-optimization/83986
6134 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6135 dependence against last_pending_memory_flush in addition to
6138 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6140 PR tree-optimization/81611
6141 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6144 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6147 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6150 2018-01-30 Richard Biener <rguenther@suse.de>
6151 Jakub Jelinek <jakub@redhat.com>
6153 PR tree-optimization/84111
6154 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6155 inner loops added during recursion, as they don't have up-to-date
6158 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6161 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6162 (can_inline_edge_by_limits_p): ... here.
6163 (can_early_inline_edge_p, check_callers,
6164 update_caller_keys, update_callee_keys, recursive_inlining,
6165 add_new_edges_to_heap, speculation_useful_p,
6166 inline_small_functions,
6167 inline_small_functions, flatten_function,
6168 inline_to_all_callers_1): Update.
6170 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6172 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6175 2018-01-30 Richard Biener <rguenther@suse.de>
6177 PR tree-optimization/83008
6178 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6179 invariant and constant vector uses in stmts when they need
6182 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6185 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6186 * configure: Regenerate.
6188 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6190 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6192 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6193 Use gen_rtx_REG rather than gen_lowpart.
6195 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6197 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6198 rather than 0 when creating partial subregs.
6200 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6202 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6205 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6208 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6209 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6210 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6211 flags. This restores the settings used before the 2017-07-24.
6212 Turning off pre increment/decrement/modify allows IVOPTS to
6213 optimize DF/SF loops where the index is an int.
6215 2018-01-29 Richard Biener <rguenther@suse.de>
6216 Kelvin Nilsen <kelvin@gcc.gnu.org>
6219 * tree-vect-stmts.c (vectorizable_call): Don't call
6220 targetm.vectorize_builtin_md_vectorized_function if callee is
6223 2018-01-22 Carl Love <cel@us.ibm.com>
6225 * doc/extend.tex: Fix typo in second arg in
6226 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6228 2018-01-29 Richard Biener <rguenther@suse.de>
6230 PR tree-optimization/84086
6231 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6232 (flush_ssaname_freelist): When SSA names were released reset
6233 the SCEV hash table.
6235 2018-01-29 Richard Biener <rguenther@suse.de>
6237 PR tree-optimization/84057
6238 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6239 removed paths when removing edges.
6241 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6243 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6244 -mfunction-return=@var{choice}.
6246 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6249 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6251 (layout::print_source_line): Likewise.
6252 (test_get_line_width_without_trailing_whitespace): Add test cases.
6254 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6257 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6260 2018-01-26 Jim Wilson <jimw@sifive.com>
6262 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6264 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6267 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6269 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6270 and CMP + SUB-immediate -> SUBS.
6272 2018-01-26 Martin Sebor <msebor@redhat.com>
6274 PR tree-optimization/83896
6275 * tree-ssa-strlen.c (get_string_len): Rename...
6276 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6277 Avoid assuming length is constant.
6278 (handle_char_store): Use HOST_WIDE_INT for string length.
6280 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6283 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6284 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6286 2018-01-26 Richard Biener <rguenther@suse.de>
6288 PR rtl-optimization/84003
6289 * dse.c (record_store): Only record redundant stores when
6290 the earlier store aliases at least all accesses the later one does.
6292 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6294 PR rtl-optimization/83985
6295 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6296 REG_CFA_RESTORE insns.
6297 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6298 REG_CFA_RESTORE insns here.
6301 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6302 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6304 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6306 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6307 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6308 (arc_init): Likewise.
6309 (arc_override_options): Likewise.
6310 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6312 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6314 * config/arc/arc.h (TARGET_DBNZ): Define.
6315 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6316 properly set the tune attribute.
6317 (dbnz): Use TARGET_DBNZ guard.
6318 * config/arc/arc.opt (mtune): Add core3 option.
6320 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6322 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6323 recognize new pic like addresses.
6324 (arc_delegitimize_address): Clean up.
6326 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6328 * config/arc/arc-arches.def: Option mrf16 valid for all
6330 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6331 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6332 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6333 * config/arc/arc-tables.opt: Regenerate.
6334 * config/arc/arc.c (arc_conditional_register_usage): Handle
6335 reduced register file case.
6336 (arc_file_start): Set must have build attributes.
6337 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6339 * config/arc/arc.opt (mrf16): Add new option.
6340 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6341 * config/arc/genmultilib.awk: Handle new mrf16 option.
6342 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6343 * config/arc/t-multilib: Regenerate.
6344 * doc/invoke.texi (ARC Options): Document mrf16 option.
6346 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6348 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6349 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6350 (arc_attribute_table): Add 'secure_call' attribute.
6351 (arc_print_operand): Print secure call operand.
6352 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6354 (arc_is_secure_call_p): New function. * config/arc/arc.md
6355 (call_i): Add support for sjli instruction.
6356 (call_value_i): Likewise.
6357 * config/arc/constraints.md (Csc): New constraint.
6359 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6360 John Eric Martin <John.Martin@emmicro-us.com>
6362 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6363 * config/arc/arc.c (_arc_jli_section): New struct.
6364 (arc_jli_section): New type.
6365 (rc_jli_sections): New static variable.
6366 (arc_handle_jli_attribute): New function.
6367 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6368 (arc_file_end): New function.
6369 (TARGET_ASM_FILE_END): Define.
6370 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6371 (arc_add_jli_section): New function.
6372 (jli_call_scan): Likewise.
6373 (arc_reorg): Call jli_call_scan.
6374 (arc_output_addsi): Remove 'S' from printing asm operand.
6375 (arc_is_jli_call_p): New function.
6376 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6378 (movhi_insn): Likewise.
6379 (movsi_insn): Likewise.
6380 (movsi_set_cc_insn): Likewise.
6381 (loadqi_update): Likewise.
6382 (load_zeroextendqisi_update): Likewise.
6383 (load_signextendqisi_update): Likewise.
6384 (loadhi_update): Likewise.
6385 (load_zeroextendhisi_update): Likewise.
6386 (load_signextendhisi_update): Likewise.
6387 (loadsi_update): Likewise.
6388 (loadsf_update): Likewise.
6389 (movsicc_insn): Likewise.
6390 (bset_insn): Likewise.
6391 (bxor_insn): Likewise.
6392 (bclr_insn): Likewise.
6393 (bmsk_insn): Likewise.
6394 (bicsi3_insn): Likewise.
6395 (cmpsi_cc_c_insn): Likewise.
6396 (movsi_ne): Likewise.
6397 (movsi_cond_exec): Likewise.
6398 (clrsbsi2): Likewise.
6409 (call_i): Remove 'S' asm letter, add jli instruction.
6410 (call_value_i): Likewise.
6411 * config/arc/arc.op (mjli-always): New option.
6412 * config/arc/constraints.md (Cji): New constraint.
6413 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6415 (subsf3_fpx): Likewise.
6416 (mulsf3_fpx): Likewise.
6417 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6419 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6421 * doc/invoke.texi (ARC): Document mjli-always option.
6423 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6425 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6426 avoid addition with 0 and use incw and decw where possible.
6428 2018-01-26 Richard Biener <rguenther@suse.de>
6430 PR tree-optimization/81082
6431 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6432 association if it requires casting to unsigned.
6433 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6434 from fold_plusminus_mult_expr to catch important cases late when
6435 range info is available.
6437 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6439 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6440 * configure.ac (hidden_linkonce): New test.
6441 * configure: Regenerate.
6442 * config.in: Regenerate.
6444 2018-01-26 Julia Koval <julia.koval@intel.com>
6446 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6447 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6448 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6449 _mm_mask_bitshuffle_epi64_mask): Fix type.
6450 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6451 USI_FTYPE_V4DI_V4DI_USI): Remove.
6452 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6453 __builtin_ia32_vpshufbitqmb256_mask,
6454 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6455 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6456 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6458 2018-01-26 Alan Modra <amodra@gmail.com>
6461 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6462 UNSPEC_VBPERMQ. Sort other unspecs.
6464 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6466 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6468 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6471 * predict.c (drop_profile): Do not push/pop cfun; update also
6473 (handle_missing_profiles): Fix logic looking for zero profiles.
6475 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6478 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6479 on functions with #pragma omp declare simd or functions with simd
6481 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6482 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6483 Remove trailing \n from warning_at calls.
6485 2018-01-25 Tom de Vries <tom@codesourcery.com>
6488 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6489 for neutered workers.
6491 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6494 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6495 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6497 2018-01-24 Jeff Law <law@redhat.com>
6500 * i386.c (get_probe_interval): Move to earlier point.
6501 (ix86_compute_frame_layout): If -fstack-clash-protection and
6502 the frame is larger than the probe interval, then use pushes
6503 to save registers rather than reg->mem moves.
6504 (ix86_expand_prologue): Remove conditional for int_registers_saved
6507 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6510 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6511 min/max for never referenced object.
6513 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6516 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6518 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6519 attributes from DECL_ATTRIBUTES (decl) without affecting
6520 DECL_ATTRIBUTES (current_function_decl).
6521 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6522 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6524 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6526 PR tree-optimization/83979
6527 * fold-const.c (fold_comparison): Use constant_boolean_node
6528 instead of boolean_{true,false}_node.
6530 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6532 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6535 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6537 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6538 Simplify the clause that sets the length attribute.
6539 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6540 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6541 clause that sets the length attribute.
6542 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6544 2018-01-24 Tom de Vries <tom@codesourcery.com>
6547 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6548 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6549 Add strict parameter.
6550 (prevent_branch_around_nothing): Insert dummy insn between branch to
6551 label and label with no ptx insn inbetween.
6552 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6554 2018-01-24 Tom de Vries <tom@codesourcery.com>
6557 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6558 for neutered threads in warp.
6559 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6561 2018-01-24 Richard Biener <rguenther@suse.de>
6563 PR tree-optimization/83176
6564 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6567 2018-01-24 Richard Biener <rguenther@suse.de>
6569 PR tree-optimization/82819
6570 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6571 code generating pluses that are no-ops in the target precision.
6573 2018-01-24 Richard Biener <rguenther@suse.de>
6576 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6578 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6580 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6581 to merge probabilities.
6582 * predict.c (probably_never_executed): Also mark as cold functions
6583 with global 0 profile and guessed local profile.
6584 * profile-count.c (profile_probability::combine_with_count): New
6586 * profile-count.h (profile_probability::operator*,
6587 profile_probability::operator*=, profile_probability::operator/,
6588 profile_probability::operator/=): Reduce precision to adjusted
6589 and set value to guessed on contradictory divisions.
6590 (profile_probability::combine_with_freq): Remove.
6591 (profile_probability::combine_wiht_count): Declare.
6592 (profile_count::force_nonzero):: Set to adjusted.
6593 (profile_count::probability_in):: Set quality to adjusted.
6594 * tree-ssa-tail-merge.c (replace_block_by): Use
6597 2018-01-23 Andrew Waterman <andrew@sifive.com>
6598 Jim Wilson <jimw@sifive.com>
6600 * config/riscv/riscv.c (riscv_stack_boundary): New.
6601 (riscv_option_override): Set riscv_stack_boundary. Handle
6602 riscv_preferred_stack_boundary_arg.
6603 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6604 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6605 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6606 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6607 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6608 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6610 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6613 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6614 of struct ix86_frame.
6615 (ix86_expand_epilogue): Likewise. Add a local variable for
6616 the reg_save_offset field in struct ix86_frame.
6618 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6620 PR tree-optimization/82604
6621 * tree-loop-distribution.c (enum partition_kind): New enum item
6622 PKIND_PARTIAL_MEMSET.
6623 (partition_builtin_p): Support above new enum item.
6624 (generate_code_for_partition): Ditto.
6625 (compute_access_range): Differentiate cases that equality can be
6626 proven at all loops, the innermost loops or no loops.
6627 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6628 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6629 (finalize_partitions, distribute_loop): Don't fuse partition of
6630 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6631 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6634 2018-01-23 Martin Liska <mliska@suse.cz>
6636 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6637 order to ignore the predictor.
6638 (PRED_POLYMORPHIC_CALL): Likewise.
6639 (PRED_RECURSIVE_CALL): Likewise.
6641 2018-01-23 Martin Liska <mliska@suse.cz>
6643 * tree-profile.c (tree_profiling): Print function header to
6644 aware reader which function we are working on.
6645 * value-prof.c (gimple_find_values_to_profile): Do not print
6646 not interesting value histograms.
6648 2018-01-23 Martin Liska <mliska@suse.cz>
6650 * profile-count.h (enum profile_quality): Add
6651 profile_uninitialized as the first value. Do not number values
6652 as they are zero based.
6653 (profile_count::verify): Update sanity check.
6654 (profile_probability::verify): Likewise.
6656 2018-01-23 Nathan Sidwell <nathan@acm.org>
6658 * doc/invoke.texi (ffor-scope): Deprecate.
6660 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6662 PR tree-optimization/83510
6663 * domwalk.c (set_all_edges_as_executable): New function.
6664 (dom_walker::dom_walker): Convert bool param
6665 "skip_unreachable_blocks" to enum reachability. Move setup of
6666 edge flags to set_all_edges_as_executable and only do it when
6667 reachability is REACHABLE_BLOCKS.
6668 * domwalk.h (enum dom_walker::reachability): New enum.
6669 (dom_walker::dom_walker): Convert bool param
6670 "skip_unreachable_blocks" to enum reachability.
6671 (set_all_edges_as_executable): New decl.
6672 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6673 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6675 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6676 but converting true to REACHABLE_BLOCKS.
6677 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6679 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6680 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6681 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6683 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6684 if check_all_array_refs will be called.
6686 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6688 * tree.c (selftest::test_location_wrappers): Add more test
6691 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6693 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6694 (selftest::test_bit_in_range): Likewise.
6696 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6699 * doc/sourcebuild.texi (vect_float): Say that the selector
6700 only describes the situation when -funsafe-math-optimizations is on.
6701 (vect_float_strict): Document.
6703 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6705 PR tree-optimization/83965
6706 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6707 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6708 instead of checking only for a reduction.
6709 (vect_recog_widen_sum_pattern): Likewise.
6711 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6713 * predict.c (probably_never_executed): Only use precise profile info.
6714 (compute_function_frequency): Skip after inlining hack since we now
6715 have quality checking.
6717 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6719 * profile-count.h (profile_probability::very_unlikely,
6720 profile_probability::unlikely, profile_probability::even): Set
6721 precision to guessed.
6723 2018-01-23 Richard Biener <rguenther@suse.de>
6725 PR tree-optimization/83963
6726 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6727 Properly terminate dominator walk when crossing the exit edge not
6728 when visiting its source block.
6730 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6733 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6734 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6736 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6738 PR tree-optimization/83957
6739 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6740 semicolon after for body surrounded by braces.
6742 PR tree-optimization/83081
6743 * profile-count.h (profile_probability::split): New method.
6744 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6745 Use profile_probability::split.
6746 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6747 when splitting a single conditional jump into 2.
6749 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6751 PR tree-optimization/69452
6752 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6755 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6757 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6758 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6759 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6761 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6763 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6764 * config/rl78/rl78.md (movdi): New define_expand.
6765 * config/rl78/rl78.c (rl78_split_movdi): New function.
6767 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6770 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6772 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6773 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6774 128-bit to produce an UNSPEC move to get the double word with the
6775 signbit and then a shift directly to do signbit.
6776 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6777 implementation with a new version that just does either a direct
6778 move or a regular move. Move memory interface to separate insns.
6779 Move insns so they are next to the expander.
6780 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6781 with signbit move. Split big and little endian case.
6782 (signbit<mode>2_dm_mem_le): Likewise.
6783 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6784 (signbit<mode>2_dm2): Likewise.
6786 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6788 * config/rl78/rl78.md (anddi3): New define_expand.
6790 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6792 * config/rl78/rl78.md (umindi3): New define_expand.
6794 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6796 * config/rl78/rl78.md (smindi3): New define_expand.
6798 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6800 * config/rl78/rl78.md (smaxdi3): New define_expand.
6802 2018-01-22 Carl Love <cel@us.ibm.com>
6804 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6805 LVX_V1TI): Add macro expansion.
6806 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6807 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6808 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6809 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6810 Change check to determine if the instruction is a byte reversing
6811 entry. Fix typo in comment.
6812 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6813 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6814 Add def_builtin calls for new builtins.
6815 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6816 Add define_insn expansion.
6818 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6820 * config/rl78/rl78.md (umaxdi3): New define_expand.
6822 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6824 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6825 for non-QImode registers.
6827 2018-01-22 Richard Biener <rguenther@suse.de>
6829 PR tree-optimization/83963
6830 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6831 including the loop exit block.
6832 (scop_detection::merge_sese): Likewise.
6833 (scop_detection::add_scop): Do it here instead.
6835 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6837 * doc/sourcebuild.texi (arm_softfloat): Document.
6839 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6842 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6843 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6844 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6846 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6847 David Edelsohn <dje.gcc@gmail.com>
6850 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6851 Change "crset eq" to "crset 2".
6852 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6853 (*call_indirect_aix<mode>_nospec): Likewise.
6854 (*call_value_indirect_aix<mode>_nospec): Likewise.
6855 (*call_indirect_elfv2<mode>_nospec): Likewise.
6856 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6857 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6858 change assembly output from . to $.
6859 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6860 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6861 (*tablejump<mode>_internal1_nospec): Likewise.
6863 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6866 * config/sh/sh_optimize_sett_clrt.cc:
6867 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6869 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6871 PR tree-optimization/83940
6872 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6873 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6874 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6875 use it to pass back the definition type.
6876 (vect_check_store_rhs): Likewise rhs_dt_out.
6877 (vect_build_gather_load_calls): Add a mask_dt argument and use
6878 it instead of a call to vect_is_simple_use.
6879 (vectorizable_store): Update calls to vect_check_load_store_mask
6880 and vect_check_store_rhs. Use the dt returned by the latter instead
6881 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6882 instead of calls to vect_is_simple_use. Pass the scalar rather
6883 than the vector operand to vect_is_simple_use when handling
6884 second and subsequent copies of an rhs value.
6885 (vectorizable_load): Update calls to vect_check_load_store_mask
6886 and vect_build_gather_load_calls. Use the cached mask_dt and
6887 gs_info.offset_dt instead of calls to vect_is_simple_use.
6889 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6892 * tree-emutls.c: Include gimplify.h.
6893 (lower_emutls_2): New function.
6894 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6895 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6896 it before further processing.
6899 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6900 UINTVAL (trueop1) instead of INTVAL (op1).
6902 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6906 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6907 INCOMING_FRAME_SP_OFFSET if not defined.
6908 (scan_trace): Add ENTRY argument. If true and
6909 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6910 emit a note to adjust the CFA offset.
6911 (create_cfi_notes): Adjust scan_trace callers.
6912 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6913 INCOMING_FRAME_SP_OFFSET in the CIE.
6914 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6915 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6917 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6918 * doc/tm.texi: Regenerated.
6920 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6922 PR rtl-optimization/83147
6923 * lra-constraints.c (remove_inheritance_pseudos): Use
6924 lra_substitute_pseudo_within_insn.
6926 2018-01-19 Tom de Vries <tom@codesourcery.com>
6927 Cesar Philippidis <cesar@codesourcery.com>
6930 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6932 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6935 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6936 spaces for function labels.
6938 2018-01-19 Martin Liska <mliska@suse.cz>
6940 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6941 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6942 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6943 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6944 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6945 (PRED_CONST_RETURN): Change from 69 to 65.
6946 (PRED_NULL_RETURN): Change from 91 to 71.
6947 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6948 (PRED_LOOP_GUARD): Change from 66 to 73.
6950 2018-01-19 Martin Liska <mliska@suse.cz>
6952 * predict.c (predict_insn_def): Add new assert.
6953 (struct branch_predictor): Change type to signed integer.
6954 (test_prediction_value_range): Amend test to cover
6956 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6957 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6958 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6959 (PRED_LOOP_IV_COMPARE): Likewise.
6960 * predict.h (PROB_UNINITIALIZED): Define new constant.
6962 2018-01-19 Martin Liska <mliska@suse.cz>
6964 * predict.c (dump_prediction): Add new format for
6965 analyze_brprob.py script which is enabled with -details
6967 * profile-count.h (precise_p): New function.
6969 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6971 PR tree-optimization/83922
6972 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6973 there are no statements that need masking.
6974 (vect_active_double_reduction_p): New function.
6975 (vect_analyze_loop_operations): Use it when handling phis that
6976 are not in the loop header.
6978 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6980 PR tree-optimization/83914
6981 * tree-vect-loop.c (vectorizable_induction): Don't convert
6982 init_expr or apply the peeling adjustment for inductions
6983 that are nested within the vectorized loop.
6985 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6987 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6990 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6994 * function.h (gimplify_parameters): Add gimple_seq * argument.
6995 * function.c: Include gimple.h and options.h.
6996 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6997 for the added local temporaries if needed.
6998 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6999 if there are any parameter cleanups, wrap whole body into a
7000 try/finally with the cleanups.
7002 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7005 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7006 Use GET_MODE_CLASS for scalar floating point.
7008 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7012 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7013 Fix call of call_cgraph_insertion_hooks.
7015 2018-01-18 Martin Sebor <msebor@redhat.com>
7017 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7019 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7022 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7025 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7028 * common.opt: (-ffile-prefix-map): New option.
7029 * opts.c (common_handle_option): Defer it.
7030 * opts-global.c (handle_common_deferred_options): Handle it.
7031 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7032 * file-prefix-map.h: New file.
7033 (remap_debug_filename, add_debug_prefix_map): ...here.
7034 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7035 * final.c (debug_prefix_map, add_debug_prefix_map
7036 remap_debug_filename): Move to...
7037 * file-prefix-map.c: New file.
7038 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7039 generalize, get rid of alloca(), use strrchr() instead of strchr().
7040 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7041 Implement in terms of add_prefix_map().
7042 (remap_macro_filename, remap_debug_filename): Implement in term of
7044 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7045 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7046 * dbxout.c: Include file-prefix-map.h.
7047 * varasm.c: Likewise.
7048 * vmsdbgout.c: Likewise.
7049 * xcoffout.c: Likewise.
7050 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7051 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7052 * doc/invoke.texi (-ffile-prefix-map): Document.
7053 (-fdebug-prefix-map): Update description.
7055 2018-01-18 Martin Liska <mliska@suse.cz>
7057 * config/i386/i386.c (indirect_thunk_name): Document that also
7059 (output_indirect_thunk): Document why both instructions
7060 (pause and lfence) are generated.
7062 2018-01-18 Richard Biener <rguenther@suse.de>
7064 PR tree-optimization/83887
7065 * graphite-scop-detection.c
7066 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7067 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7068 (scop_detection::merge_sese): Re-implement with a flood-fill
7069 algorithm that properly finds a SESE region if it exists.
7071 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7074 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7075 pointer_diff optimizations use view_convert instead of convert.
7077 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7079 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7080 Generate different code for -mno-speculate-indirect-jumps.
7081 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7082 (*call_indirect_aix<mode>): Disable for
7083 -mno-speculate-indirect-jumps.
7084 (*call_indirect_aix<mode>_nospec): New define_insn.
7085 (*call_value_indirect_aix<mode>): Disable for
7086 -mno-speculate-indirect-jumps.
7087 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7088 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7089 -mno-speculate-indirect-jumps.
7090 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7092 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7094 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7095 long double type, set the flags for noting the default long double
7096 type, even if we don't pass or return a long double type.
7098 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7101 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7104 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7106 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7107 support for merge[hl].
7108 (fold_mergehl_helper): New helper function.
7109 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7110 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7111 (altivec_vmrglw_direct): Add xxmrglw insn.
7113 2018-01-17 Andrew Waterman <andrew@sifive.com>
7115 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7116 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7118 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7121 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7122 call the lto_location_cache before reading the
7123 DECL_SOURCE_LOCATION of the types.
7125 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7126 Richard Sandiford <richard.sandiford@linaro.org>
7128 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7129 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7130 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7131 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7132 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7134 * config/aarch64/constraints.md (aarch64_movti_operand):
7136 * config/aarch64/predicates.md (Uti): Add new constraint.
7138 2018-01-17 Carl Love <cel@us.ibm.com>
7140 * config/rs6000/vsx.md (define_expand xl_len_r,
7141 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7142 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7144 (define_expand, define_insn): Move the shift left from the
7145 define_insn to the define_expand for lxvl and stxvl instructions.
7146 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7147 and XL_LEN_R definitions to PURE.
7149 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7151 * config/i386/i386.c (indirect_thunk_name): Declare regno
7152 as unsigned int. Compare regno with INVALID_REGNUM.
7153 (output_indirect_thunk): Ditto.
7154 (output_indirect_thunk_function): Ditto.
7155 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7156 in the call to output_indirect_thunk_function.
7158 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7161 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7162 rather than the size of inner_type to determine the stack slot size
7163 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7165 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7168 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7171 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7173 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7174 endian Linux systems to optionally enable multilibs for selecting
7175 the long double type if the user configured an explicit type.
7176 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7177 have no long double multilibs if not defined.
7178 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7179 warn if the user used -mabi={ieee,ibm}longdouble and we built
7180 multilibs for long double.
7181 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7182 appropriate multilib option.
7183 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7185 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7186 for building long double multilibs.
7187 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7189 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7191 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7194 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7196 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7199 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7202 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7205 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7207 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7208 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7210 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7212 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7213 different rtl trees depending on TARGET_64BIT.
7214 (rs6000_gen_lvx): Likewise.
7216 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7218 * config/visium/visium.md (nop): Tweak comment.
7219 (hazard_nop): Likewise.
7221 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7223 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7224 -mspeculate-indirect-jumps.
7225 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7226 for -mno-speculate-indirect-jumps.
7227 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7228 (*call_value_indirect_elfv2<mode>): Disable for
7229 -mno-speculate-indirect-jumps.
7230 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7231 (indirect_jump): Emit different RTL for
7232 -mno-speculate-indirect-jumps.
7233 (*indirect_jump<mode>): Disable for
7234 -mno-speculate-indirect-jumps.
7235 (*indirect_jump<mode>_nospec): New define_insn.
7236 (tablejump): Emit different RTL for
7237 -mno-speculate-indirect-jumps.
7238 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7239 (tablejumpsi_nospec): New define_expand.
7240 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7241 (tablejumpdi_nospec): New define_expand.
7242 (*tablejump<mode>_internal1): Disable for
7243 -mno-speculate-indirect-jumps.
7244 (*tablejump<mode>_internal1_nospec): New define_insn.
7245 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7248 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7250 * caller-save.c (insert_save): Drop unnecessary parameter. All
7253 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7254 Richard Biener <rguenth@suse.de>
7257 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7258 return early, inline manually is_gimple_sizepos. Make sure if we
7259 call gimplify_expr we don't end up with a gimple constant.
7260 * tree.c (variably_modified_type_p): Don't return true for
7261 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7262 * gimplify.h (is_gimple_sizepos): Remove.
7264 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7266 PR tree-optimization/83857
7267 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7268 vectorizable_live_operation for pure SLP statements.
7269 (vectorizable_live_operation): Handle PHIs.
7271 2018-01-16 Richard Biener <rguenther@suse.de>
7273 PR tree-optimization/83867
7274 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7275 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7277 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7280 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7281 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7282 If off is not INTEGER_CST, issue a may not be aligned warning
7283 rather than isn't aligned. Use isn%'t rather than isn't.
7284 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7286 <case MULT_EXPR>: Improve the case when bottom and one of the
7287 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7288 operand, in that case check if the other operand is multiple of
7289 bottom divided by the INTEGER_CST operand.
7291 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7294 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7295 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7296 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7297 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7298 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7299 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7300 (pa_function_arg_size): New function.
7302 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7304 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7305 in a separate statement.
7307 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7309 PR tree-optimization/83847
7310 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7311 group gathers and scatters.
7313 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7315 PR rtl-optimization/86620
7316 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7318 PR rtl-optimization/83213
7319 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7320 to last if both are JUMP_INSNs.
7322 PR tree-optimization/83843
7323 * gimple-ssa-store-merging.c
7324 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7325 store_immediate_info for bswap/nop orig_stores.
7327 2018-01-15 Andrew Waterman <andrew@sifive.com>
7329 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7331 <UDIV>: Increase cost if !TARGET_DIV.
7333 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7335 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7336 (define_attr "cr_logical_3op"): New.
7337 (cceq_ior_compare): Adjust.
7338 (cceq_ior_compare_complement): Adjust.
7339 (*cceq_rev_compare): Adjust.
7340 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7341 (is_cracked_insn): Adjust.
7342 (insn_must_be_first_in_group): Adjust.
7343 * config/rs6000/40x.md: Adjust.
7344 * config/rs6000/440.md: Adjust.
7345 * config/rs6000/476.md: Adjust.
7346 * config/rs6000/601.md: Adjust.
7347 * config/rs6000/603.md: Adjust.
7348 * config/rs6000/6xx.md: Adjust.
7349 * config/rs6000/7450.md: Adjust.
7350 * config/rs6000/7xx.md: Adjust.
7351 * config/rs6000/8540.md: Adjust.
7352 * config/rs6000/cell.md: Adjust.
7353 * config/rs6000/e300c2c3.md: Adjust.
7354 * config/rs6000/e500mc.md: Adjust.
7355 * config/rs6000/e500mc64.md: Adjust.
7356 * config/rs6000/e5500.md: Adjust.
7357 * config/rs6000/e6500.md: Adjust.
7358 * config/rs6000/mpc.md: Adjust.
7359 * config/rs6000/power4.md: Adjust.
7360 * config/rs6000/power5.md: Adjust.
7361 * config/rs6000/power6.md: Adjust.
7362 * config/rs6000/power7.md: Adjust.
7363 * config/rs6000/power8.md: Adjust.
7364 * config/rs6000/power9.md: Adjust.
7365 * config/rs6000/rs64.md: Adjust.
7366 * config/rs6000/titan.md: Adjust.
7368 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7370 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7371 ix86_indirect_branch_register logic.
7373 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7375 * config/i386/constraints.md (Bs): Update
7376 ix86_indirect_branch_register check. Don't check
7377 ix86_indirect_branch_register with GOT_memory_operand.
7379 * config/i386/predicates.md (GOT_memory_operand): Don't check
7380 ix86_indirect_branch_register here.
7381 (GOT32_symbol_operand): Likewise.
7383 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7385 * config/i386/predicates.md (constant_call_address_operand):
7386 Rewrite ix86_indirect_branch_register logic.
7387 (sibcall_insn_operand): Likewise.
7389 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7391 * config/i386/constraints.md (Bs): Replace
7392 ix86_indirect_branch_thunk_register with
7393 ix86_indirect_branch_register.
7395 * config/i386/i386.md (indirect_jump): Likewise.
7396 (tablejump): Likewise.
7397 (*sibcall_memory): Likewise.
7398 (*sibcall_value_memory): Likewise.
7399 Peepholes of indirect call and jump via memory: Likewise.
7400 * config/i386/i386.opt: Likewise.
7401 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7402 (GOT_memory_operand): Likewise.
7403 (call_insn_operand): Likewise.
7404 (sibcall_insn_operand): Likewise.
7405 (GOT32_symbol_operand): Likewise.
7407 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7410 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7411 type rather than type addr's type points to.
7412 (expand_omp_atomic_mutex): Likewise.
7413 (expand_omp_atomic): Likewise.
7415 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7418 * config/i386/i386.c (output_indirect_thunk_function): Use
7419 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7420 for __x86_return_thunk.
7422 2018-01-15 Richard Biener <rguenther@suse.de>
7425 * expmed.c (extract_bit_field_1): Fix typo.
7427 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7430 * config/arm/iterators.md (VF): New mode iterator.
7431 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7432 Remove integer-related logic from pattern.
7433 (neon_vabd<mode>_3): Likewise.
7435 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7438 * common.opt (fstrict-overflow): No longer an alias.
7439 (fwrapv-pointer): New option.
7440 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7441 also for pointer types based on flag_wrapv_pointer.
7442 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7443 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7444 opts->x_flag_wrapv got set.
7445 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7446 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7447 POINTER_TYPE_OVERFLOW_UNDEFINED.
7448 * match.pd: Likewise in address comparison pattern.
7449 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7451 2018-01-15 Richard Biener <rguenther@suse.de>
7454 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7455 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7456 Reset type names to their identifier if their TYPE_DECL doesn't
7457 have linkage (and thus is used for ODR and devirt).
7458 (save_debug_info_for_decl): Remove.
7459 (save_debug_info_for_type): Likewise.
7460 (add_tree_to_fld_list): Adjust.
7461 * tree-pretty-print.c (dump_generic_node): Make dumping of
7462 type names more robust.
7464 2018-01-15 Richard Biener <rguenther@suse.de>
7466 * BASE-VER: Bump to 8.0.1.
7468 2018-01-14 Martin Sebor <msebor@redhat.com>
7471 * builtins.c (check_access): Avoid warning when the no-warning bit
7474 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7476 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7477 * ira-color (allocno_hard_regs_compare): Likewise.
7479 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7482 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7483 Use .pushsection/.popsection.
7485 2018-01-14 Martin Sebor <msebor@redhat.com>
7488 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7490 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7492 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7493 entry from extra_headers.
7494 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7495 extra_headers, make the list bitwise identical to the i?86-*-* one.
7497 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7499 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7500 -mcmodel=large with -mindirect-branch=thunk,
7501 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7502 -mfunction-return=thunk-extern.
7503 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7504 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7505 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7507 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7509 * config/i386/i386.c (print_reg): Print the name of the full
7510 integer register without '%'.
7511 (ix86_print_operand): Handle 'V'.
7512 * doc/extend.texi: Document 'V' modifier.
7514 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7516 * config/i386/constraints.md (Bs): Disallow memory operand for
7517 -mindirect-branch-register.
7519 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7520 (GOT_memory_operand): Likewise.
7521 (call_insn_operand): Likewise.
7522 (sibcall_insn_operand): Likewise.
7523 (GOT32_symbol_operand): Likewise.
7524 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7525 for -mindirect-branch-register.
7526 (tablejump): Likewise.
7527 (*sibcall_memory): Likewise.
7528 (*sibcall_value_memory): Likewise.
7529 Disallow peepholes of indirect call and jump via memory for
7530 -mindirect-branch-register.
7531 (*call_pop): Replace m with Bw.
7532 (*call_value_pop): Likewise.
7533 (*sibcall_pop_memory): Replace m with Bs.
7534 * config/i386/i386.opt (mindirect-branch-register): New option.
7535 * doc/invoke.texi: Document -mindirect-branch-register option.
7537 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7539 * config/i386/i386-protos.h (ix86_output_function_return): New.
7540 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7541 set function_return_type.
7542 (indirect_thunk_name): Add ret_p to indicate thunk for function
7544 (output_indirect_thunk_function): Pass false to
7545 indirect_thunk_name.
7546 (ix86_output_indirect_branch_via_reg): Likewise.
7547 (ix86_output_indirect_branch_via_push): Likewise.
7548 (output_indirect_thunk_function): Create alias for function
7549 return thunk if regno < 0.
7550 (ix86_output_function_return): New function.
7551 (ix86_handle_fndecl_attribute): Handle function_return.
7552 (ix86_attribute_table): Add function_return.
7553 * config/i386/i386.h (machine_function): Add
7554 function_return_type.
7555 * config/i386/i386.md (simple_return_internal): Use
7556 ix86_output_function_return.
7557 (simple_return_internal_long): Likewise.
7558 * config/i386/i386.opt (mfunction-return=): New option.
7559 (indirect_branch): Mention -mfunction-return=.
7560 * doc/extend.texi: Document function_return function attribute.
7561 * doc/invoke.texi: Document -mfunction-return= option.
7563 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7565 * config/i386/i386-opts.h (indirect_branch): New.
7566 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7567 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7568 with local indirect jump when converting indirect call and jump.
7569 (ix86_set_indirect_branch_type): New.
7570 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7571 (indirectlabelno): New.
7572 (indirect_thunk_needed): Likewise.
7573 (indirect_thunk_bnd_needed): Likewise.
7574 (indirect_thunks_used): Likewise.
7575 (indirect_thunks_bnd_used): Likewise.
7576 (INDIRECT_LABEL): Likewise.
7577 (indirect_thunk_name): Likewise.
7578 (output_indirect_thunk): Likewise.
7579 (output_indirect_thunk_function): Likewise.
7580 (ix86_output_indirect_branch_via_reg): Likewise.
7581 (ix86_output_indirect_branch_via_push): Likewise.
7582 (ix86_output_indirect_branch): Likewise.
7583 (ix86_output_indirect_jmp): Likewise.
7584 (ix86_code_end): Call output_indirect_thunk_function if needed.
7585 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7587 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7588 (ix86_attribute_table): Add indirect_branch.
7589 * config/i386/i386.h (machine_function): Add indirect_branch_type
7590 and has_local_indirect_jump.
7591 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7593 (tablejump): Likewise.
7594 (*indirect_jump): Use ix86_output_indirect_jmp.
7595 (*tablejump_1): Likewise.
7596 (simple_return_indirect_internal): Likewise.
7597 * config/i386/i386.opt (mindirect-branch=): New option.
7598 (indirect_branch): New.
7601 (thunk-inline): Likewise.
7602 (thunk-extern): Likewise.
7603 * doc/extend.texi: Document indirect_branch function attribute.
7604 * doc/invoke.texi: Document -mindirect-branch= option.
7606 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7609 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7611 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7613 * ipa-inline.c (want_inline_small_function_p): Return false if
7614 inlining has already failed with CIF_FINAL_ERROR.
7615 (update_caller_keys): Call want_inline_small_function_p before
7617 (update_callee_keys): Likewise.
7619 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7621 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7623 (rs6000_quadword_masked_address_p): Likewise.
7624 (quad_aligned_load_p): Likewise.
7625 (quad_aligned_store_p): Likewise.
7626 (const_load_sequence_p): Add comment to describe the outer-most loop.
7627 (mimic_memory_attributes_and_flags): New function.
7628 (rs6000_gen_stvx): Likewise.
7629 (replace_swapped_aligned_store): Likewise.
7630 (rs6000_gen_lvx): Likewise.
7631 (replace_swapped_aligned_load): Likewise.
7632 (replace_swapped_load_constant): Capitalize argument name in
7633 comment describing this function.
7634 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7635 and stores that access quad-word aligned addresses and replace
7636 with stvx or lvx instructions when appropriate.
7637 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7638 New function prototype.
7639 (rs6000_quadword_masked_address_p): Likewise.
7640 (rs6000_gen_lvx): Likewise.
7641 (rs6000_gen_stvx): Likewise.
7642 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7643 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7644 when memory address is aligned.
7645 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7646 this split to select lvx instruction when memory address is aligned.
7647 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7648 instruction when memory address is aligned.
7649 (*vsx_le_perm_load_v16qi): Likewise.
7650 (four unnamed splitters): Modify to select the stvx instruction
7651 when memory is aligned.
7653 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7655 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7656 which appears in the queue multiple times.
7658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7659 Alan Hayward <alan.hayward@arm.com>
7660 David Sherwood <david.sherwood@arm.com>
7662 * tree-vectorizer.h (vec_lower_bound): New structure.
7663 (_loop_vec_info): Add check_nonzero and lower_bounds.
7664 (LOOP_VINFO_CHECK_NONZERO): New macro.
7665 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7666 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7667 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7668 fields. Make seg_len the distance travelled, not including the
7670 (dr_direction_indicator): Declare.
7671 (dr_zero_step_indicator): Likewise.
7672 (dr_known_forward_stride_p): Likewise.
7673 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7675 (runtime_alias_check_p): Allow runtime alias checks with
7677 (operator ==): Compare access_size and align.
7678 (prune_runtime_alias_test_list): Rework for new distinction between
7679 the access_size and seg_len.
7680 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7682 (get_segment_min_max): New function.
7683 (create_intersect_range_checks): Use it.
7684 (dr_step_indicator): New function.
7685 (dr_direction_indicator): Likewise.
7686 (dr_zero_step_indicator): Likewise.
7687 (dr_known_forward_stride_p): Likewise.
7688 * tree-loop-distribution.c (data_ref_segment_size): Return
7689 DR_STEP * (niters - 1).
7690 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7692 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7693 (vect_preserves_scalar_order_p): New function, split out from...
7694 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7695 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7696 (vect_vfa_access_size): New function.
7697 (vect_vfa_align): Likewise.
7698 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7699 (dump_lower_bound): New function.
7700 (vect_check_lower_bound): Likewise.
7701 (vect_small_gap_p): Likewise.
7702 (vectorizable_with_step_bound_p): Likewise.
7703 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7704 depencies if the vectorization factor is 1. Convert the checks
7705 for nonzero steps into checks on the bounds of DR_STEP. Try using
7706 a bunds check for variable steps if the minimum required step is
7707 relatively small. Update calls to the dr_with_seg_len
7708 constructor and to vect_compile_time_alias.
7709 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7711 (vect_loop_versioning): Call it.
7712 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7714 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7716 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7717 Alan Hayward <alan.hayward@arm.com>
7718 David Sherwood <david.sherwood@arm.com>
7720 * doc/sourcebuild.texi (vect_scatter_store): Document.
7721 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7723 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7725 * genopinit.c (main): Add supports_vec_scatter_store and
7726 supports_vec_scatter_store_cached to target_optabs.
7727 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7728 IFN_MASK_SCATTER_STORE.
7729 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7731 * internal-fn.h (internal_store_fn_p): Declare.
7732 (internal_fn_stored_value_index): Likewise.
7733 * internal-fn.c (scatter_store_direct): New macro.
7734 (expand_scatter_store_optab_fn): New function.
7735 (direct_scatter_store_optab_supported_p): New macro.
7736 (internal_store_fn_p): New function.
7737 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7738 IFN_MASK_SCATTER_STORE.
7739 (internal_fn_mask_index): Likewise.
7740 (internal_fn_stored_value_index): New function.
7741 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7743 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7744 * optabs-query.c (supports_vec_scatter_store_p): New function.
7745 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7746 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7747 true for scatter stores.
7748 (vect_gather_scatter_fn_p): Handle scatter stores too.
7749 (vect_check_gather_scatter): Consider using scatter stores if
7750 supports_vec_scatter_store_p.
7751 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7753 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7754 internal_fn_stored_value_index.
7755 (check_load_store_masking): Handle scatter stores too.
7756 (vect_get_store_rhs): Make public.
7757 (vectorizable_call): Use internal_store_fn_p.
7758 (vectorizable_store): Handle scatter store internal functions.
7759 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7760 when deciding whether the end of the group has been reached.
7761 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7762 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7763 (mask_scatter_store<mode>): New insns.
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_gather_scatter_fn_p): Declare.
7770 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7771 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7773 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7774 Use vect_truncate_gather_scatter_offset if we can't treat the
7775 operation as a normal gather load or scatter store.
7776 (get_group_load_store_type): Take the gather_scatter_info
7777 as argument. Try using a gather load or scatter store for
7778 single-element groups.
7779 (get_load_store_type): Update calls to get_group_load_store_type
7780 and vect_use_strided_gather_scatters_p.
7782 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7783 Alan Hayward <alan.hayward@arm.com>
7784 David Sherwood <david.sherwood@arm.com>
7786 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7787 optional tree argument.
7788 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7790 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7791 but continue to use the current value as a fallback.
7792 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7793 to compare the updates.
7794 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7795 (get_load_store_type): Use it when handling a strided access.
7796 (vect_get_strided_load_store_ops): New function.
7797 (vect_get_data_ptr_increment): Likewise.
7798 (vectorizable_load): Handle strided gather loads. Always pass
7799 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7801 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7802 Alan Hayward <alan.hayward@arm.com>
7803 David Sherwood <david.sherwood@arm.com>
7805 * doc/md.texi (gather_load@var{m}): Document.
7806 (mask_gather_load@var{m}): Likewise.
7807 * genopinit.c (main): Add supports_vec_gather_load and
7808 supports_vec_gather_load_cached to target_optabs.
7809 * optabs-tree.c (init_tree_optimization_optabs): Use
7810 ggc_cleared_alloc to allocate target_optabs.
7811 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7812 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7814 * internal-fn.h (internal_load_fn_p): Declare.
7815 (internal_gather_scatter_fn_p): Likewise.
7816 (internal_fn_mask_index): Likewise.
7817 (internal_gather_scatter_fn_supported_p): Likewise.
7818 * internal-fn.c (gather_load_direct): New macro.
7819 (expand_gather_load_optab_fn): New function.
7820 (direct_gather_load_optab_supported_p): New macro.
7821 (direct_internal_fn_optab): New function.
7822 (internal_load_fn_p): Likewise.
7823 (internal_gather_scatter_fn_p): Likewise.
7824 (internal_fn_mask_index): Likewise.
7825 (internal_gather_scatter_fn_supported_p): Likewise.
7826 * optabs-query.c (supports_at_least_one_mode_p): New function.
7827 (supports_vec_gather_load_p): Likewise.
7828 * optabs-query.h (supports_vec_gather_load_p): Declare.
7829 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7830 and memory_type field.
7831 (NUM_PATTERNS): Bump to 15.
7832 * tree-vect-data-refs.c: Include internal-fn.h.
7833 (vect_gather_scatter_fn_p): New function.
7834 (vect_describe_gather_scatter_call): Likewise.
7835 (vect_check_gather_scatter): Try using internal functions for
7836 gather loads. Recognize existing calls to a gather load function.
7837 (vect_analyze_data_refs): Consider using gather loads if
7838 supports_vec_gather_load_p.
7839 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7840 (vect_get_gather_scatter_offset_type): Likewise.
7841 (vect_convert_mask_for_vectype): Likewise.
7842 (vect_add_conversion_to_patterm): Likewise.
7843 (vect_try_gather_scatter_pattern): Likewise.
7844 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7845 (vect_vect_recog_func_ptrs): Add it.
7846 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7847 internal_fn_mask_index and internal_gather_scatter_fn_p.
7848 (check_load_store_masking): Take the gather_scatter_info as an
7849 argument and handle gather loads.
7850 (vect_get_gather_scatter_ops): New function.
7851 (vectorizable_call): Check internal_load_fn_p.
7852 (vectorizable_load): Likewise. Handle gather load internal
7854 (vectorizable_store): Update call to check_load_store_masking.
7855 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7856 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7857 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7858 (aarch64_gather_scale_operand_d): New predicates.
7859 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7860 (mask_gather_load<mode>): New insns.
7862 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7863 Alan Hayward <alan.hayward@arm.com>
7864 David Sherwood <david.sherwood@arm.com>
7866 * optabs.def (fold_left_plus_optab): New optab.
7867 * doc/md.texi (fold_left_plus_@var{m}): Document.
7868 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7869 * internal-fn.c (fold_left_direct): Define.
7870 (expand_fold_left_optab_fn): Likewise.
7871 (direct_fold_left_optab_supported_p): Likewise.
7872 * fold-const-call.c (fold_const_fold_left): New function.
7873 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7874 * tree-parloops.c (valid_reduction_p): New function.
7875 (gather_scalar_reductions): Use it.
7876 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7877 (vect_finish_replace_stmt): Declare.
7878 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7879 (needs_fold_left_reduction_p): New function, split out from...
7880 (vect_is_simple_reduction): ...here. Accept reductions that
7881 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7882 (vect_force_simple_reduction): Also store the reduction type in
7883 the assignment's STMT_VINFO_REDUC_TYPE.
7884 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7885 (merge_with_identity): New function.
7886 (vect_expand_fold_left): Likewise.
7887 (vectorize_fold_left_reduction): Likewise.
7888 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7889 scalar phi in place for it. Check for target support and reject
7890 cases that would reassociate the operation. Defer the transform
7891 phase to vectorize_fold_left_reduction.
7892 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7893 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7894 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7896 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7898 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7899 call to ifc_temp_var.
7901 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7902 Alan Hayward <alan.hayward@arm.com>
7903 David Sherwood <david.sherwood@arm.com>
7905 * target.def (legitimize_address_displacement): Take the original
7906 offset as a poly_int.
7907 * targhooks.h (default_legitimize_address_displacement): Update
7909 * targhooks.c (default_legitimize_address_displacement): Likewise.
7910 * doc/tm.texi: Regenerate.
7911 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7912 as an argument, moving assert of ad->disp == ad->disp_term to...
7913 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7914 Try calling targetm.legitimize_address_displacement before expanding
7915 the address rather than afterwards, and adjust for the new interface.
7916 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7917 Match the new hook interface. Handle SVE addresses.
7918 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7921 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7923 * Makefile.in (OBJS): Add early-remat.o.
7924 * target.def (select_early_remat_modes): New hook.
7925 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7926 * doc/tm.texi: Regenerate.
7927 * targhooks.h (default_select_early_remat_modes): Declare.
7928 * targhooks.c (default_select_early_remat_modes): New function.
7929 * timevar.def (TV_EARLY_REMAT): New timevar.
7930 * passes.def (pass_early_remat): New pass.
7931 * tree-pass.h (make_pass_early_remat): Declare.
7932 * early-remat.c: New file.
7933 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7935 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7938 Alan Hayward <alan.hayward@arm.com>
7939 David Sherwood <david.sherwood@arm.com>
7941 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7942 vfm1 with a bound_epilog parameter.
7943 (vect_do_peeling): Update calls accordingly, and move the prologue
7944 call earlier in the function. Treat the base bound_epilog as 0 for
7945 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7946 this base when peeling for gaps.
7947 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7948 with fully-masked loops.
7949 (vect_estimate_min_profitable_iters): Handle the single peeled
7950 iteration in that case.
7952 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7953 Alan Hayward <alan.hayward@arm.com>
7954 David Sherwood <david.sherwood@arm.com>
7956 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7957 single-element interleaving even if the size is not a power of 2.
7958 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7959 accesses for single-element interleaving if the group size is
7962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7963 Alan Hayward <alan.hayward@arm.com>
7964 David Sherwood <david.sherwood@arm.com>
7966 * doc/md.texi (fold_extract_last_@var{m}): Document.
7967 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7968 * optabs.def (fold_extract_last_optab): New optab.
7969 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7970 * internal-fn.c (fold_extract_direct): New macro.
7971 (expand_fold_extract_optab_fn): Likewise.
7972 (direct_fold_extract_optab_supported_p): Likewise.
7973 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7974 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7975 EXTRACT_LAST_REDUCTION.
7976 (get_initial_def_for_reduction): Do not create an initial vector
7977 for EXTRACT_LAST_REDUCTION reductions.
7978 (vectorizable_reduction): Leave the scalar phi in place for
7979 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7980 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7981 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7982 transform phase to vectorizable_condition.
7983 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7985 (vect_finish_stmt_generation): ...here.
7986 (vect_finish_replace_stmt): New function.
7987 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7988 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7990 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7992 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7993 Alan Hayward <alan.hayward@arm.com>
7994 David Sherwood <david.sherwood@arm.com>
7996 * doc/md.texi (extract_last_@var{m}): Document.
7997 * optabs.def (extract_last_optab): New optab.
7998 * internal-fn.def (EXTRACT_LAST): New internal function.
7999 * internal-fn.c (cond_unary_direct): New macro.
8000 (expand_cond_unary_optab_fn): Likewise.
8001 (direct_cond_unary_optab_supported_p): Likewise.
8002 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8003 loops using EXTRACT_LAST.
8004 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8005 (extract_last_<mode>): ...this optab.
8006 (vec_extract<mode><Vel>): Update accordingly.
8008 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8009 Alan Hayward <alan.hayward@arm.com>
8010 David Sherwood <david.sherwood@arm.com>
8012 * target.def (empty_mask_is_expensive): New hook.
8013 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8014 * doc/tm.texi: Regenerate.
8015 * targhooks.h (default_empty_mask_is_expensive): Declare.
8016 * targhooks.c (default_empty_mask_is_expensive): New function.
8017 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8018 if the target says that empty masks are expensive.
8019 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8021 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8023 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8024 Alan Hayward <alan.hayward@arm.com>
8025 David Sherwood <david.sherwood@arm.com>
8027 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8028 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8029 (vect_use_loop_mask_for_alignment_p): New function.
8030 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8031 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8032 niters_skip argument. Make sure that the first niters_skip elements
8033 of the first iteration are inactive.
8034 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8035 Update call to vect_set_loop_masks_directly.
8036 (get_misalign_in_elems): New function, split out from...
8037 (vect_gen_prolog_loop_niters): ...here.
8038 (vect_update_init_of_dr): Take a code argument that specifies whether
8039 the adjustment should be added or subtracted.
8040 (vect_update_init_of_drs): Likewise.
8041 (vect_prepare_for_masked_peels): New function.
8042 (vect_do_peeling): Skip prologue peeling if we're using a mask
8043 instead. Update call to vect_update_inits_of_drs.
8044 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8046 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8047 alignment. Do not include the number of peeled iterations in
8048 the minimum threshold in that case.
8049 (vectorizable_induction): Adjust the start value down by
8050 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8051 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8052 Take the number of skipped iterations into account when calculating
8054 * tree-vect-stmts.c (vect_gen_while_not): New function.
8056 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8057 Alan Hayward <alan.hayward@arm.com>
8058 David Sherwood <david.sherwood@arm.com>
8060 * doc/sourcebuild.texi (vect_fully_masked): Document.
8061 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8063 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8065 (vect_analyze_loop_2): ...here. Don't check the vectorization
8066 factor against the number of loop iterations if the loop is
8069 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8070 Alan Hayward <alan.hayward@arm.com>
8071 David Sherwood <david.sherwood@arm.com>
8073 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8074 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8075 (dump_groups): Update accordingly.
8076 (iv_use::mem_type): New member variable.
8077 (address_p): New function.
8078 (record_use): Add a mem_type argument and initialize the new
8080 (record_group_use): Add a mem_type argument. Use address_p.
8081 Remove obsolete null checks of base_object. Update call to record_use.
8082 (find_interesting_uses_op): Update call to record_group_use.
8083 (find_interesting_uses_cond): Likewise.
8084 (find_interesting_uses_address): Likewise.
8085 (get_mem_type_for_internal_fn): New function.
8086 (find_address_like_use): Likewise.
8087 (find_interesting_uses_stmt): Try find_address_like_use before
8088 calling find_interesting_uses_op.
8089 (addr_offset_valid_p): Use the iv mem_type field as the type
8090 of the addressed memory.
8091 (add_autoinc_candidates): Likewise.
8092 (get_address_cost): Likewise.
8093 (split_small_address_groups_p): Use address_p.
8094 (split_address_groups): Likewise.
8095 (add_iv_candidate_for_use): Likewise.
8096 (autoinc_possible_for_pair): Likewise.
8097 (rewrite_groups): Likewise.
8098 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8099 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8100 (get_alias_ptr_type_for_ptr_address): New function.
8101 (rewrite_use_address): Rewrite address uses in calls that were
8102 identified by find_address_like_use.
8104 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8105 Alan Hayward <alan.hayward@arm.com>
8106 David Sherwood <david.sherwood@arm.com>
8108 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8110 * gimple-expr.h (is_gimple_addressable: Likewise.
8111 * gimple-expr.c (is_gimple_address): Likewise.
8112 * internal-fn.c (expand_call_mem_ref): New function.
8113 (expand_mask_load_optab_fn): Use it.
8114 (expand_mask_store_optab_fn): Likewise.
8116 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8117 Alan Hayward <alan.hayward@arm.com>
8118 David Sherwood <david.sherwood@arm.com>
8120 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8121 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8122 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8123 (cond_umax@var{mode}): Document.
8124 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8125 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8126 (cond_umin_optab, cond_umax_optab): New optabs.
8127 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8128 (COND_IOR, COND_XOR): New internal functions.
8129 * internal-fn.h (get_conditional_internal_fn): Declare.
8130 * internal-fn.c (cond_binary_direct): New macro.
8131 (expand_cond_binary_optab_fn): Likewise.
8132 (direct_cond_binary_optab_supported_p): Likewise.
8133 (get_conditional_internal_fn): New function.
8134 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8135 Cope with reduction statements that are vectorized as calls rather
8137 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8138 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8139 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8140 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8141 (UNSPEC_COND_EOR): New unspecs.
8142 (optab): Add mappings for them.
8143 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8144 (sve_int_op, sve_fp_op): New int attributes.
8146 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8147 Alan Hayward <alan.hayward@arm.com>
8148 David Sherwood <david.sherwood@arm.com>
8150 * optabs.def (while_ult_optab): New optab.
8151 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8152 * internal-fn.def (WHILE_ULT): New internal function.
8153 * internal-fn.h (direct_internal_fn_supported_p): New override
8154 that takes two types as argument.
8155 * internal-fn.c (while_direct): New macro.
8156 (expand_while_optab_fn): New function.
8157 (convert_optab_supported_p): Likewise.
8158 (direct_while_optab_supported_p): New macro.
8159 * wide-int.h (wi::udiv_ceil): New function.
8160 * tree-vectorizer.h (rgroup_masks): New structure.
8161 (vec_loop_masks): New typedef.
8162 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8164 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8165 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8166 (vect_max_vf): New function.
8167 (slpeel_make_loop_iterate_ntimes): Delete.
8168 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8169 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8170 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8171 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8172 internal-fn.h, stor-layout.h and optabs-query.h.
8173 (vect_set_loop_mask): New function.
8174 (add_preheader_seq): Likewise.
8175 (add_header_seq): Likewise.
8176 (interleave_supported_p): Likewise.
8177 (vect_maybe_permute_loop_masks): Likewise.
8178 (vect_set_loop_masks_directly): Likewise.
8179 (vect_set_loop_condition_masked): Likewise.
8180 (vect_set_loop_condition_unmasked): New function, split out from
8181 slpeel_make_loop_iterate_ntimes.
8182 (slpeel_make_loop_iterate_ntimes): Rename to..
8183 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8184 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8185 (vect_do_peeling): Update call accordingly.
8186 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8188 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8189 mask_compare_type, can_fully_mask_p and fully_masked_p.
8190 (release_vec_loop_masks): New function.
8191 (_loop_vec_info): Use it to free the loop masks.
8192 (can_produce_all_loop_masks_p): New function.
8193 (vect_get_max_nscalars_per_iter): Likewise.
8194 (vect_verify_full_masking): Likewise.
8195 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8196 retries, and free the mask rgroups before retrying. Check loop-wide
8197 reasons for disallowing fully-masked loops. Make the final decision
8198 about whether use a fully-masked loop or not.
8199 (vect_estimate_min_profitable_iters): Do not assume that peeling
8200 for the number of iterations will be needed for fully-masked loops.
8201 (vectorizable_reduction): Disable fully-masked loops.
8202 (vectorizable_live_operation): Likewise.
8203 (vect_halve_mask_nunits): New function.
8204 (vect_double_mask_nunits): Likewise.
8205 (vect_record_loop_mask): Likewise.
8206 (vect_get_loop_mask): Likewise.
8207 (vect_transform_loop): Handle the case in which the final loop
8208 iteration might handle a partial vector. Call vect_set_loop_condition
8209 instead of slpeel_make_loop_iterate_ntimes.
8210 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8211 (check_load_store_masking): New function.
8212 (prepare_load_store_mask): Likewise.
8213 (vectorizable_store): Handle fully-masked loops.
8214 (vectorizable_load): Likewise.
8215 (supportable_widening_operation): Use vect_halve_mask_nunits for
8217 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8218 (vect_gen_while): New function.
8219 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8220 (aarch64_uqdec<mode>): New insn.
8222 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8223 Alan Hayward <alan.hayward@arm.com>
8224 David Sherwood <david.sherwood@arm.com>
8226 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8227 (reduc_xor_scal_optab): New optabs.
8228 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8229 (reduc_xor_scal_@var{m}): Document.
8230 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8231 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8233 * fold-const-call.c (fold_const_call): Handle them.
8234 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8235 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8236 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8237 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8238 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8239 (UNSPEC_XORV): New unspecs.
8240 (optab): Add entries for them.
8241 (BITWISEV): New int iterator.
8242 (bit_reduc_op): New int attributes.
8244 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8245 Alan Hayward <alan.hayward@arm.com>
8246 David Sherwood <david.sherwood@arm.com>
8248 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8249 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8250 * optabs.def (vec_shl_insert_optab): New optab.
8251 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8252 (duplicate_and_interleave): Likewise.
8253 * tree-vect-loop.c: Include internal-fn.h.
8254 (neutral_op_for_slp_reduction): New function, split out from
8255 get_initial_defs_for_reduction.
8256 (get_initial_def_for_reduction): Handle option 2 for variable-length
8257 vectors by loading the neutral value into a vector and then shifting
8258 the initial value into element 0.
8259 (get_initial_defs_for_reduction): Replace the code argument with
8260 the neutral value calculated by neutral_op_for_slp_reduction.
8261 Use gimple_build_vector for constant-length vectors.
8262 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8263 but the first group_size elements have a neutral value.
8264 Use duplicate_and_interleave otherwise.
8265 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8266 Update call to get_initial_defs_for_reduction. Handle SLP
8267 reductions for variable-length vectors by creating one vector
8268 result for each scalar result, with the elements associated
8269 with other scalar results stubbed out with the neutral value.
8270 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8271 Require IFN_VEC_SHL_INSERT for double reductions on
8272 variable-length vectors, or SLP reductions that have
8273 a neutral value. Require can_duplicate_and_interleave_p
8274 support for variable-length unchained SLP reductions if there
8275 is no neutral value, such as for MIN/MAX reductions. Also require
8276 the number of vector elements to be a multiple of the number of
8277 SLP statements when doing variable-length unchained SLP reductions.
8278 Update call to vect_create_epilog_for_reduction.
8279 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8280 and remove initial values.
8281 (duplicate_and_interleave): Make public.
8282 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8283 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8285 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8286 Alan Hayward <alan.hayward@arm.com>
8287 David Sherwood <david.sherwood@arm.com>
8289 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8290 (can_duplicate_and_interleave_p): New function.
8291 (vect_get_and_check_slp_defs): Take the vector of statements
8292 rather than just the current one. Remove excess parentheses.
8293 Restriction rejectinon of vect_constant_def and vect_external_def
8294 for variable-length vectors to boolean types, or types for which
8295 can_duplicate_and_interleave_p is false.
8296 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8297 (duplicate_and_interleave): New function.
8298 (vect_get_constant_vectors): Use gimple_build_vector for
8299 constant-length vectors and suitable variable-length constant
8300 vectors. Use duplicate_and_interleave for other variable-length
8301 vectors. Don't defer the update when inserting new statements.
8303 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8304 Alan Hayward <alan.hayward@arm.com>
8305 David Sherwood <david.sherwood@arm.com>
8307 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8308 min_profitable_iters doesn't go negative.
8310 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8311 Alan Hayward <alan.hayward@arm.com>
8312 David Sherwood <david.sherwood@arm.com>
8314 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8315 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8316 * optabs.def (vec_mask_load_lanes_optab): New optab.
8317 (vec_mask_store_lanes_optab): Likewise.
8318 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8319 (MASK_STORE_LANES): Likewise.
8320 * internal-fn.c (mask_load_lanes_direct): New macro.
8321 (mask_store_lanes_direct): Likewise.
8322 (expand_mask_load_optab_fn): Handle masked operations.
8323 (expand_mask_load_lanes_optab_fn): New macro.
8324 (expand_mask_store_optab_fn): Handle masked operations.
8325 (expand_mask_store_lanes_optab_fn): New macro.
8326 (direct_mask_load_lanes_optab_supported_p): Likewise.
8327 (direct_mask_store_lanes_optab_supported_p): Likewise.
8328 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8330 (vect_load_lanes_supported): Likewise.
8331 * tree-vect-data-refs.c (strip_conversion): New function.
8332 (can_group_stmts_p): Likewise.
8333 (vect_analyze_data_ref_accesses): Use it instead of checking
8334 for a pair of assignments.
8335 (vect_store_lanes_supported): Take a masked_p parameter.
8336 (vect_load_lanes_supported): Likewise.
8337 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8338 vect_store_lanes_supported and vect_load_lanes_supported.
8339 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8340 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8341 parameter. Don't allow gaps for masked accesses.
8342 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8343 and vect_load_lanes_supported.
8344 (get_load_store_type): Take a masked_p parameter and update
8345 call to get_group_load_store_type.
8346 (vectorizable_store): Update call to get_load_store_type.
8347 Handle IFN_MASK_STORE_LANES.
8348 (vectorizable_load): Update call to get_load_store_type.
8349 Handle IFN_MASK_LOAD_LANES.
8351 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8352 Alan Hayward <alan.hayward@arm.com>
8353 David Sherwood <david.sherwood@arm.com>
8355 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8357 * config/aarch64/aarch64-protos.h
8358 (aarch64_sve_struct_memory_operand_p): Declare.
8359 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8360 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8361 (VPRED, vpred): Handle SVE structure modes.
8362 * config/aarch64/constraints.md (Utx): New constraint.
8363 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8364 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8365 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8366 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8367 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8368 structure modes. Split into pieces after RA.
8369 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8370 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8372 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8373 SVE structure modes.
8374 (aarch64_classify_address): Likewise.
8375 (sizetochar): Move earlier in file.
8376 (aarch64_print_operand): Handle SVE register lists.
8377 (aarch64_array_mode): New function.
8378 (aarch64_sve_struct_memory_operand_p): Likewise.
8379 (TARGET_ARRAY_MODE): Redefine.
8381 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8382 Alan Hayward <alan.hayward@arm.com>
8383 David Sherwood <david.sherwood@arm.com>
8385 * target.def (array_mode): New target hook.
8386 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8387 * doc/tm.texi: Regenerate.
8388 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8389 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8390 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8392 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8395 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8396 Alan Hayward <alan.hayward@arm.com>
8397 David Sherwood <david.sherwood@arm.com>
8399 * fold-const.c (fold_binary_loc): Check the argument types
8400 rather than the result type when testing for a vector operation.
8402 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8404 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8405 * doc/tm.texi: Regenerate.
8407 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8408 Alan Hayward <alan.hayward@arm.com>
8409 David Sherwood <david.sherwood@arm.com>
8411 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8412 (sve): Document new AArch64 extension.
8413 * doc/md.texi (w): Extend the description of the AArch64
8414 constraint to include SVE vectors.
8415 (Upl, Upa): Document new AArch64 predicate constraints.
8416 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8418 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8419 (msve-vector-bits=): New option.
8420 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8421 SVE when these are disabled.
8422 (sve): New extension.
8423 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8424 modes. Adjust their number of units based on aarch64_sve_vg.
8425 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8426 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8427 aarch64_addr_query_type.
8428 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8429 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8430 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8431 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8432 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8433 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8434 (aarch64_simd_imm_zero_p): Delete.
8435 (aarch64_check_zero_based_sve_index_immediate): Declare.
8436 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8437 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8438 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8439 (aarch64_sve_float_mul_immediate_p): Likewise.
8440 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8442 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8443 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8444 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8445 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8446 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8447 (aarch64_regmode_natural_size): Likewise.
8448 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8449 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8451 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8452 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8453 for VG and the SVE predicate registers.
8454 (V_ALIASES): Add a "z"-prefixed alias.
8455 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8456 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8457 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8458 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8459 (REG_CLASS_NAMES): Add entries for them.
8460 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8461 and the predicate registers.
8462 (aarch64_sve_vg): Declare.
8463 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8464 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8465 (REGMODE_NATURAL_SIZE): Define.
8466 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8468 * config/aarch64/aarch64.c: Include cfgrtl.h.
8469 (simd_immediate_info): Add a constructor for series vectors,
8470 and an associated step field.
8471 (aarch64_sve_vg): New variable.
8472 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8473 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8474 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8475 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8476 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8477 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8478 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8479 (aarch64_get_mask_mode): New functions.
8480 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8481 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8482 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8483 predicate modes and predicate registers. Explicitly restrict
8484 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8485 to store a vector mode if it is recognized by
8486 aarch64_classify_vector_mode.
8487 (aarch64_regmode_natural_size): New function.
8488 (aarch64_hard_regno_caller_save_mode): Return the original mode
8490 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8491 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8492 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8493 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8495 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8496 does not overlap dest if the function is frame-related. Handle
8498 (aarch64_split_add_offset): New function.
8499 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8500 them aarch64_add_offset.
8501 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8502 and update call to aarch64_sub_sp.
8503 (aarch64_add_cfa_expression): New function.
8504 (aarch64_expand_prologue): Pass extra temporary registers to the
8505 functions above. Handle the case in which we need to emit new
8506 DW_CFA_expressions for registers that were originally saved
8507 relative to the stack pointer, but now have to be expressed
8508 relative to the frame pointer.
8509 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8511 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8512 IP0 and IP1 values for SVE frames.
8513 (aarch64_expand_vec_series): New function.
8514 (aarch64_expand_sve_widened_duplicate): Likewise.
8515 (aarch64_expand_sve_const_vector): Likewise.
8516 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8517 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8518 into the register, rather than emitting a SET directly.
8519 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8520 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8521 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8522 (offset_9bit_signed_scaled_p): New functions.
8523 (aarch64_replicate_bitmask_imm): New function.
8524 (aarch64_bitmask_imm): Use it.
8525 (aarch64_cannot_force_const_mem): Reject expressions involving
8526 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8527 (aarch64_classify_index): Handle SVE indices, by requiring
8528 a plain register index with a scale that matches the element size.
8529 (aarch64_classify_address): Handle SVE addresses. Assert that
8530 the mode of the address is VOIDmode or an integer mode.
8531 Update call to aarch64_classify_symbol.
8532 (aarch64_classify_symbolic_expression): Update call to
8533 aarch64_classify_symbol.
8534 (aarch64_const_vec_all_in_range_p): New function.
8535 (aarch64_print_vector_float_operand): Likewise.
8536 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8537 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8538 and the FP immediates 1.0 and 0.5.
8539 (aarch64_print_address_internal): Handle SVE addresses.
8540 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8541 (aarch64_regno_regclass): Handle predicate registers.
8542 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8544 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8545 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8546 (aarch64_convert_sve_vector_bits): New function.
8547 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8548 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8550 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8551 Handle SVE vector and predicate modes. Accept VL-based constants
8552 that need only one temporary register, and VL offsets that require
8553 no temporary registers.
8554 (aarch64_conditional_register_usage): Mark the predicate registers
8555 as fixed if SVE isn't available.
8556 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8557 Return true for SVE vector and predicate modes.
8558 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8559 rather than an unsigned int. Handle SVE modes.
8560 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8562 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8564 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8565 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8566 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8567 (aarch64_sve_float_mul_immediate_p): New functions.
8568 (aarch64_sve_valid_immediate): New function.
8569 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8570 Explicitly reject structure modes. Check for INDEX constants.
8571 Handle PTRUE and PFALSE constants.
8572 (aarch64_check_zero_based_sve_index_immediate): New function.
8573 (aarch64_simd_imm_zero_p): Delete.
8574 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8575 vector modes. Accept constants in the range of CNT[BHWD].
8576 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8577 ask for an Advanced SIMD mode.
8578 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8579 (aarch64_simd_vector_alignment): Handle SVE predicates.
8580 (aarch64_vectorize_preferred_vector_alignment): New function.
8581 (aarch64_simd_vector_alignment_reachable): Use it instead of
8583 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8584 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8586 (MAX_VECT_LEN): Delete.
8587 (expand_vec_perm_d): Add a vec_flags field.
8588 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8589 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8590 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8592 (aarch64_evpc_rev): Rename to...
8593 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8594 (aarch64_evpc_rev_global): New function.
8595 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8596 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8598 (aarch64_evpc_sve_tbl): New function.
8599 (aarch64_expand_vec_perm_const_1): Update after rename of
8600 aarch64_evpc_rev. Handle SVE permutes too, trying
8601 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8602 than aarch64_evpc_tbl.
8603 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8604 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8605 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8606 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8607 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8608 (aarch64_expand_sve_vcond): New functions.
8609 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8610 of aarch64_vector_mode_p.
8611 (aarch64_dwarf_poly_indeterminate_value): New function.
8612 (aarch64_compute_pressure_classes): Likewise.
8613 (aarch64_can_change_mode_class): Likewise.
8614 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8615 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8616 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8617 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8618 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8619 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8620 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8621 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8623 (Dn, Dl, Dr): Accept const as well as const_vector.
8624 (Dz): Likewise. Compare against CONST0_RTX.
8625 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8626 of "vector" where appropriate.
8627 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8628 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8629 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8630 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8631 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8632 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8633 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8634 (v_int_equiv): Extend to SVE modes.
8635 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8637 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8638 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8639 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8640 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8641 (SVE_COND_FP_CMP): New int iterators.
8642 (perm_hilo): Handle the new unpack unspecs.
8643 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8645 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8646 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8647 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8648 (aarch64_equality_operator, aarch64_constant_vector_operand)
8649 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8650 (aarch64_sve_nonimmediate_operand): Likewise.
8651 (aarch64_sve_general_operand): Likewise.
8652 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8653 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8654 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8655 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8656 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8657 (aarch64_sve_float_arith_immediate): Likewise.
8658 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8659 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8660 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8661 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8662 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8663 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8664 (aarch64_sve_float_arith_operand): Likewise.
8665 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8666 (aarch64_sve_float_mul_operand): Likewise.
8667 (aarch64_sve_vec_perm_operand): Likewise.
8668 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8669 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8670 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8671 as well as const_vector.
8672 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8673 in file. Use CONST0_RTX and CONSTM1_RTX.
8674 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8675 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8676 Use aarch64_simd_imm_zero.
8677 * config/aarch64/aarch64-sve.md: New file.
8678 * config/aarch64/aarch64.md: Include it.
8679 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8680 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8681 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8682 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8683 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8684 (sve): New attribute.
8685 (enabled): Disable instructions with the sve attribute unless
8687 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8688 aarch64_expand_mov_immediate.
8689 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8690 CNT[BHSD] immediates.
8691 (movti): Split CONST_POLY_INT moves into two halves.
8692 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8693 Split additions that need a temporary here if the destination
8694 is the stack pointer.
8695 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8696 (*add<mode>3_poly_1): New instruction.
8697 (set_clobber_cc): New expander.
8699 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8701 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8702 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8703 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8704 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8705 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8706 Change innermode from fixed_mode_size to machine_mode.
8707 (simplify_subreg): Update call accordingly. Handle a constant-sized
8708 subreg of a variable-length CONST_VECTOR.
8710 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8711 Alan Hayward <alan.hayward@arm.com>
8712 David Sherwood <david.sherwood@arm.com>
8714 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8715 (add_offset_to_base): New function, split out from...
8716 (create_mem_ref): ...here. When handling a scale other than 1,
8717 check first whether the address is valid without the offset.
8718 Add it into the base if so, leaving the index and scale as-is.
8720 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8723 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8724 fold_for_warn before checking if arg2 is INTEGER_CST.
8726 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8728 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8729 (store_multiple_operation): Delete.
8730 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8731 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8732 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8733 guarded by TARGET_STRING.
8734 (rs6000_output_load_multiple): Delete.
8735 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8736 OPTION_MASK_STRING / TARGET_STRING handling.
8737 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8738 (const rs6000_opt_masks) <"string">: Change mask to 0.
8739 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8740 (MASK_STRING): Delete.
8741 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8743 (load_multiple): Delete.
8750 (store_multiple): Delete.
8757 (movmemsi_8reg): Delete.
8758 (corresponding unnamed define_insn): Delete.
8759 (movmemsi_6reg): Delete.
8760 (corresponding unnamed define_insn): Delete.
8761 (movmemsi_4reg): Delete.
8762 (corresponding unnamed define_insn): Delete.
8763 (movmemsi_2reg): Delete.
8764 (corresponding unnamed define_insn): Delete.
8765 (movmemsi_1reg): Delete.
8766 (corresponding unnamed define_insn): Delete.
8767 * config/rs6000/rs6000.opt (mno-string): New.
8768 (mstring): Replace by deprecation warning stub.
8769 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8771 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8773 * regrename.c (regrename_do_replace): If replacing the same
8774 reg multiple times, try to reuse last created gen_raw_REG.
8777 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8778 main to workaround a bug in GDB.
8780 2018-01-12 Tom de Vries <tom@codesourcery.com>
8783 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8785 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8787 PR rtl-optimization/80481
8788 * ira-color.c (get_cap_member): New function.
8789 (allocnos_conflict_by_live_ranges_p): Use it.
8790 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8791 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8793 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8796 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8797 (*saddl_se_1): Ditto.
8799 (*ssubl_se_1): Ditto.
8801 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8803 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8804 rather than wi::to_widest for DR_INITs.
8805 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8806 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8807 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8809 (vect_analyze_group_access_1): Note that here.
8811 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8813 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8814 polynomial type sizes.
8816 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8818 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8819 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8820 (gimple_add_tmp_var): Likewise.
8822 2018-01-12 Martin Liska <mliska@suse.cz>
8824 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8825 (gimple_alloc_sizes): Likewise.
8826 (dump_gimple_statistics): Use PRIu64 in printf format.
8827 * gimple.h: Change uint64_t to int.
8829 2018-01-12 Martin Liska <mliska@suse.cz>
8831 * tree-core.h: Use uint64_t instead of int.
8832 * tree.c (tree_node_counts): Likewise.
8833 (tree_node_sizes): Likewise.
8834 (dump_tree_statistics): Use PRIu64 in printf format.
8836 2018-01-12 Martin Liska <mliska@suse.cz>
8838 * Makefile.in: As qsort_chk is implemented in vec.c, add
8839 vec.o to linkage of gencfn-macros.
8840 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8841 passing the info to record_node_allocation_statistics.
8842 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8844 * ggc-common.c (struct ggc_usage): Add operator== and use
8845 it in operator< and compare function.
8846 * mem-stats.h (struct mem_usage): Likewise.
8847 * vec.c (struct vec_usage): Remove operator< and compare
8848 function. Can be simply inherited.
8850 2018-01-12 Martin Jambor <mjambor@suse.cz>
8853 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8854 * tree-ssa-math-opts.c: Include domwalk.h.
8855 (convert_mult_to_fma_1): New function.
8856 (fma_transformation_info): New type.
8857 (fma_deferring_state): Likewise.
8858 (cancel_fma_deferring): New function.
8859 (result_of_phi): Likewise.
8860 (last_fma_candidate_feeds_initial_phi): Likewise.
8861 (convert_mult_to_fma): Added deferring logic, split actual
8862 transformation to convert_mult_to_fma_1.
8863 (math_opts_dom_walker): New type.
8864 (math_opts_dom_walker::after_dom_children): New method, body moved
8865 here from pass_optimize_widening_mul::execute, added deferring logic
8867 (pass_optimize_widening_mul::execute): Moved most of code to
8868 math_opts_dom_walker::after_dom_children.
8869 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8870 * config/i386/i386.c (ix86_option_override_internal): Added
8871 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8873 2018-01-12 Richard Biener <rguenther@suse.de>
8876 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8877 inline instance vars.
8879 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8882 * config/rx/rx.c (rx_is_restricted_memory_address):
8885 2018-01-12 Richard Biener <rguenther@suse.de>
8887 PR tree-optimization/80846
8888 * target.def (split_reduction): New target hook.
8889 * targhooks.c (default_split_reduction): New function.
8890 * targhooks.h (default_split_reduction): Declare.
8891 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8892 target requests first reduce vectors by combining low and high
8894 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8895 (get_vectype_for_scalar_type_and_size): Export.
8896 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8897 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8898 * doc/tm.texi: Regenerate.
8899 * config/i386/i386.c (ix86_split_reduction): Implement
8900 TARGET_VECTORIZE_SPLIT_REDUCTION.
8902 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8905 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8906 in PIC mode except for TARGET_VXWORKS_RTP.
8907 * config/sparc/sparc.c: Include cfgrtl.h.
8908 (TARGET_INIT_PIC_REG): Define.
8909 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8910 (sparc_pic_register_p): New predicate.
8911 (sparc_legitimate_address_p): Use it.
8912 (sparc_legitimize_pic_address): Likewise.
8913 (sparc_delegitimize_address): Likewise.
8914 (sparc_mode_dependent_address_p): Likewise.
8915 (gen_load_pcrel_sym): Remove 4th parameter.
8916 (load_got_register): Adjust call to above. Remove obsolete stuff.
8917 (sparc_expand_prologue): Do not call load_got_register here.
8918 (sparc_flat_expand_prologue): Likewise.
8919 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8920 (sparc_use_pseudo_pic_reg): New function.
8921 (sparc_init_pic_reg): Likewise.
8922 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8923 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8925 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8927 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8928 Add item for branch_cost.
8930 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8932 PR rtl-optimization/83565
8933 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8934 not extend the result to a larger mode for rotate operations.
8935 (num_sign_bit_copies1): Likewise.
8937 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8940 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8942 Use values-Xc.o for -pedantic.
8943 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8945 2018-01-12 Martin Liska <mliska@suse.cz>
8948 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8950 (possible_polymorphic_call_targets): Use it.
8951 (ipa_devirt): Likewise.
8953 2018-01-12 Martin Liska <mliska@suse.cz>
8955 * profile-count.h (enum profile_quality): Use 0 as invalid
8956 enum value of profile_quality.
8958 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8960 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8961 -mext-string options.
8963 2018-01-12 Richard Biener <rguenther@suse.de>
8965 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8966 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8967 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8969 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8971 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8973 * configure.ac (--with-long-double-format): Add support for the
8974 configuration option to change the default long double format on
8976 * config.gcc (powerpc*-linux*-*): Likewise.
8977 * configure: Regenerate.
8978 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8979 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8980 used without modification.
8982 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8984 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8985 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8986 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8987 MISC_BUILTIN_SPEC_BARRIER.
8988 (rs6000_init_builtins): Likewise.
8989 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8991 (speculation_barrier): New define_insn.
8992 * doc/extend.texi: Document __builtin_speculation_barrier.
8994 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8997 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8998 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8999 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9001 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9002 integral modes instead of "ss" and "sd".
9003 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9004 vectors with 32-bit and 64-bit elements.
9005 (vecdupssescalarmodesuffix): New mode attribute.
9006 (vec_dup<mode>): Use it.
9008 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9011 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9012 frame if argument is passed on stack.
9014 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9017 * ree.c (combine_reaching_defs): Optimize also
9018 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9019 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9021 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9024 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9026 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9029 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9030 after they are computed.
9032 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9034 PR tree-optimization/83695
9035 * gimple-loop-linterchange.cc
9036 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9037 reset cached scev information after interchange.
9038 (pass_linterchange::execute): Remove call to scev_reset_htab.
9040 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9042 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9043 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9044 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9045 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9046 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9047 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9048 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9049 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9050 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9051 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9052 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9053 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9054 (V_lane_reg): Likewise.
9055 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9057 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9058 (vfmal_lane_low<mode>_intrinsic,
9059 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9060 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9061 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9062 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9063 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9064 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9066 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9068 * config/arm/arm-cpus.in (fp16fml): New feature.
9069 (ALL_SIMD): Add fp16fml.
9070 (armv8.2-a): Add fp16fml as an option.
9071 (armv8.3-a): Likewise.
9072 (armv8.4-a): Add fp16fml as part of fp16.
9073 * config/arm/arm.h (TARGET_FP16FML): Define.
9074 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9076 * config/arm/arm-modes.def (V2HF): Define.
9077 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9078 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9079 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9080 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9081 vfmsl_low, vfmsl_high): New set of builtins.
9082 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9083 (vfml_op): New code attribute.
9084 (VFMLHALVES): New int iterator.
9085 (VFML, VFMLSEL): New mode attributes.
9086 (V_reg): Define mapping for V2HF.
9087 (V_hi, V_lo): New mode attributes.
9088 (VF_constraint): Likewise.
9089 (vfml_half, vfml_half_selector): New int attributes.
9090 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9092 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9093 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9095 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9096 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9097 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9098 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9100 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9101 Document new effective target and option set.
9103 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9105 * config/arm/arm-cpus.in (armv8_4): New feature.
9106 (ARMv8_4a): New fgroup.
9107 (armv8.4-a): New arch.
9108 * config/arm/arm-tables.opt: Regenerate.
9109 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9110 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9111 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9112 Add matching rules for -march=armv8.4-a and extensions.
9113 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9115 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9118 * config/rx/rx.md (BW): New mode attribute.
9119 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9121 2018-01-11 Richard Biener <rguenther@suse.de>
9123 PR tree-optimization/83435
9124 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9125 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9126 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9128 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9129 Alan Hayward <alan.hayward@arm.com>
9130 David Sherwood <david.sherwood@arm.com>
9132 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9134 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9135 (aarch64_print_address_internal): Use it to check for a zero offset.
9137 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9138 Alan Hayward <alan.hayward@arm.com>
9139 David Sherwood <david.sherwood@arm.com>
9141 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9142 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9143 Return a poly_int64 rather than a HOST_WIDE_INT.
9144 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9145 rather than a HOST_WIDE_INT.
9146 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9147 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9148 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9149 final_offset from HOST_WIDE_INT to poly_int64.
9150 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9151 to_constant when getting the number of units in an Advanced SIMD
9153 (aarch64_builtin_vectorized_function): Check for a constant number
9155 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9157 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9158 attribute instead of GET_MODE_NUNITS.
9159 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9160 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9161 GET_MODE_SIZE for fixed-size registers.
9162 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9163 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9164 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9165 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9166 (aarch64_print_operand, aarch64_print_address_internal)
9167 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9168 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9169 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9170 Handle polynomial GET_MODE_SIZE.
9171 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9172 wider than SImode without modification.
9173 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9174 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9175 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9176 passing and returning SVE modes.
9177 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9178 rather than GEN_INT.
9179 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9180 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9181 (aarch64_allocate_and_probe_stack_space): Likewise.
9182 (aarch64_layout_frame): Cope with polynomial offsets.
9183 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9184 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9186 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9187 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9188 poly_int64 rather than a HOST_WIDE_INT.
9189 (aarch64_get_separate_components, aarch64_process_components)
9190 (aarch64_expand_prologue, aarch64_expand_epilogue)
9191 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9192 (aarch64_anchor_offset): New function, split out from...
9193 (aarch64_legitimize_address): ...here.
9194 (aarch64_builtin_vectorization_cost): Handle polynomial
9195 TYPE_VECTOR_SUBPARTS.
9196 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9198 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9199 number of elements from the PARALLEL rather than the mode.
9200 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9201 rather than GET_MODE_BITSIZE.
9202 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9203 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9204 (aarch64_expand_vec_perm_const_1): Handle polynomial
9205 d->perm.length () and d->perm elements.
9206 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9207 Apply to_constant to d->perm elements.
9208 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9209 polynomial CONST_VECTOR_NUNITS.
9210 (aarch64_move_pointer): Take amount as a poly_int64 rather
9212 (aarch64_progress_pointer): Avoid temporary variable.
9213 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9214 the mode attribute instead of GET_MODE.
9216 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9217 Alan Hayward <alan.hayward@arm.com>
9218 David Sherwood <david.sherwood@arm.com>
9220 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9221 x exists before using it.
9222 (aarch64_add_constant_internal): Rename to...
9223 (aarch64_add_offset_1): ...this. Replace regnum with separate
9224 src and dest rtxes. Handle the case in which they're different,
9225 including when the offset is zero. Replace scratchreg with an rtx.
9226 Use 2 additions if there is no spare register into which we can
9227 move a 16-bit constant.
9228 (aarch64_add_constant): Delete.
9229 (aarch64_add_offset): Replace reg with separate src and dest
9230 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9231 Use aarch64_add_offset_1.
9232 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9233 an rtx rather than an int. Take the delta as a poly_int64
9234 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9235 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9236 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9237 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9238 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9240 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9241 aarch64_add_constant.
9243 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9245 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9246 Use scalar_float_mode.
9248 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9250 * config/aarch64/aarch64-simd.md
9251 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9252 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9253 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9254 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9255 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9256 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9257 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9258 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9259 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9260 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9262 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9265 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9266 targ_options->x_arm_arch_string is non NULL.
9268 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9270 * config/aarch64/aarch64.h
9271 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9273 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9276 * expmed.c (emit_store_flag_force): Swap if const op0
9277 and change VOIDmode to mode of op0.
9279 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9281 PR rtl-optimization/83761
9282 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9283 than bytes to mode_for_size.
9285 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9288 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9289 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9292 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9295 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9296 when in layout mode.
9297 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9298 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9301 2018-01-10 Michael Collison <michael.collison@arm.com>
9303 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9304 * config/aarch64/aarch64-option-extension.def: Add
9305 AARCH64_OPT_EXTENSION of 'fp16fml'.
9306 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9307 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9308 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9309 * config/aarch64/constraints.md (Ui7): New constraint.
9310 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9311 (VFMLA_SEL_W): Ditto.
9314 (VFMLA16_LOW): New int iterator.
9315 (VFMLA16_HIGH): Ditto.
9316 (UNSPEC_FMLAL): New unspec.
9317 (UNSPEC_FMLSL): Ditto.
9318 (UNSPEC_FMLAL2): Ditto.
9319 (UNSPEC_FMLSL2): Ditto.
9320 (f16mac): New code attribute.
9321 * config/aarch64/aarch64-simd-builtins.def
9322 (aarch64_fmlal_lowv2sf): Ditto.
9323 (aarch64_fmlsl_lowv2sf): Ditto.
9324 (aarch64_fmlalq_lowv4sf): Ditto.
9325 (aarch64_fmlslq_lowv4sf): Ditto.
9326 (aarch64_fmlal_highv2sf): Ditto.
9327 (aarch64_fmlsl_highv2sf): Ditto.
9328 (aarch64_fmlalq_highv4sf): Ditto.
9329 (aarch64_fmlslq_highv4sf): Ditto.
9330 (aarch64_fmlal_lane_lowv2sf): Ditto.
9331 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9332 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9333 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9334 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9335 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9336 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9337 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9338 (aarch64_fmlal_lane_highv2sf): Ditto.
9339 (aarch64_fmlsl_lane_highv2sf): Ditto.
9340 (aarch64_fmlal_laneq_highv2sf): Ditto.
9341 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9342 (aarch64_fmlalq_lane_highv4sf): Ditto.
9343 (aarch64_fmlsl_lane_highv4sf): Ditto.
9344 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9345 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9346 * config/aarch64/aarch64-simd.md:
9347 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9348 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9349 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9350 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9351 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9352 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9353 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9354 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9355 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9356 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9357 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9358 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9359 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9360 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9361 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9362 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9363 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9364 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9365 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9366 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9367 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9368 (vfmlsl_low_u32): Ditto.
9369 (vfmlalq_low_u32): Ditto.
9370 (vfmlslq_low_u32): Ditto.
9371 (vfmlal_high_u32): Ditto.
9372 (vfmlsl_high_u32): Ditto.
9373 (vfmlalq_high_u32): Ditto.
9374 (vfmlslq_high_u32): Ditto.
9375 (vfmlal_lane_low_u32): Ditto.
9376 (vfmlsl_lane_low_u32): Ditto.
9377 (vfmlal_laneq_low_u32): Ditto.
9378 (vfmlsl_laneq_low_u32): Ditto.
9379 (vfmlalq_lane_low_u32): Ditto.
9380 (vfmlslq_lane_low_u32): Ditto.
9381 (vfmlalq_laneq_low_u32): Ditto.
9382 (vfmlslq_laneq_low_u32): Ditto.
9383 (vfmlal_lane_high_u32): Ditto.
9384 (vfmlsl_lane_high_u32): Ditto.
9385 (vfmlal_laneq_high_u32): Ditto.
9386 (vfmlsl_laneq_high_u32): Ditto.
9387 (vfmlalq_lane_high_u32): Ditto.
9388 (vfmlslq_lane_high_u32): Ditto.
9389 (vfmlalq_laneq_high_u32): Ditto.
9390 (vfmlslq_laneq_high_u32): Ditto.
9391 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9392 (AARCH64_FL_FOR_ARCH8_4): New.
9393 (AARCH64_ISA_F16FML): New ISA flag.
9394 (TARGET_F16FML): New feature flag for fp16fml.
9395 (doc/invoke.texi): Document new fp16fml option.
9397 2018-01-10 Michael Collison <michael.collison@arm.com>
9399 * config/aarch64/aarch64-builtins.c:
9400 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9401 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9402 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9403 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9404 (AARCH64_ISA_SHA3): New ISA flag.
9405 (TARGET_SHA3): New feature flag for sha3.
9406 * config/aarch64/iterators.md (sha512_op): New int attribute.
9407 (CRYPTO_SHA512): New int iterator.
9408 (UNSPEC_SHA512H): New unspec.
9409 (UNSPEC_SHA512H2): Ditto.
9410 (UNSPEC_SHA512SU0): Ditto.
9411 (UNSPEC_SHA512SU1): Ditto.
9412 * config/aarch64/aarch64-simd-builtins.def
9413 (aarch64_crypto_sha512hqv2di): New builtin.
9414 (aarch64_crypto_sha512h2qv2di): Ditto.
9415 (aarch64_crypto_sha512su0qv2di): Ditto.
9416 (aarch64_crypto_sha512su1qv2di): Ditto.
9417 (aarch64_eor3qv8hi): Ditto.
9418 (aarch64_rax1qv2di): Ditto.
9419 (aarch64_xarqv2di): Ditto.
9420 (aarch64_bcaxqv8hi): Ditto.
9421 * config/aarch64/aarch64-simd.md:
9422 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9423 (aarch64_crypto_sha512su0qv2di): Ditto.
9424 (aarch64_crypto_sha512su1qv2di): Ditto.
9425 (aarch64_eor3qv8hi): Ditto.
9426 (aarch64_rax1qv2di): Ditto.
9427 (aarch64_xarqv2di): Ditto.
9428 (aarch64_bcaxqv8hi): Ditto.
9429 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9430 (vsha512h2q_u64): Ditto.
9431 (vsha512su0q_u64): Ditto.
9432 (vsha512su1q_u64): Ditto.
9433 (veor3q_u16): Ditto.
9434 (vrax1q_u64): Ditto.
9436 (vbcaxq_u16): Ditto.
9437 * config/arm/types.md (crypto_sha512): New type attribute.
9438 (crypto_sha3): Ditto.
9439 (doc/invoke.texi): Document new sha3 option.
9441 2018-01-10 Michael Collison <michael.collison@arm.com>
9443 * config/aarch64/aarch64-builtins.c:
9444 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9445 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9446 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9447 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9448 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9449 (AARCH64_ISA_SM4): New ISA flag.
9450 (TARGET_SM4): New feature flag for sm4.
9451 * config/aarch64/aarch64-simd-builtins.def
9452 (aarch64_sm3ss1qv4si): Ditto.
9453 (aarch64_sm3tt1aq4si): Ditto.
9454 (aarch64_sm3tt1bq4si): Ditto.
9455 (aarch64_sm3tt2aq4si): Ditto.
9456 (aarch64_sm3tt2bq4si): Ditto.
9457 (aarch64_sm3partw1qv4si): Ditto.
9458 (aarch64_sm3partw2qv4si): Ditto.
9459 (aarch64_sm4eqv4si): Ditto.
9460 (aarch64_sm4ekeyqv4si): Ditto.
9461 * config/aarch64/aarch64-simd.md:
9462 (aarch64_sm3ss1qv4si): Ditto.
9463 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9464 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9465 (aarch64_sm4eqv4si): Ditto.
9466 (aarch64_sm4ekeyqv4si): Ditto.
9467 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9468 (sm3part_op): Ditto.
9469 (CRYPTO_SM3TT): Ditto.
9470 (CRYPTO_SM3PART): Ditto.
9471 (UNSPEC_SM3SS1): New unspec.
9472 (UNSPEC_SM3TT1A): Ditto.
9473 (UNSPEC_SM3TT1B): Ditto.
9474 (UNSPEC_SM3TT2A): Ditto.
9475 (UNSPEC_SM3TT2B): Ditto.
9476 (UNSPEC_SM3PARTW1): Ditto.
9477 (UNSPEC_SM3PARTW2): Ditto.
9478 (UNSPEC_SM4E): Ditto.
9479 (UNSPEC_SM4EKEY): Ditto.
9480 * config/aarch64/constraints.md (Ui2): New constraint.
9481 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9482 * config/arm/types.md (crypto_sm3): New type attribute.
9483 (crypto_sm4): Ditto.
9484 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9485 (vsm3tt1aq_u32): Ditto.
9486 (vsm3tt1bq_u32): Ditto.
9487 (vsm3tt2aq_u32): Ditto.
9488 (vsm3tt2bq_u32): Ditto.
9489 (vsm3partw1q_u32): Ditto.
9490 (vsm3partw2q_u32): Ditto.
9491 (vsm4eq_u32): Ditto.
9492 (vsm4ekeyq_u32): Ditto.
9493 (doc/invoke.texi): Document new sm4 option.
9495 2018-01-10 Michael Collison <michael.collison@arm.com>
9497 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9498 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9499 (AARCH64_FL_FOR_ARCH8_4): New.
9500 (AARCH64_FL_V8_4): New flag.
9501 (doc/invoke.texi): Document new armv8.4-a option.
9503 2018-01-10 Michael Collison <michael.collison@arm.com>
9505 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9506 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9507 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9508 * config/aarch64/aarch64-option-extension.def: Add
9509 AARCH64_OPT_EXTENSION of 'sha2'.
9510 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9511 (crypto): Disable sha2 and aes if crypto disabled.
9512 (crypto): Enable aes and sha2 if enabled.
9513 (simd): Disable sha2 and aes if simd disabled.
9514 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9516 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9517 (TARGET_SHA2): New feature flag for sha2.
9518 (TARGET_AES): New feature flag for aes.
9519 * config/aarch64/aarch64-simd.md:
9520 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9521 conditional on TARGET_AES.
9522 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9523 (aarch64_crypto_sha1hsi): Make pattern conditional
9525 (aarch64_crypto_sha1hv4si): Ditto.
9526 (aarch64_be_crypto_sha1hv4si): Ditto.
9527 (aarch64_crypto_sha1su1v4si): Ditto.
9528 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9529 (aarch64_crypto_sha1su0v4si): Ditto.
9530 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9531 (aarch64_crypto_sha256su0v4si): Ditto.
9532 (aarch64_crypto_sha256su1v4si): Ditto.
9533 (doc/invoke.texi): Document new aes and sha2 options.
9535 2018-01-10 Martin Sebor <msebor@redhat.com>
9537 PR tree-optimization/83781
9538 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9541 2018-01-11 Martin Sebor <msebor@gmail.com>
9542 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9544 PR tree-optimization/83501
9545 PR tree-optimization/81703
9547 * tree-ssa-strlen.c (get_string_cst): Rename...
9548 (get_string_len): ...to this. Handle global constants.
9549 (handle_char_store): Adjust.
9551 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9552 Jim Wilson <jimw@sifive.com>
9554 * config/riscv/riscv-protos.h (riscv_output_return): New.
9555 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9556 (riscv_attribute_table, riscv_output_return),
9557 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9558 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9559 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9561 (riscv_expand_prologue): Add early return for naked function.
9562 (riscv_expand_epilogue): Likewise.
9563 (riscv_function_ok_for_sibcall): Return false for naked function.
9564 (riscv_set_current_function): New.
9565 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9566 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9567 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9568 * doc/extend.texi (RISC-V Function Attributes): New.
9570 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9572 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9573 check for 128-bit long double before checking TCmode.
9574 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9575 128-bit long doubles before checking TFmode or TCmode.
9576 (FLOAT128_IBM_P): Likewise.
9578 2018-01-10 Martin Sebor <msebor@redhat.com>
9580 PR tree-optimization/83671
9581 * builtins.c (c_strlen): Unconditionally return zero for the empty
9583 Use -Warray-bounds for warnings.
9584 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9585 for non-constant array indices with COMPONENT_REF, arrays of
9586 arrays, and pointers to arrays.
9587 (gimple_fold_builtin_strlen): Determine and set length range for
9588 non-constant character arrays.
9590 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9593 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9596 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9598 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9600 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9603 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9604 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9605 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9606 indexed_or_indirect_operand predicate.
9607 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9608 (*vsx_le_perm_load_v8hi): Likewise.
9609 (*vsx_le_perm_load_v16qi): Likewise.
9610 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9611 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9612 (*vsx_le_perm_store_v8hi): Likewise.
9613 (*vsx_le_perm_store_v16qi): Likewise.
9614 (eight unnamed splitters): Likewise.
9616 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9618 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9619 * config/rs6000/emmintrin.h: Likewise.
9620 * config/rs6000/mmintrin.h: Likewise.
9621 * config/rs6000/xmmintrin.h: Likewise.
9623 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9626 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9628 * tree.c (tree_nop_conversion): Return true for location wrapper
9630 (maybe_wrap_with_location): New function.
9631 (selftest::check_strip_nops): New function.
9632 (selftest::test_location_wrappers): New function.
9633 (selftest::tree_c_tests): Call it.
9634 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9635 (maybe_wrap_with_location): New decl.
9636 (EXPR_LOCATION_WRAPPER_P): New macro.
9637 (location_wrapper_p): New inline function.
9638 (tree_strip_any_location_wrapper): New inline function.
9640 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9643 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9644 stack_realign_offset for the largest alignment of stack slot
9646 (ix86_find_max_used_stack_alignment): New function.
9647 (ix86_finalize_stack_frame_flags): Use it. Set
9648 max_used_stack_alignment if we don't realign stack.
9649 * config/i386/i386.h (machine_function): Add
9650 max_used_stack_alignment.
9652 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9654 * config/arm/arm.opt (-mbranch-cost): New option.
9655 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9658 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9661 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9662 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9664 2018-01-10 Richard Biener <rguenther@suse.de>
9667 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9668 early out so it also covers the case where we have a non-NULL
9671 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9673 PR tree-optimization/83753
9674 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9675 for non-strided grouped accesses if the number of elements is 1.
9677 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9680 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9681 * i386.h (TARGET_USE_GATHER): Define.
9682 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9684 2018-01-10 Martin Liska <mliska@suse.cz>
9687 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9688 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9690 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9691 CLEANUP_NO_PARTITIONING is not set.
9693 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9695 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9696 for vectors, as a partial revert of r254296.
9697 * rtl.h (const_vec_p): Delete.
9698 (const_vec_duplicate_p): Don't test for vector CONSTs.
9699 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9700 * expmed.c (make_tree): Likewise.
9703 * common.md (E, F): Use CONSTANT_P instead of checking for
9705 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9706 checking for CONST_VECTOR.
9708 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9711 * predict.c (force_edge_cold): Handle in more sane way edges
9714 2018-01-09 Carl Love <cel@us.ibm.com>
9716 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9718 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9719 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9720 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9721 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9722 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9723 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9724 * config/rs6000/rs6000-protos.h: Add extern defition for
9725 rs6000_generate_float2_double_code.
9726 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9728 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9729 (float2_v2df): Add define_expand.
9731 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9734 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9735 op_mode in the force_to_mode call.
9737 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9739 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9740 instead of checking each element individually.
9741 (aarch64_evpc_uzp): Likewise.
9742 (aarch64_evpc_zip): Likewise.
9743 (aarch64_evpc_ext): Likewise.
9744 (aarch64_evpc_rev): Likewise.
9745 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9746 instead of checking each element individually. Return true without
9748 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9749 whether all selected elements come from the same input, instead of
9750 checking each element individually. Remove calls to gen_rtx_REG,
9751 start_sequence and end_sequence and instead assert that no rtl is
9754 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9756 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9757 order of HIGH and CONST checks.
9759 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9761 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9762 if the destination isn't an SSA_NAME.
9764 2018-01-09 Richard Biener <rguenther@suse.de>
9766 PR tree-optimization/83668
9767 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9769 (canonicalize_loop_form): ... here, renamed from ...
9770 (canonicalize_loop_closed_ssa_form): ... this and amended to
9771 swap successor edges for loop exit blocks to make us use
9772 the RPO order we need for initial schedule generation.
9774 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9776 PR tree-optimization/64811
9777 * match.pd: When optimizing comparisons with Inf, avoid
9778 introducing or losing exceptions from comparisons with NaN.
9780 2018-01-09 Martin Liska <mliska@suse.cz>
9783 * asan.c (shadow_mem_size): Add gcc_assert.
9785 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9787 Don't save registers in main().
9790 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9791 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9792 * config/avr/avr.c (avr_set_current_function): Don't error if
9793 naked, OS_task or OS_main are specified at the same time.
9794 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9796 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9798 * common/config/avr/avr-common.c (avr_option_optimization_table):
9799 Switch on -mmain-is-OS_task for optimizing compilations.
9801 2018-01-09 Richard Biener <rguenther@suse.de>
9803 PR tree-optimization/83572
9804 * graphite.c: Include cfganal.h.
9805 (graphite_transform_loops): Connect infinite loops to exit
9806 and remove fake edges at the end.
9808 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9810 * ipa-inline.c (edge_badness): Revert accidental checkin.
9812 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9815 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9816 symbols; not inline clones.
9818 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9821 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9822 hard registers. Formatting fixes.
9824 PR preprocessor/83722
9825 * gcc.c (try_generate_repro): Pass
9826 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9827 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9830 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9831 Kito Cheng <kito.cheng@gmail.com>
9833 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9834 (riscv_leaf_function_p): Delete.
9835 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9837 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9839 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9841 (do_ifelse): New function.
9842 (do_isel): New function.
9843 (do_sub3): New function.
9844 (do_add3): New function.
9845 (do_load_mask_compare): New function.
9846 (do_overlap_load_compare): New function.
9847 (expand_compare_loop): New function.
9848 (expand_block_compare): Call expand_compare_loop() when appropriate.
9849 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9851 (-mblock-compare-inline-loop-limit): New option.
9853 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9856 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9857 Reverse order of second and third operands in first alternative.
9858 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9859 of first and second elements in UNSPEC_VPERMR vector.
9860 (altivec_expand_vec_perm_le): Likewise.
9862 2018-01-08 Jeff Law <law@redhat.com>
9864 PR rtl-optimizatin/81308
9865 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9866 (process_switch): If group_case_labels makes a change, then set
9868 (pass_convert_switch::execute): If a switch is converted, then
9869 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9871 PR rtl-optimization/81308
9872 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9875 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9877 PR target/83663 - Revert r255946
9878 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9879 generation for cases where splatting a value is not useful.
9880 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9881 across a vec_duplicate and a paradoxical subreg forming a vector
9882 mode to a vec_concat.
9884 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9886 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9887 -march=armv8.3-a variants.
9888 * config/arm/t-multilib: Likewise.
9889 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9891 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9893 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9895 (cceq_ior_compare_complement): Give it a name so I can use it, and
9896 change boolean_or_operator predicate to boolean_operator so it can
9897 be used to generate a crand.
9898 (eqne): New code iterator.
9899 (bd/bd_neg): New code_attrs.
9900 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9901 a single define_insn.
9902 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9903 decrement (bdnzt/bdnzf/bdzt/bdzf).
9904 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9905 with the new names of the branch decrement patterns, and added the
9906 names of the branch decrement conditional patterns.
9908 2018-01-08 Richard Biener <rguenther@suse.de>
9910 PR tree-optimization/83563
9911 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9914 2018-01-08 Richard Biener <rguenther@suse.de>
9917 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9919 2018-01-08 Richard Biener <rguenther@suse.de>
9921 PR tree-optimization/83685
9922 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9923 references to abnormals.
9925 2018-01-08 Richard Biener <rguenther@suse.de>
9928 * dwarf2out.c (output_indirect_strings): Handle empty
9929 skeleton_debug_str_hash.
9930 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9932 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9934 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9935 (emit_store_direct): Likewise.
9936 (arc_trampoline_adjust_address): Likewise.
9937 (arc_asm_trampoline_template): New function.
9938 (arc_initialize_trampoline): Use asm_trampoline_template.
9939 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9940 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9941 * config/arc/arc.md (flush_icache): Delete pattern.
9943 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9945 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9946 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9949 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9952 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9953 by not USED_FOR_TARGET.
9954 (make_pass_resolve_sw_modes): Likewise.
9956 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9958 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9961 2018-01-08 Richard Biener <rguenther@suse.de>
9964 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9966 2018-01-08 Richard Biener <rguenther@suse.de>
9969 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9971 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9974 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9975 basic blocks with a small number of successors.
9976 (convert_control_dep_chain_into_preds): Improve handling of
9978 (dump_predicates): Split apart into...
9979 (dump_pred_chain): ...here...
9980 (dump_pred_info): ...and here.
9981 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9982 (can_chain_union_be_invalidated_p): Improve check for invalidation
9984 (uninit_uses_cannot_happen): Avoid unnecessary if
9985 convert_control_dep_chain_into_preds yielded nothing.
9987 2018-01-06 Martin Sebor <msebor@redhat.com>
9989 PR tree-optimization/83640
9990 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9991 subtracting negative offset from size.
9992 (builtin_access::overlap): Adjust offset bounds of the access to fall
9993 within the size of the object if possible.
9995 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9997 PR rtl-optimization/83699
9998 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9999 extract_bit_field_as_subreg to cases in which the extracted
10000 value is also a vector.
10002 * lra-constraints.c (process_alt_operands): Test for the equivalence
10003 substitutions when detecting a possible reload cycle.
10005 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10008 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10009 by default if flag_selective_schedling{,2}. Formatting fixes.
10011 PR rtl-optimization/83682
10012 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10013 if it has non-VECTOR_MODE element mode.
10014 (vec_duplicate_p): Likewise.
10016 PR middle-end/83694
10017 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10018 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10020 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10023 * config/i386/i386-builtin.def
10024 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10025 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10026 Require also OPTION_MASK_ISA_AVX512F in addition to
10027 OPTION_MASK_ISA_GFNI.
10028 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10029 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10030 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10031 to OPTION_MASK_ISA_GFNI.
10032 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10033 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10034 OPTION_MASK_ISA_AVX512BW.
10035 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10036 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10037 addition to OPTION_MASK_ISA_GFNI.
10038 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10039 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10040 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10041 to OPTION_MASK_ISA_GFNI.
10042 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10043 a requirement for all ISAs rather than any of them with a few
10045 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10047 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10048 bitmasks to be enabled with 3 exceptions, instead of requiring any
10049 enabled ISA with lots of exceptions.
10050 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10051 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10052 Change avx512bw in isa attribute to avx512f.
10053 * config/i386/sgxintrin.h: Add license boilerplate.
10054 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10055 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10056 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10057 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10059 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10060 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10061 temporarily sse2 rather than sse if not enabled already.
10064 * config/i386/sse.md (VI248_VLBW): Rename to ...
10065 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10066 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10067 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10068 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10069 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10070 mode iterator instead of VI248_VLBW.
10072 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10074 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10075 (record_modified): Skip clobbers; add debug output.
10076 (param_change_prob): Use sreal frequencies.
10078 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10080 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10081 punt for user-aligned variables.
10083 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10085 * tree-chrec.c (chrec_contains_symbols): Return true for
10088 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10091 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10092 of (x|y) == x for BICS pattern.
10094 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10096 PR tree-optimization/83605
10097 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10098 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10101 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10103 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10104 * config/epiphany/rtems.h: New file.
10106 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10107 Uros Bizjak <ubizjak@gmail.com>
10110 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10111 QIreg_operand instead of register_operand predicate.
10112 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10113 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10114 comments instead of -fmitigate[-_]rop.
10116 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10119 * cgraphunit.c (symbol_table::compile): Switch to text_section
10120 before calling assembly_start debug hook.
10121 * run-rtl-passes.c (run_rtl_passes): Likewise.
10124 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10126 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10127 range_int_cst_p rather than !symbolic_range_p before calling
10128 extract_range_from_multiplicative_op_1.
10130 2018-01-04 Jeff Law <law@redhat.com>
10132 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10133 redundant test in assertion.
10135 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10137 * doc/rtl.texi: Document machine_mode wrapper classes.
10139 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10141 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10142 using tree_to_uhwi.
10144 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10146 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10147 the VEC_PERM_EXPR fold to fail.
10149 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10152 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10153 to switched_sections.
10155 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10158 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10159 test for d.testing.
10161 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10164 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10165 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10167 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10170 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10171 is BLKmode and bitpos not zero or mode change is needed.
10173 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10176 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10179 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10182 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10183 instead of MULT rtx. Update all corresponding splitters.
10184 (*saddl_se): Ditto.
10185 (*ssub<modesuffix>): Ditto.
10186 (*ssubl_se): Ditto.
10187 (*cmp_sadd_di): Update split patterns.
10188 (*cmp_sadd_si): Ditto.
10189 (*cmp_sadd_sidi): Ditto.
10190 (*cmp_ssub_di): Ditto.
10191 (*cmp_ssub_si): Ditto.
10192 (*cmp_ssub_sidi): Ditto.
10193 * config/alpha/predicates.md (const23_operand): New predicate.
10194 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10195 Look for ASHIFT, not MULT inner operand.
10196 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10198 2018-01-04 Martin Liska <mliska@suse.cz>
10200 PR gcov-profile/83669
10201 * gcov.c (output_intermediate_file): Add version to intermediate
10203 * doc/gcov.texi: Document new field 'version' in intermediate
10204 file format. Fix location of '-k' option of gcov command.
10206 2018-01-04 Martin Liska <mliska@suse.cz>
10209 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10211 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10213 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10215 2018-01-03 Martin Sebor <msebor@redhat.com>
10217 PR tree-optimization/83655
10218 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10219 checking calls with invalid arguments.
10221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10223 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10224 (vectorizable_mask_load_store): Delete.
10225 (vectorizable_call): Return false for masked loads and stores.
10226 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10227 instead of gimple_assign_rhs1.
10228 (vectorizable_load): Handle IFN_MASK_LOAD.
10229 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10233 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10235 (vectorizable_mask_load_store): ...here.
10236 (vectorizable_load): ...and here.
10238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10240 * tree-vect-stmts.c (vect_build_all_ones_mask)
10241 (vect_build_zero_merge_argument): New functions, split out from...
10242 (vectorizable_load): ...here.
10244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10246 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10248 (vectorizable_mask_load_store): ...here.
10249 (vectorizable_store): ...and here.
10251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10253 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10255 (vectorizable_mask_load_store): ...here.
10257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10259 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10260 (vect_model_store_cost): Take a vec_load_store_type instead of a
10262 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10263 (vect_model_store_cost): Take a vec_load_store_type instead of a
10265 (vectorizable_mask_load_store): Update accordingly.
10266 (vectorizable_store): Likewise.
10267 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10269 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10271 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10272 IFN_MASK_LOAD calls here rather than...
10273 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
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 * expmed.c (extract_bit_field_1): For vector extracts,
10280 fall back to extract_bit_field_as_subreg if vec_extract
10283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10284 Alan Hayward <alan.hayward@arm.com>
10285 David Sherwood <david.sherwood@arm.com>
10287 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10288 they are variable or constant sized.
10289 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10290 slots for constant-sized data.
10292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10293 Alan Hayward <alan.hayward@arm.com>
10294 David Sherwood <david.sherwood@arm.com>
10296 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10297 handling COND_EXPRs with boolean comparisons, try to find a better
10298 basis for the mask type than the boolean itself.
10300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10302 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10303 is calculated and how it can be overridden.
10304 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10305 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10307 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10311 Alan Hayward <alan.hayward@arm.com>
10312 David Sherwood <david.sherwood@arm.com>
10314 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10315 Remove the mode argument.
10316 (aarch64_simd_valid_immediate): Remove the mode and inverse
10318 * config/aarch64/iterators.md (bitsize): New iterator.
10319 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10320 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10321 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10322 aarch64_simd_valid_immediate.
10323 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10324 (aarch64_reg_or_bic_imm): Likewise.
10325 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10326 with an insn_type enum and msl with a modifier_type enum.
10327 Replace element_width with a scalar_mode. Change the shift
10328 to unsigned int. Add constructors for scalar_float_mode and
10329 scalar_int_mode elements.
10330 (aarch64_vect_float_const_representable_p): Delete.
10331 (aarch64_can_const_movi_rtx_p)
10332 (aarch64_simd_scalar_immediate_valid_for_move)
10333 (aarch64_simd_make_constant): Update call to
10334 aarch64_simd_valid_immediate.
10335 (aarch64_advsimd_valid_immediate_hs): New function.
10336 (aarch64_advsimd_valid_immediate): Likewise.
10337 (aarch64_simd_valid_immediate): Remove mode and inverse
10338 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10339 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10340 and aarch64_float_const_representable_p on the result.
10341 (aarch64_output_simd_mov_immediate): Remove mode argument.
10342 Update call to aarch64_simd_valid_immediate and use of
10343 simd_immediate_info.
10344 (aarch64_output_scalar_simd_mov_immediate): Update call
10347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10348 Alan Hayward <alan.hayward@arm.com>
10349 David Sherwood <david.sherwood@arm.com>
10351 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10352 (mode_nunits): Likewise CONST_MODE_NUNITS.
10353 * machmode.def (ADJUST_NUNITS): Document.
10354 * genmodes.c (mode_data::need_nunits_adj): New field.
10355 (blank_mode): Update accordingly.
10356 (adj_nunits): New variable.
10357 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10359 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10360 listed in adj_nunits.
10361 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10362 listed in adj_nunits. Don't emit case statements for such modes.
10363 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10364 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10365 nothing if adj_nunits is nonnull.
10366 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10367 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10368 (emit_mode_fbit): Update use of print_maybe_const_decl.
10369 (emit_move_size): Likewise. Treat the array as non-const
10371 (emit_mode_adjustments): Handle adj_nunits.
10373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10375 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10376 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10377 (VECTOR_MODES): Use it.
10378 (make_vector_modes): Take the prefix as an argument.
10380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10381 Alan Hayward <alan.hayward@arm.com>
10382 David Sherwood <david.sherwood@arm.com>
10384 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10385 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10386 for MODE_VECTOR_BOOL.
10387 * machmode.def (VECTOR_BOOL_MODE): Document.
10388 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10389 (make_vector_bool_mode): New function.
10390 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10392 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10393 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10395 * stor-layout.c (int_mode_for_mode): Likewise.
10396 * tree.c (build_vector_type_for_mode): Likewise.
10397 * varasm.c (output_constant_pool_2): Likewise.
10398 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10399 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10400 for MODE_VECTOR_BOOL.
10401 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10402 of mode class checks.
10403 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10404 instead of a list of mode class checks.
10405 (expand_vector_scalar_condition): Likewise.
10406 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10409 Alan Hayward <alan.hayward@arm.com>
10410 David Sherwood <david.sherwood@arm.com>
10412 * machmode.h (mode_size): Change from unsigned short to
10414 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10415 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10416 or if measurement_type is not polynomial.
10417 (fixed_size_mode::includes_p): Check for constant-sized modes.
10418 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10419 return a poly_uint16 rather than an unsigned short.
10420 (emit_mode_size): Change the type of mode_size from unsigned short
10421 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10422 (emit_mode_adjustments): Cope with polynomial vector sizes.
10423 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10425 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10427 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10428 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10429 * caller-save.c (setup_save_areas): Likewise.
10430 (replace_reg_with_saved_mem): Likewise.
10431 * calls.c (emit_library_call_value_1): Likewise.
10432 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10433 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10434 (gen_lowpart_for_combine): Likewise.
10435 * convert.c (convert_to_integer_1): Likewise.
10436 * cse.c (equiv_constant, cse_insn): Likewise.
10437 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10438 (cselib_subst_to_values): Likewise.
10439 * dce.c (word_dce_process_block): Likewise.
10440 * df-problems.c (df_word_lr_mark_ref): Likewise.
10441 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10442 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10443 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10444 (rtl_for_decl_location): Likewise.
10445 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10446 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10447 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10448 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10449 (expand_expr_real_1): Likewise.
10450 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10451 (pad_below): Likewise.
10452 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10453 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10454 * ira.c (get_subreg_tracking_sizes): Likewise.
10455 * ira-build.c (ira_create_allocno_objects): Likewise.
10456 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10457 (ira_sort_regnos_for_alter_reg): Likewise.
10458 * ira-costs.c (record_operand_costs): Likewise.
10459 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10460 (resolve_simple_move): Likewise.
10461 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10462 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10463 (lra_constraints): Likewise.
10464 (CONST_POOL_OK_P): Reject variable-sized modes.
10465 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10466 (add_pseudo_to_slot, lra_spill): Likewise.
10467 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10468 * optabs-query.c (get_best_extraction_insn): Likewise.
10469 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10470 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10471 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10472 * recog.c (offsettable_address_addr_space_p): Likewise.
10473 * regcprop.c (maybe_mode_change): Likewise.
10474 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10475 * regrename.c (build_def_use): Likewise.
10476 * regstat.c (dump_reg_info): Likewise.
10477 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10478 (find_reloads, find_reloads_subreg_address): Likewise.
10479 * reload1.c (eliminate_regs_1): Likewise.
10480 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10481 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10482 (simplify_binary_operation_1, simplify_subreg): Likewise.
10483 * targhooks.c (default_function_arg_padding): Likewise.
10484 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10485 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10486 (verify_gimple_assign_ternary): Likewise.
10487 * tree-inline.c (estimate_move_cost): Likewise.
10488 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10489 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10490 (get_address_cost_ainc): Likewise.
10491 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10492 (vect_supportable_dr_alignment): Likewise.
10493 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10494 (vectorizable_reduction): Likewise.
10495 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10496 (vectorizable_operation, vectorizable_load): Likewise.
10497 * tree.c (build_same_sized_truth_vector_type): Likewise.
10498 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10499 * var-tracking.c (emit_note_insn_var_location): Likewise.
10500 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10501 (ADDR_VEC_ALIGN): Likewise.
10503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10504 Alan Hayward <alan.hayward@arm.com>
10505 David Sherwood <david.sherwood@arm.com>
10507 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10509 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10510 or if measurement_type is polynomial.
10511 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10512 * combine.c (make_extraction): Likewise.
10513 * dse.c (find_shift_sequence): Likewise.
10514 * dwarf2out.c (mem_loc_descriptor): Likewise.
10515 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10516 (extract_bit_field, extract_low_bits): Likewise.
10517 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10518 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10519 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10520 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10521 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10522 * reload.c (find_reloads): Likewise.
10523 * reload1.c (alter_reg): Likewise.
10524 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10525 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10526 * tree-if-conv.c (predicate_mem_writes): Likewise.
10527 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10528 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10529 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10530 * valtrack.c (dead_debug_insert_temp): Likewise.
10531 * varasm.c (mergeable_constant_section): Likewise.
10532 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10535 Alan Hayward <alan.hayward@arm.com>
10536 David Sherwood <david.sherwood@arm.com>
10538 * expr.c (expand_assignment): Cope with polynomial mode sizes
10539 when assigning to a CONCAT.
10541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10542 Alan Hayward <alan.hayward@arm.com>
10543 David Sherwood <david.sherwood@arm.com>
10545 * machmode.h (mode_precision): Change from unsigned short to
10547 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10549 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10550 or if measurement_type is not polynomial.
10551 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10552 in which the mode is already known to be a scalar_int_mode.
10553 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10554 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10556 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10557 for GET_MODE_PRECISION.
10558 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10559 for GET_MODE_PRECISION.
10560 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10562 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10563 (expand_field_assignment, make_extraction): Likewise.
10564 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10565 (get_last_value): Likewise.
10566 * convert.c (convert_to_integer_1): Likewise.
10567 * cse.c (cse_insn): Likewise.
10568 * expr.c (expand_expr_real_1): Likewise.
10569 * lra-constraints.c (simplify_operand_subreg): Likewise.
10570 * optabs-query.c (can_atomic_load_p): Likewise.
10571 * optabs.c (expand_atomic_load): Likewise.
10572 (expand_atomic_store): Likewise.
10573 * ree.c (combine_reaching_defs): Likewise.
10574 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10575 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10576 * tree.h (type_has_mode_precision_p): Likewise.
10577 * ubsan.c (instrument_si_overflow): Likewise.
10579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10580 Alan Hayward <alan.hayward@arm.com>
10581 David Sherwood <david.sherwood@arm.com>
10583 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10584 polynomial numbers of units.
10585 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10586 (valid_vector_subparts_p): New function.
10587 (build_vector_type): Remove temporary shim and take the number
10588 of units as a poly_uint64 rather than an int.
10589 (build_opaque_vector_type): Take the number of units as a
10590 poly_uint64 rather than an int.
10591 * tree.c (build_vector_from_ctor): Handle polynomial
10592 TYPE_VECTOR_SUBPARTS.
10593 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10594 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10595 (build_vector_from_val): If the number of units is variable,
10596 use build_vec_duplicate_cst for constant operands and
10597 VEC_DUPLICATE_EXPR otherwise.
10598 (make_vector_type): Remove temporary is_constant ().
10599 (build_vector_type, build_opaque_vector_type): Take the number of
10600 units as a poly_uint64 rather than an int.
10601 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10603 * cfgexpand.c (expand_debug_expr): Likewise.
10604 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10605 (store_constructor, expand_expr_real_1): Likewise.
10606 (const_scalar_mask_from_tree): Likewise.
10607 * fold-const-call.c (fold_const_reduction): Likewise.
10608 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10609 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10610 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10611 (fold_relational_const): Likewise.
10612 (native_interpret_vector): Likewise. Change the size from an
10613 int to an unsigned int.
10614 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10615 TYPE_VECTOR_SUBPARTS.
10616 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10617 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10618 duplicating a non-constant operand into a variable-length vector.
10619 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10620 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10621 * ipa-icf.c (sem_variable::equals): Likewise.
10622 * match.pd: Likewise.
10623 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10624 * print-tree.c (print_node): Likewise.
10625 * stor-layout.c (layout_type): Likewise.
10626 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10627 * tree-cfg.c (verify_gimple_comparison): Likewise.
10628 (verify_gimple_assign_binary): Likewise.
10629 (verify_gimple_assign_ternary): Likewise.
10630 (verify_gimple_assign_single): Likewise.
10631 * tree-pretty-print.c (dump_generic_node): Likewise.
10632 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10633 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10634 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10635 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10636 (vect_shift_permute_load_chain): Likewise.
10637 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10638 (expand_vector_condition, optimize_vector_constructor): Likewise.
10639 (lower_vec_perm, get_compute_type): Likewise.
10640 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10641 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10642 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10643 (vect_recog_mask_conversion_pattern): Likewise.
10644 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10645 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10646 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10647 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10648 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10649 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10650 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10651 (supportable_widening_operation): Likewise.
10652 (supportable_narrowing_operation): Likewise.
10653 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10655 * varasm.c (output_constant): Likewise.
10657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10658 Alan Hayward <alan.hayward@arm.com>
10659 David Sherwood <david.sherwood@arm.com>
10661 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10662 so that both the length == 3 and length != 3 cases set up their
10663 own permute vectors. Add comments explaining why we know the
10664 number of elements is constant.
10665 (vect_permute_load_chain): Likewise.
10667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10668 Alan Hayward <alan.hayward@arm.com>
10669 David Sherwood <david.sherwood@arm.com>
10671 * machmode.h (mode_nunits): Change from unsigned char to
10673 (ONLY_FIXED_SIZE_MODES): New macro.
10674 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10675 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10676 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10678 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10679 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10680 or if measurement_type is not polynomial.
10681 * genmodes.c (ZERO_COEFFS): New macro.
10682 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10684 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10685 Use ZERO_COEFFS when emitting initializers.
10686 * data-streamer.h (bp_pack_poly_value): New function.
10687 (bp_unpack_poly_value): Likewise.
10688 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10689 for GET_MODE_NUNITS.
10690 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10691 for GET_MODE_NUNITS.
10692 * tree.c (make_vector_type): Remove temporary shim and make
10693 the real function take the number of units as a poly_uint64
10694 rather than an int.
10695 (build_vector_type_for_mode): Handle polynomial nunits.
10696 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10697 * emit-rtl.c (const_vec_series_p_1): Likewise.
10698 (gen_rtx_CONST_VECTOR): Likewise.
10699 * fold-const.c (test_vec_duplicate_folding): Likewise.
10700 * genrecog.c (validate_pattern): Likewise.
10701 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10702 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10703 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10704 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10705 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10706 * rtlanal.c (subreg_get_info): Likewise.
10707 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10708 (vect_grouped_load_supported): Likewise.
10709 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10710 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10711 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10712 (simplify_const_unary_operation, simplify_binary_operation_1)
10713 (simplify_const_binary_operation, simplify_ternary_operation)
10714 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10715 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10716 instead of CONST_VECTOR_NUNITS.
10717 * varasm.c (output_constant_pool_2): Likewise.
10718 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10719 explicit-encoded elements in the XVEC for variable-length vectors.
10721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10723 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10726 Alan Hayward <alan.hayward@arm.com>
10727 David Sherwood <david.sherwood@arm.com>
10729 * coretypes.h (fixed_size_mode): Declare.
10730 (fixed_size_mode_pod): New typedef.
10731 * builtins.h (target_builtins::x_apply_args_mode)
10732 (target_builtins::x_apply_result_mode): Change type to
10733 fixed_size_mode_pod.
10734 * builtins.c (apply_args_size, apply_result_size, result_vector)
10735 (expand_builtin_apply_args_1, expand_builtin_apply)
10736 (expand_builtin_return): Update accordingly.
10738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10740 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10741 * cselib.c (cselib_hash_rtx): Likewise.
10742 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10743 CONST_VECTOR encoding.
10745 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10746 Jeff Law <law@redhat.com>
10749 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10750 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10751 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10752 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10755 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10756 explicitly probe *sp in a noreturn function if there were any callee
10757 register saves or frame pointer is needed.
10759 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10762 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10763 BLKmode for ternary, binary or unary expressions.
10766 * var-tracking.c (delete_vta_debug_insn): New inline function.
10767 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10768 insns from get_insns () to NULL instead of each bb separately.
10769 Use delete_vta_debug_insn. No longer static.
10770 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10771 delete_vta_debug_insns callers.
10772 * rtl.h (delete_vta_debug_insns): Declare.
10773 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10774 instead of variable_tracking_main.
10776 2018-01-03 Martin Sebor <msebor@redhat.com>
10778 PR tree-optimization/83603
10779 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10780 arguments past the endof the argument list in functions declared
10781 without a prototype.
10782 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10783 Avoid checking when arguments are null.
10785 2018-01-03 Martin Sebor <msebor@redhat.com>
10788 * doc/extend.texi (attribute const): Fix a typo.
10789 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10790 issuing -Wsuggest-attribute for void functions.
10792 2018-01-03 Martin Sebor <msebor@redhat.com>
10794 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10795 offset_int::from instead of wide_int::to_shwi.
10796 (maybe_diag_overlap): Remove assertion.
10797 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10798 * gimple-ssa-sprintf.c (format_directive): Same.
10799 (parse_directive): Same.
10800 (sprintf_dom_walker::compute_format_length): Same.
10801 (try_substitute_return_value): Same.
10803 2018-01-03 Jeff Law <law@redhat.com>
10805 PR middle-end/83654
10806 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10807 non-constant residual for zero at runtime and avoid probing in
10808 that case. Reorganize code for trailing problem to mirror handling
10811 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10813 PR tree-optimization/83501
10814 * tree-ssa-strlen.c (get_string_cst): New.
10815 (handle_char_store): Call get_string_cst.
10817 2018-01-03 Martin Liska <mliska@suse.cz>
10819 PR tree-optimization/83593
10820 * tree-ssa-strlen.c: Include tree-cfg.h.
10821 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10822 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10823 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10825 (strlen_dom_walker::before_dom_children): Call
10826 gimple_purge_dead_eh_edges. Dump tranformation with details
10828 (strlen_dom_walker::before_dom_children): Update call by adding
10829 new argument cleanup_eh.
10830 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10832 2018-01-03 Martin Liska <mliska@suse.cz>
10835 * cif-code.def (VARIADIC_THUNK): New enum value.
10836 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10839 2018-01-03 Jan Beulich <jbeulich@suse.com>
10841 * sse.md (mov<mode>_internal): Tighten condition for when to use
10842 vmovdqu<ssescalarsize> for TI and OI modes.
10844 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10846 Update copyright years.
10848 2018-01-03 Martin Liska <mliska@suse.cz>
10851 * ipa-visibility.c (function_and_variable_visibility): Skip
10852 functions with noipa attribure.
10854 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10856 * gcc.c (process_command): Update copyright notice dates.
10857 * gcov-dump.c (print_version): Ditto.
10858 * gcov.c (print_version): Ditto.
10859 * gcov-tool.c (print_version): Ditto.
10860 * gengtype.c (create_file): Ditto.
10861 * doc/cpp.texi: Bump @copying's copyright year.
10862 * doc/cppinternals.texi: Ditto.
10863 * doc/gcc.texi: Ditto.
10864 * doc/gccint.texi: Ditto.
10865 * doc/gcov.texi: Ditto.
10866 * doc/install.texi: Ditto.
10867 * doc/invoke.texi: Ditto.
10869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10871 * vector-builder.h (vector_builder::m_full_nelts): Change from
10872 unsigned int to poly_uint64.
10873 (vector_builder::full_nelts): Update prototype accordingly.
10874 (vector_builder::new_vector): Likewise.
10875 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10876 (vector_builder::operator ==): Likewise.
10877 (vector_builder::finalize): Likewise.
10878 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10879 Take the number of elements as a poly_uint64 rather than an
10881 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10882 from unsigned int to poly_uint64.
10883 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10884 (vec_perm_indices::new_vector): Likewise.
10885 (vec_perm_indices::length): Likewise.
10886 (vec_perm_indices::nelts_per_input): Likewise.
10887 (vec_perm_indices::input_nelts): Likewise.
10888 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10889 number of elements per input as a poly_uint64 rather than an
10890 unsigned int. Use the original encoding for variable-length
10891 vectors, rather than clamping each individual element.
10892 For the second and subsequent elements in each pattern,
10893 clamp the step and base before clamping their sum.
10894 (vec_perm_indices::series_p): Handle polynomial element counts.
10895 (vec_perm_indices::all_in_range_p): Likewise.
10896 (vec_perm_indices_to_tree): Likewise.
10897 (vec_perm_indices_to_rtx): Likewise.
10898 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10899 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10900 (tree_vector_builder::new_binary_operation): Handle polynomial
10901 element counts. Return false if we need to know the number
10902 of elements at compile time.
10903 * fold-const.c (fold_vec_perm): Punt if the number of elements
10904 isn't known at compile time.
10906 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10908 * vec-perm-indices.h (vec_perm_builder): Change element type
10909 from HOST_WIDE_INT to poly_int64.
10910 (vec_perm_indices::element_type): Update accordingly.
10911 (vec_perm_indices::clamp): Handle polynomial element_types.
10912 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10913 (vec_perm_indices::all_in_range_p): Likewise.
10914 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10916 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10917 polynomial vec_perm_indices element types.
10918 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10919 * fold-const.c (fold_vec_perm): Likewise.
10920 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10921 * tree-vect-generic.c (lower_vec_perm): Likewise.
10922 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10923 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10924 element type to HOST_WIDE_INT.
10926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10927 Alan Hayward <alan.hayward@arm.com>
10928 David Sherwood <david.sherwood@arm.com>
10930 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10931 rather than an int. Use plus_constant.
10932 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10933 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10936 Alan Hayward <alan.hayward@arm.com>
10937 David Sherwood <david.sherwood@arm.com>
10939 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10940 a HOST_WIDE_INT to a poly_int64.
10942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10943 Alan Hayward <alan.hayward@arm.com>
10944 David Sherwood <david.sherwood@arm.com>
10946 * calls.c (load_register_parameters): Cope with polynomial
10947 mode sizes. Require a constant size for BLKmode parameters
10948 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10949 forces a parameter to be padded at the lsb end in order to
10950 fill a complete number of words, require the parameter size
10951 to be ordered wrt UNITS_PER_WORD.
10953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10954 Alan Hayward <alan.hayward@arm.com>
10955 David Sherwood <david.sherwood@arm.com>
10957 * reload1.c (spill_stack_slot_width): Change element type
10958 from unsigned int to poly_uint64_pod.
10959 (alter_reg): Treat mode sizes as polynomial.
10961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10962 Alan Hayward <alan.hayward@arm.com>
10963 David Sherwood <david.sherwood@arm.com>
10965 * reload.c (complex_word_subreg_p): New function.
10966 (reload_inner_reg_of_subreg, push_reload): Use it.
10968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10969 Alan Hayward <alan.hayward@arm.com>
10970 David Sherwood <david.sherwood@arm.com>
10972 * lra-constraints.c (process_alt_operands): Reject matched
10973 operands whose sizes aren't ordered.
10974 (match_reload): Refer to this check here.
10976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10977 Alan Hayward <alan.hayward@arm.com>
10978 David Sherwood <david.sherwood@arm.com>
10980 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10981 that the mode size is in the set {1, 2, 4, 8, 16}.
10983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10984 Alan Hayward <alan.hayward@arm.com>
10985 David Sherwood <david.sherwood@arm.com>
10987 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10988 Use plus_constant instead of gen_rtx_PLUS.
10990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10991 Alan Hayward <alan.hayward@arm.com>
10992 David Sherwood <david.sherwood@arm.com>
10994 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10995 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10996 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10997 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10998 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10999 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11000 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11001 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11002 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11003 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11005 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11006 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11007 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11008 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11009 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11010 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11011 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11012 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11013 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11014 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11016 * expr.c (emit_move_resolve_push): Treat the input and result
11017 of PUSH_ROUNDING as a poly_int64.
11018 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11019 (emit_push_insn): Likewise.
11020 * lra-eliminations.c (mark_not_eliminable): Likewise.
11021 * recog.c (push_operand): Likewise.
11022 * reload1.c (elimination_effects): Likewise.
11023 * rtlanal.c (nonzero_bits1): Likewise.
11024 * calls.c (store_one_arg): Likewise. Require the padding to be
11025 known at compile time.
11027 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11028 Alan Hayward <alan.hayward@arm.com>
11029 David Sherwood <david.sherwood@arm.com>
11031 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11032 Use plus_constant instead of gen_rtx_PLUS.
11034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11035 Alan Hayward <alan.hayward@arm.com>
11036 David Sherwood <david.sherwood@arm.com>
11038 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11039 rather than an int.
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 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11046 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11047 via stack temporaries. Treat the mode size as polynomial too.
11049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11050 Alan Hayward <alan.hayward@arm.com>
11051 David Sherwood <david.sherwood@arm.com>
11053 * expr.c (expand_expr_real_2): When handling conversions involving
11054 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11055 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11056 as a poly_uint64 too.
11058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11059 Alan Hayward <alan.hayward@arm.com>
11060 David Sherwood <david.sherwood@arm.com>
11062 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11065 Alan Hayward <alan.hayward@arm.com>
11066 David Sherwood <david.sherwood@arm.com>
11068 * combine.c (can_change_dest_mode): Handle polynomial
11069 REGMODE_NATURAL_SIZE.
11070 * expmed.c (store_bit_field_1): Likewise.
11071 * expr.c (store_constructor): Likewise.
11072 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11073 and polynomial REGMODE_NATURAL_SIZE.
11074 (gen_lowpart_common): Likewise.
11075 * reginfo.c (record_subregs_of_mode): Likewise.
11076 * rtlanal.c (read_modify_subreg_p): Likewise.
11078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11079 Alan Hayward <alan.hayward@arm.com>
11080 David Sherwood <david.sherwood@arm.com>
11082 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11083 numbers of elements.
11085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11086 Alan Hayward <alan.hayward@arm.com>
11087 David Sherwood <david.sherwood@arm.com>
11089 * match.pd: Cope with polynomial numbers of vector elements.
11091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11092 Alan Hayward <alan.hayward@arm.com>
11093 David Sherwood <david.sherwood@arm.com>
11095 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11096 in a POINTER_PLUS_EXPR.
11098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11099 Alan Hayward <alan.hayward@arm.com>
11100 David Sherwood <david.sherwood@arm.com>
11102 * omp-simd-clone.c (simd_clone_subparts): New function.
11103 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11104 (ipa_simd_modify_function_body): Likewise.
11106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11107 Alan Hayward <alan.hayward@arm.com>
11108 David Sherwood <david.sherwood@arm.com>
11110 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11111 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11112 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11113 (expand_vector_condition, vector_element): Likewise.
11114 (subparts_gt): New function.
11115 (get_compute_type): Use subparts_gt.
11116 (count_type_subparts): Delete.
11117 (expand_vector_operations_1): Use subparts_gt instead of
11118 count_type_subparts.
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-data-refs.c (vect_no_alias_p): Replace with...
11125 (vect_compile_time_alias): ...this new function. Do the calculation
11126 on poly_ints rather than trees.
11127 (vect_prune_runtime_alias_test_list): Update call accordingly.
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-slp.c (vect_build_slp_tree_1): Handle polynomial
11135 (vect_schedule_slp_instance): Likewise.
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-slp.c (vect_get_and_check_slp_defs): Reject
11142 constant and extern definitions for variable-length vectors.
11143 (vect_get_constant_vectors): Note that the number of units
11144 is known to be constant.
11146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11147 Alan Hayward <alan.hayward@arm.com>
11148 David Sherwood <david.sherwood@arm.com>
11150 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11151 of units as polynomial. Choose between WIDE and NARROW based
11154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11155 Alan Hayward <alan.hayward@arm.com>
11156 David Sherwood <david.sherwood@arm.com>
11158 * tree-vect-stmts.c (simd_clone_subparts): New function.
11159 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11161 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11162 Alan Hayward <alan.hayward@arm.com>
11163 David Sherwood <david.sherwood@arm.com>
11165 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11166 vectors as polynomial. Use build_index_vector for
11167 IFN_GOMP_SIMD_LANE.
11169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11170 Alan Hayward <alan.hayward@arm.com>
11171 David Sherwood <david.sherwood@arm.com>
11173 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11174 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11175 for variable-length vectors.
11176 (vectorizable_mask_load_store): Treat the number of units as
11177 polynomial, asserting that it is constant if the condition has
11178 already been enforced.
11179 (vectorizable_store, vectorizable_load): Likewise.
11181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11182 Alan Hayward <alan.hayward@arm.com>
11183 David Sherwood <david.sherwood@arm.com>
11185 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11186 of units as polynomial. Punt if we can't tell at compile time
11187 which vector contains the final result.
11189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11190 Alan Hayward <alan.hayward@arm.com>
11191 David Sherwood <david.sherwood@arm.com>
11193 * tree-vect-loop.c (vectorizable_induction): Treat the number
11194 of units as polynomial. Punt on SLP inductions. Use an integer
11195 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11196 cast of such a series for variable-length floating-point
11199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11200 Alan Hayward <alan.hayward@arm.com>
11201 David Sherwood <david.sherwood@arm.com>
11203 * tree.h (build_index_vector): Declare.
11204 * tree.c (build_index_vector): New function.
11205 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11206 of units as polynomial, forcibly converting it to a constant if
11207 vectorizable_reduction has already enforced the condition.
11208 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11209 to create a {1,2,3,...} vector.
11210 (vectorizable_reduction): Treat the number of units as polynomial.
11211 Choose vectype_in based on the largest scalar element size rather
11212 than the smallest number of units. Enforce the restrictions
11215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11216 Alan Hayward <alan.hayward@arm.com>
11217 David Sherwood <david.sherwood@arm.com>
11219 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11220 number of units as polynomial.
11222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11223 Alan Hayward <alan.hayward@arm.com>
11224 David Sherwood <david.sherwood@arm.com>
11226 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11227 * target.def (autovectorize_vector_sizes): Return the vector sizes
11228 by pointer, using vector_sizes rather than a bitmask.
11229 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11230 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11231 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11233 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11234 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11235 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11236 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11237 * omp-general.c (omp_max_vf): Likewise.
11238 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11239 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11240 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11241 * tree-vect-slp.c (vect_slp_bb): Likewise.
11242 * doc/tm.texi: Regenerate.
11243 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11245 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11246 the vector size as a poly_uint64 rather than an unsigned int.
11247 (current_vector_size): Change from an unsigned int to a poly_uint64.
11248 (get_vectype_for_scalar_type): Update accordingly.
11249 * tree.h (build_truth_vector_type): Take the size and number of
11250 units as a poly_uint64 rather than an unsigned int.
11251 (build_vector_type): Add a temporary overload that takes
11252 the number of units as a poly_uint64 rather than an unsigned int.
11253 * tree.c (make_vector_type): Likewise.
11254 (build_truth_vector_type): Take the number of units as a poly_uint64
11255 rather than an unsigned int.
11257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11258 Alan Hayward <alan.hayward@arm.com>
11259 David Sherwood <david.sherwood@arm.com>
11261 * target.def (get_mask_mode): Take the number of units and length
11262 as poly_uint64s rather than unsigned ints.
11263 * targhooks.h (default_get_mask_mode): Update accordingly.
11264 * targhooks.c (default_get_mask_mode): Likewise.
11265 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11266 * doc/tm.texi: Regenerate.
11268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11269 Alan Hayward <alan.hayward@arm.com>
11270 David Sherwood <david.sherwood@arm.com>
11272 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11273 * omp-general.c (omp_max_vf): Likewise.
11274 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11275 (expand_omp_simd): Handle polynomial safelen.
11276 * omp-low.c (omplow_simd_context): Add a default constructor.
11277 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11278 (lower_rec_simd_input_clauses): Update accordingly.
11279 (lower_rec_input_clauses): Likewise.
11281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11282 Alan Hayward <alan.hayward@arm.com>
11283 David Sherwood <david.sherwood@arm.com>
11285 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11286 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11287 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11288 (vect_analyze_slp_cost): Likewise.
11289 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11290 (vect_model_load_cost): Likewise.
11292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11293 Alan Hayward <alan.hayward@arm.com>
11294 David Sherwood <david.sherwood@arm.com>
11296 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11297 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11298 from an unsigned int * to a poly_uint64_pod *.
11299 (calculate_unrolling_factor): New function.
11300 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11303 Alan Hayward <alan.hayward@arm.com>
11304 David Sherwood <david.sherwood@arm.com>
11306 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11307 from an unsigned int to a poly_uint64.
11308 (_loop_vec_info::slp_unrolling_factor): Likewise.
11309 (_loop_vec_info::vectorization_factor): Change from an int
11311 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11312 (vect_get_num_vectors): New function.
11313 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11314 (vect_get_num_copies): Use vect_get_num_vectors.
11315 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11316 to an unsigned int *.
11317 (vect_analyze_data_refs): Change min_vf from an int * to a
11319 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11320 than an unsigned HOST_WIDE_INT.
11321 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11322 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11323 to an unsigned int *.
11324 (vect_analyze_data_ref_dependences): Likewise.
11325 (vect_compute_data_ref_alignment): Handle polynomial vf.
11326 (vect_enhance_data_refs_alignment): Likewise.
11327 (vect_prune_runtime_alias_test_list): Likewise.
11328 (vect_shift_permute_load_chain): Likewise.
11329 (vect_supportable_dr_alignment): Likewise.
11330 (dependence_distance_ge_vf): Take the vectorization factor as a
11331 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11332 (vect_analyze_data_refs): Change min_vf from an int * to a
11334 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11335 vfm1 as a poly_uint64 rather than an int. Make the same change
11336 for the returned bound_scalar.
11337 (vect_gen_vector_loop_niters): Handle polynomial vf.
11338 (vect_do_peeling): Likewise. Update call to
11339 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11340 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11342 * tree-vect-loop.c (vect_determine_vectorization_factor)
11343 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11344 (vect_get_known_peeling_cost): Likewise.
11345 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11346 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11347 (vect_transform_loop): Likewise. Use the lowest possible VF when
11348 updating the upper bounds of the loop.
11349 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11350 rather than an int.
11351 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11352 polynomial unroll factors.
11353 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11354 (vect_make_slp_decision): Likewise.
11355 (vect_supported_load_permutation_p): Likewise, and polynomial
11357 (vect_analyze_slp_cost): Handle polynomial vf.
11358 (vect_slp_analyze_node_operations): Likewise.
11359 (vect_slp_analyze_bb_1): Likewise.
11360 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11361 than an unsigned HOST_WIDE_INT.
11362 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11363 (vectorizable_load): Handle polynomial vf.
11364 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11366 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11369 Alan Hayward <alan.hayward@arm.com>
11370 David Sherwood <david.sherwood@arm.com>
11372 * match.pd: Handle bit operations involving three constants
11373 and try to fold one pair.
11375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11377 * tree-vect-loop-manip.c: Include gimple-fold.h.
11378 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11379 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11380 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11381 Add a path that uses a step of VF instead of 1, but disable it
11383 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11384 and niters_no_overflow parameters. Update calls to
11385 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11386 Create a new SSA name if the latter choses to use a ste other
11387 than zero, and return it via niters_vector_mult_vf_var.
11388 * tree-vect-loop.c (vect_transform_loop): Update calls to
11389 vect_do_peeling, vect_gen_vector_loop_niters and
11390 slpeel_make_loop_iterate_ntimes.
11391 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11392 (vect_gen_vector_loop_niters): Update declarations after above changes.
11394 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11396 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11397 128-bit round to integer instructions.
11398 (ceil<mode>2): Likewise.
11399 (btrunc<mode>2): Likewise.
11400 (round<mode>2): Likewise.
11402 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11404 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11405 unaligned VSX load/store on P8/P9.
11406 (expand_block_clear): Allow the use of unaligned VSX
11407 load/store on P8/P9.
11409 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11411 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11413 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11414 swap associated with both a load and a store.
11416 2018-01-02 Andrew Waterman <andrew@sifive.com>
11418 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11419 * config/riscv/riscv.md (clear_cache): Use it.
11421 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11423 * web.c: Remove out-of-date comment.
11425 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11427 * expr.c (fixup_args_size_notes): Check that any existing
11428 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11429 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11430 (emit_single_push_insn): ...here.
11432 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11434 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11435 (const_vector_encoded_nelts): New function.
11436 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11437 (const_vector_int_elt, const_vector_elt): Declare.
11438 * emit-rtl.c (const_vector_int_elt_1): New function.
11439 (const_vector_elt): Likewise.
11440 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11441 of CONST_VECTOR_ELT.
11443 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11445 * expr.c: Include rtx-vector-builder.h.
11446 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11447 directly on the tree encoding.
11448 (const_vector_from_tree): Likewise.
11449 * optabs.c: Include rtx-vector-builder.h.
11450 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11451 sequence of "u" values.
11452 * vec-perm-indices.c: Include rtx-vector-builder.h.
11453 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11454 directly on the vec_perm_indices encoding.
11456 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11458 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11459 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11460 * rtx-vector-builder.h: New file.
11461 * rtx-vector-builder.c: Likewise.
11462 * rtl.h (rtx_def::u2): Add a const_vector field.
11463 (CONST_VECTOR_NPATTERNS): New macro.
11464 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11465 (CONST_VECTOR_DUPLICATE_P): Likewise.
11466 (CONST_VECTOR_STEPPED_P): Likewise.
11467 (CONST_VECTOR_ENCODED_ELT): Likewise.
11468 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11469 (unwrap_const_vec_duplicate): Likewise.
11470 (const_vec_series_p): Check for a non-duplicated vector encoding.
11471 Say that the function only returns true for integer vectors.
11472 * emit-rtl.c: Include rtx-vector-builder.h.
11473 (gen_const_vec_duplicate_1): Delete.
11474 (gen_const_vector): Call gen_const_vec_duplicate instead of
11475 gen_const_vec_duplicate_1.
11476 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11477 (gen_const_vec_duplicate): Use rtx_vector_builder.
11478 (gen_const_vec_series): Likewise.
11479 (gen_rtx_CONST_VECTOR): Likewise.
11480 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11481 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11482 Build a new vector rather than modifying a CONST_VECTOR in-place.
11483 (handle_special_swappables): Update call accordingly.
11484 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11485 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11486 Build a new vector rather than modifying a CONST_VECTOR in-place.
11487 (handle_special_swappables): Update call accordingly.
11489 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11491 * simplify-rtx.c (simplify_const_binary_operation): Use
11492 CONST_VECTOR_ELT instead of XVECEXP.
11494 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11496 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11497 the selector elements to be different from the data elements
11498 if the selector is a VECTOR_CST.
11499 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11500 ssizetype for the selector.
11502 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11504 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11505 before testing each element individually.
11506 * tree-vect-generic.c (lower_vec_perm): Likewise.
11508 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11510 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11511 * selftest-run-tests.c (selftest::run_tests): Call it.
11512 * vector-builder.h (vector_builder::operator ==): New function.
11513 (vector_builder::operator !=): Likewise.
11514 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11515 (vec_perm_indices::all_from_input_p): New function.
11516 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11517 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11518 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11519 instead of reading the VECTOR_CST directly. Detect whether both
11520 vector inputs are the same before constructing the vec_perm_indices,
11521 and update the number of inputs argument accordingly. Use the
11522 utility functions added above. Only construct sel2 if we need to.
11524 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11526 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11527 the broadcast of the low byte.
11528 (expand_mult_highpart): Use an explicit encoding for the permutes.
11529 * optabs-query.c (can_mult_highpart_p): Likewise.
11530 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11531 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11532 (vectorizable_bswap): Likewise.
11533 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11534 explicit encoding for the power-of-2 permutes.
11535 (vect_permute_store_chain): Likewise.
11536 (vect_grouped_load_supported): Likewise.
11537 (vect_permute_load_chain): Likewise.
11539 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11541 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11542 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11543 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11544 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11545 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11546 (vect_gen_perm_mask_any): Likewise.
11548 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11550 * int-vector-builder.h: New file.
11551 * vec-perm-indices.h: Include int-vector-builder.h.
11552 (vec_perm_indices): Redefine as an int_vector_builder.
11553 (auto_vec_perm_indices): Delete.
11554 (vec_perm_builder): Redefine as a stand-alone class.
11555 (vec_perm_indices::vec_perm_indices): New function.
11556 (vec_perm_indices::clamp): Likewise.
11557 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11558 (vec_perm_indices::new_vector): New function.
11559 (vec_perm_indices::new_expanded_vector): Update for new
11560 vec_perm_indices class.
11561 (vec_perm_indices::rotate_inputs): New function.
11562 (vec_perm_indices::all_in_range_p): Operate directly on the
11563 encoded form, without computing elided elements.
11564 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11565 encoding. Update for new vec_perm_indices class.
11566 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11567 the given vec_perm_builder.
11568 (expand_vec_perm_var): Update vec_perm_builder constructor.
11569 (expand_mult_highpart): Use vec_perm_builder instead of
11570 auto_vec_perm_indices.
11571 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11572 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11573 or double series encoding as appropriate.
11574 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11575 vec_perm_indices instead of auto_vec_perm_indices.
11576 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11577 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11578 (vect_permute_store_chain): Likewise.
11579 (vect_grouped_load_supported): Likewise.
11580 (vect_permute_load_chain): Likewise.
11581 (vect_shift_permute_load_chain): Likewise.
11582 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11583 (vect_transform_slp_perm_load): Likewise.
11584 (vect_schedule_slp_instance): Likewise.
11585 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11586 (vectorizable_mask_load_store): Likewise.
11587 (vectorizable_bswap): Likewise.
11588 (vectorizable_store): Likewise.
11589 (vectorizable_load): Likewise.
11590 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11591 vec_perm_indices instead of auto_vec_perm_indices. Use
11592 tree_to_vec_perm_builder to read the vector from a tree.
11593 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11594 vec_perm_builder instead of a vec_perm_indices.
11595 (have_whole_vector_shift): Use vec_perm_builder and
11596 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11597 truncation to calc_vec_perm_mask_for_shift.
11598 (vect_create_epilog_for_reduction): Likewise.
11599 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11600 from auto_vec_perm_indices to vec_perm_indices.
11601 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11602 instead of changing individual elements.
11603 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11604 the vector in d.perm.
11605 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11606 from auto_vec_perm_indices to vec_perm_indices.
11607 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11608 instead of changing individual elements.
11609 (arm_vectorize_vec_perm_const): Use new_vector to install
11610 the vector in d.perm.
11611 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11612 Update vec_perm_builder constructor.
11613 (rs6000_expand_interleave): Likewise.
11614 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11615 (rs6000_expand_interleave): Likewise.
11617 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11619 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11620 to qimode could truncate the indices.
11621 * optabs.c (expand_vec_perm_var): Likewise.
11623 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11625 * Makefile.in (OBJS): Add vec-perm-indices.o.
11626 * vec-perm-indices.h: New file.
11627 * vec-perm-indices.c: Likewise.
11628 * target.h (vec_perm_indices): Replace with a forward class
11630 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11631 * optabs.h: Include vec-perm-indices.h.
11632 (expand_vec_perm): Delete.
11633 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11634 (expand_vec_perm_const): Declare.
11635 * target.def (vec_perm_const_ok): Replace with...
11636 (vec_perm_const): ...this new hook.
11637 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11638 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11639 * doc/tm.texi: Regenerate.
11640 * optabs.def (vec_perm_const): Delete.
11641 * doc/md.texi (vec_perm_const): Likewise.
11642 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11643 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11644 expand_vec_perm for constant permutation vectors. Assert that
11645 the mode of variable permutation vectors is the integer equivalent
11646 of the mode that is being permuted.
11647 * optabs-query.h (selector_fits_mode_p): Declare.
11648 * optabs-query.c: Include vec-perm-indices.h.
11649 (selector_fits_mode_p): New function.
11650 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11651 is defined, instead of checking whether the vec_perm_const_optab
11652 exists. Use targetm.vectorize.vec_perm_const instead of
11653 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11654 fit in the vector mode before using a variable permute.
11655 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11656 vec_perm_indices instead of an rtx.
11657 (expand_vec_perm): Replace with...
11658 (expand_vec_perm_const): ...this new function. Take the selector
11659 as a vec_perm_indices rather than an rtx. Also take the mode of
11660 the selector. Update call to shift_amt_for_vec_perm_mask.
11661 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11662 Use vec_perm_indices::new_expanded_vector to expand the original
11663 selector into bytes. Check whether the indices fit in the vector
11664 mode before using a variable permute.
11665 (expand_vec_perm_var): Make global.
11666 (expand_mult_highpart): Use expand_vec_perm_const.
11667 * fold-const.c: Includes vec-perm-indices.h.
11668 * tree-ssa-forwprop.c: Likewise.
11669 * tree-vect-data-refs.c: Likewise.
11670 * tree-vect-generic.c: Likewise.
11671 * tree-vect-loop.c: Likewise.
11672 * tree-vect-slp.c: Likewise.
11673 * tree-vect-stmts.c: Likewise.
11674 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11676 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11677 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11678 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11679 (aarch64_vectorize_vec_perm_const): ...this new function.
11680 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11681 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11682 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11683 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11684 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11685 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11686 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11688 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11689 check for NEON modes.
11690 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11691 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11692 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11693 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11695 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11696 the old VEC_PERM_CONST conditions.
11697 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11698 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11699 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11700 (ia64_vectorize_vec_perm_const_ok): Merge into...
11701 (ia64_vectorize_vec_perm_const): ...this new function.
11702 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11703 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11704 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11705 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11706 * config/mips/mips.c (mips_expand_vec_perm_const)
11707 (mips_vectorize_vec_perm_const_ok): Merge into...
11708 (mips_vectorize_vec_perm_const): ...this new function.
11709 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11710 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11711 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11712 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11713 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11714 (rs6000_expand_vec_perm_const): Delete.
11715 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11717 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11718 (altivec_expand_vec_perm_const_le): Take each operand individually.
11719 Operate on constant selectors rather than rtxes.
11720 (altivec_expand_vec_perm_const): Likewise. Update call to
11721 altivec_expand_vec_perm_const_le.
11722 (rs6000_expand_vec_perm_const): Delete.
11723 (rs6000_vectorize_vec_perm_const_ok): Delete.
11724 (rs6000_vectorize_vec_perm_const): New function.
11725 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11726 an element count and rtx array.
11727 (rs6000_expand_extract_even): Update call accordingly.
11728 (rs6000_expand_interleave): Likewise.
11729 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11730 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11731 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11732 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11733 (rs6000_expand_vec_perm_const): Delete.
11734 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11735 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11736 (altivec_expand_vec_perm_const_le): Take each operand individually.
11737 Operate on constant selectors rather than rtxes.
11738 (altivec_expand_vec_perm_const): Likewise. Update call to
11739 altivec_expand_vec_perm_const_le.
11740 (rs6000_expand_vec_perm_const): Delete.
11741 (rs6000_vectorize_vec_perm_const_ok): Delete.
11742 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11743 reference to the SPE evmerge intructions.
11744 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11745 an element count and rtx array.
11746 (rs6000_expand_extract_even): Update call accordingly.
11747 (rs6000_expand_interleave): Likewise.
11748 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11749 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11751 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11753 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11755 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11756 vector mode and that that mode matches the mode of the data
11758 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11759 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11760 directly using expand_vec_perm_1 when forcing selectors into
11762 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11764 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11766 * optabs-query.h (can_vec_perm_p): Delete.
11767 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11768 * optabs-query.c (can_vec_perm_p): Split into...
11769 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11770 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11771 particular selector is valid.
11772 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11773 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11774 (vect_grouped_load_supported): Likewise.
11775 (vect_shift_permute_load_chain): Likewise.
11776 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11777 (vect_transform_slp_perm_load): Likewise.
11778 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11779 (vectorizable_bswap): Likewise.
11780 (vect_gen_perm_mask_checked): Likewise.
11781 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11782 implementations of variable permutation vectors into account
11783 when deciding which selector to use.
11784 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11785 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11786 with a false third argument.
11787 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11788 to test whether the constant selector is valid and can_vec_perm_var_p
11789 to test whether a variable selector is valid.
11791 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11793 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11794 * optabs-query.c (can_vec_perm_p): Likewise.
11795 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11796 instead of vec_perm_indices.
11797 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11798 (vect_gen_perm_mask_checked): Likewise,
11799 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11800 (vect_gen_perm_mask_checked): Likewise,
11802 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11804 * optabs-query.h (qimode_for_vec_perm): Declare.
11805 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11806 (qimode_for_vec_perm): ...this new function.
11807 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11809 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11811 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11812 does not have a conditional at the top.
11814 2018-01-02 Richard Biener <rguenther@suse.de>
11816 * ipa-inline.c (big_speedup_p): Fix expression.
11818 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11821 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11824 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11828 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11829 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11830 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11831 cond_taken_branch_cost 3->4.
11833 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11835 PR tree-optimization/83581
11836 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11837 TODO_cleanup_cfg if any changes have been made.
11839 PR middle-end/83608
11840 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11841 convert_modes if target mode has the right side, but different mode
11844 PR middle-end/83609
11845 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11846 last argument when extracting from CONCAT. If either from_real or
11847 from_imag is NULL, use expansion through memory. If result is not
11848 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11849 the parts directly to inner mode, if even that fails, use expansion
11852 PR middle-end/83623
11853 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11854 check for bswap in mode rather than HImode and use that in expand_unop
11857 Copyright (C) 2018 Free Software Foundation, Inc.
11859 Copying and distribution of this file, with or without modification,
11860 are permitted in any medium without royalty provided the copyright
11861 notice and this notice are preserved.