1 2020-03-01 Martin Sebor <msebor@redhat.com>
4 * calls.c (init_attr_rdwr_indices): Correctly handle attribute.
6 2020-03-01 Martin Sebor <msebor@redhat.com>
9 * tree-ssa-strlen.c (count_nonzero_bytes): Set the size to that
10 of a pointer in the outermost ADDR_EXPRs.
12 2020-02-28 Jeff Law <law@redhat.com>
14 * config/v850/v850.h (STATIC_CHAIN_REGNUM): Change to r19.
15 * config/v850/v850.c (v850_asm_trampoline_template): Update
18 2020-02-28 Michael Meissner <meissner@linux.ibm.com>
21 * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
24 2020-02-28 Martin Liska <mliska@suse.cz>
27 * configure.ac: Improve detection of ld_date by requiring
28 either two dashes or none.
29 * configure: Regenerate.
31 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
33 PR rtl-optimization/93564
34 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
35 do not honor reg alloc order.
37 2020-02-27 Joel Hutton <Joel.Hutton@arm.com>
40 * config/aarch64/aarch64.c (aarch64_override_options): Fix
41 misleading warning string.
43 2020-02-27 Martin Sebor <msebor@redhat.com>
45 * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
47 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
50 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
51 Split the insn into two parts. This insn only does variable
52 extract from a register.
53 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
54 variable extract from memory.
55 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
56 only does variable extract from a register.
57 (vsx_extract_v4sf_var_load): New insn, do variable extract from
59 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
60 into two parts. This insn only does variable extract from a
62 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
63 do variable extract from memory.
65 2020-02-27 Martin Jambor <mjambor@suse.cz>
66 Feng Xue <fxue@os.amperecomputing.com>
69 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
70 new function calls_same_node_or_its_all_contexts_clone_p.
71 (cgraph_edge_brings_value_p): Use it.
72 (cgraph_edge_brings_value_p): Likewise.
73 (self_recursive_pass_through_p): Return false if caller is a clone.
74 (self_recursive_agg_pass_through_p): Likewise.
76 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
79 * alias.c (ends_tbaa_access_path_p): Break out from ...
80 (component_uses_parent_alias_set_from): ... here.
81 * alias.h (ends_tbaa_access_path_p): Declare.
82 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
83 handle trailing arrays past end of tbaa access path.
84 (aliasing_component_refs_p): ... here; likewise.
85 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
86 path; disambiguate also past end of it.
87 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
90 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
92 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
93 beginning of the file.
94 (vcreate_bf16, vcombine_bf16): New.
95 (vdup_n_bf16, vdupq_n_bf16): New.
96 (vdup_lane_bf16, vdup_laneq_bf16): New.
97 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
98 (vduph_lane_bf16, vduph_laneq_bf16): New.
99 (vset_lane_bf16, vsetq_lane_bf16): New.
100 (vget_lane_bf16, vgetq_lane_bf16): New.
101 (vget_high_bf16, vget_low_bf16): New.
102 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
103 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
104 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
105 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
106 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
107 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
108 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
109 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
110 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
111 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
112 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
113 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
114 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
115 (vreinterpretq_bf16_p128): New.
116 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
117 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
118 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
119 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
120 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
121 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
122 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
123 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
124 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
125 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
126 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
127 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
128 (vreinterpretq_p128_bf16): New.
129 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
131 (V_elem_l): Likewise.
134 (V_DOUBLE): Likewise.
135 (VDQX): Add V4BF and V8BF.
136 (V_two_elem, V_three_elem, V_four_elem): Likewise.
139 (V_double_vector_mode): Likewise.
140 (V_cmp_result): Likewise.
141 (V_uf_sclr): Likewise.
142 (V_sz_elem): Likewise.
143 (Is_d_reg): Likewise.
144 (V_mode_nunits): Likewise.
145 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
147 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
149 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
150 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
151 (<expander><mode>3<exec>): Likewise.
152 (<expander><mode>3): New.
153 (v<expander><mode>3): New.
154 (<expander><mode>3): New.
155 (<expander><mode>3<exec>): Rename to ...
156 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
157 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
159 2020-02-27 Alexandre Oliva <oliva@adacore.com>
161 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
164 2020-02-27 Richard Biener <rguenther@suse.de>
166 PR tree-optimization/93508
167 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
168 non-_CHK variants. Valueize their length arguments.
170 2020-02-27 Richard Biener <rguenther@suse.de>
172 PR tree-optimization/93953
173 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
174 to the hash-map entry.
176 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
178 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
180 2020-02-27 Mark Williams <mwilliams@fb.com>
182 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
183 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
184 -ffile-prefix-map and -fmacro-prefix-map.
185 * lto-streamer-out.c: Include file-prefix-map.h.
186 (lto_output_location): Remap the file part of locations.
188 2020-02-27 Jakub Jelinek <jakub@redhat.com>
191 * gimplify.c (gimplify_init_constructor): Don't promote readonly
192 DECL_REGISTER variables to TREE_STATIC.
194 PR tree-optimization/93582
195 PR tree-optimization/93945
196 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
197 non-zero INTEGER_CST second argument and ref->offset or ref->size
198 not a multiple of BITS_PER_UNIT.
200 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
202 * doc/install.texi (Binaries): Update description of BullFreeware.
204 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
208 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
209 C++ Language Options, Warning Options, and Static Analyzer
210 Options lists. Document negative form of options enabled by
211 default. Move some things around to more accurately sort
212 warnings by category.
213 (C++ Dialect Options, Warning Options, Static Analyzer
214 Options): Document negative form of options when enabled by
215 default. Move some things around to more accurately sort
216 warnings by category. Add some missing index entries.
219 2020-02-26 Carl Love <cel@us.ibm.com>
222 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
223 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
224 for the vector unsigned short arguments. It is also listed as the
225 name of the built-in for arguments vector unsigned short,
226 vector unsigned int and vector unsigned long long built-ins. The
227 name of the builtins for these arguments should be:
228 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
229 __builtin_crypto_vpmsumd respectively.
231 2020-02-26 Richard Biener <rguenther@suse.de>
233 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
234 and load permutation.
236 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
239 * optabs-tree.c (supportable_convert_operation): Reject types with
242 2020-02-26 David Malcolm <dmalcolm@redhat.com>
244 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
246 2020-02-26 Jakub Jelinek <jakub@redhat.com>
248 PR tree-optimization/93820
249 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
250 argument to ALL_INTEGER_CST_P boolean.
251 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
252 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
253 adjacent INTEGER_CST store into merged_store->only_constants like
256 2020-02-25 Jakub Jelinek <jakub@redhat.com>
259 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
261 * cfghooks.c (verify_flow_info): Likewise.
262 * predict.c (combine_predictions_for_bb): Likewise.
263 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
264 sucessor -> successor.
265 (find_traces_1_round): Fix comment typo, destinarion -> destination.
266 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
268 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
269 message typo, sucessors -> successors.
271 2020-02-25 Martin Sebor <msebor@redhat.com>
273 * doc/extend.texi (attribute access): Correct an example.
275 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
277 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
279 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
281 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
282 (VD): Enable for V4BF.
284 (VQ): Enable for V8BF.
287 (VDBL, Vdbl): Add V4BF.
288 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
289 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
290 (bfloat16x8x2_t): Likewise.
291 (bfloat16x4x3_t): Likewise.
292 (bfloat16x8x3_t): Likewise.
293 (bfloat16x4x4_t): Likewise.
294 (bfloat16x8x4_t): Likewise.
295 (vcombine_bf16): New.
296 (vld1_bf16, vld1_bf16_x2): New.
297 (vld1_bf16_x3, vld1_bf16_x4): New.
298 (vld1q_bf16, vld1q_bf16_x2): New.
299 (vld1q_bf16_x3, vld1q_bf16_x4): New.
300 (vld1_lane_bf16): New.
301 (vld1q_lane_bf16): New.
302 (vld1_dup_bf16): New.
303 (vld1q_dup_bf16): New.
306 (vld2_dup_bf16): New.
307 (vld2q_dup_bf16): New.
310 (vld3_dup_bf16): New.
311 (vld3q_dup_bf16): New.
314 (vld4_dup_bf16): New.
315 (vld4q_dup_bf16): New.
316 (vst1_bf16, vst1_bf16_x2): New.
317 (vst1_bf16_x3, vst1_bf16_x4): New.
318 (vst1q_bf16, vst1q_bf16_x2): New.
319 (vst1q_bf16_x3, vst1q_bf16_x4): New.
320 (vst1_lane_bf16): New.
321 (vst1q_lane_bf16): New.
329 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
331 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
332 (VALL_F16): Likewise.
333 (VALLDI_F16): Likewise.
336 (vswap_width_name): Likewise.
337 (VSWAP_WIDTH): Likewise.
341 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
342 (vget_lane_bf16, vgetq_lane_bf16): New.
344 (vdup_n_bf16, vdupq_n_bf16): New.
345 (vdup_lane_bf16, vdup_laneq_bf16): New.
346 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
347 (vduph_lane_bf16, vduph_laneq_bf16): New.
348 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
349 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
350 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
351 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
352 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
353 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
354 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
355 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
356 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
357 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
358 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
359 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
360 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
361 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
362 (vreinterpretq_bf16_p128): New.
363 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
364 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
365 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
366 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
367 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
368 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
369 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
370 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
371 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
372 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
373 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
374 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
375 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
376 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
377 (vreinterpretq_p128_bf16): New.
379 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
381 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
382 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
383 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
384 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
385 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
386 * config/arm/iterators.md (VSF2BF): New attribute.
387 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
388 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
389 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
391 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
393 * config/arm/arm.md (required_for_purecode): New attribute.
394 (enabled): Handle required_for_purecode.
395 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
396 work with -mpure-code.
398 2020-02-25 Jakub Jelinek <jakub@redhat.com>
400 PR rtl-optimization/93908
401 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
404 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
406 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
408 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
410 * doc/install.texi (--enable-checking): Adjust wording.
412 2020-02-25 Richard Biener <rguenther@suse.de>
414 PR tree-optimization/93868
415 * tree-vect-slp.c (slp_copy_subtree): New function.
416 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
417 re-arranging stmts in it.
419 2020-02-25 Jakub Jelinek <jakub@redhat.com>
422 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
423 dummy function and remove it at the end.
426 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
427 paramter -> parameter.
428 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
429 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
431 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
433 * doc/install.texi (--enable-checking): Properly document current
435 (--enable-stage1-checking): Minor clarification about bootstrap.
437 2020-02-24 David Malcolm <dmalcolm@redhat.com>
440 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
441 -fanalyzer-checker=taint is also required.
442 (-fanalyzer-checker=): Note that providing this option enables the
443 given checker, and doing so may be required for checkers that are
446 2020-02-24 David Malcolm <dmalcolm@redhat.com>
448 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
449 significant control flow events; add a "3" which shows all
450 control flow events; the old "3" becomes "4".
452 2020-02-24 Jakub Jelinek <jakub@redhat.com>
454 PR tree-optimization/93582
455 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
456 pd.offset and pd.size to be counted in bits rather than bytes, add
457 support for maxsizei that is not a multiple of BITS_PER_UNIT and
458 handle bitfield stores and loads.
459 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
460 uncomparable quantities - bytes vs. bits. Allow push_partial_def
461 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
462 pd.offset/pd.size to be counted in bits rather than bytes.
463 Formatting fix. Rename shadowed len variable to buflen.
465 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
466 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
469 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
470 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
471 * opts-common.c (parse_options_from_collect_gcc_options): New function.
472 (prepend_xassembler_to_collect_as_options): Likewise.
473 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
474 (prepend_xassembler_to_collect_as_options): Likewise.
475 * lto-opts.c (lto_write_options): Stream assembler options
476 in COLLECT_AS_OPTIONS.
477 * lto-wrapper.c (xassembler_options_error): New static variable.
478 (get_options_from_collect_gcc_options): Move parsing options code to
479 parse_options_from_collect_gcc_options and call it.
480 (merge_and_complain): Validate -Xassembler options.
481 (append_compiler_options): Handle OPT_Xassembler.
482 (run_gcc): Append command line -Xassembler options to
484 * doc/invoke.texi: Add documentation about using Xassembler
487 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
489 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
491 (riscv_rtx_costs): Update cost model for LTGT.
493 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
495 PR rtl-optimization/93564
496 * ira-color.c (struct update_cost_queue_elem): New member start.
497 (queue_update_cost, get_next_update_cost): Add new arg start.
498 (allocnos_conflict_p): New function.
499 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
500 Add checking conflicts with allocnos_conflict_p.
501 (update_costs_from_prefs, restore_costs_from_copies): Adjust
502 update_costs_from_allocno calls.
503 (update_conflict_hard_regno_costs): Add checking conflicts with
504 allocnos_conflict_p. Adjust calls of queue_update_cost and
505 get_next_update_cost.
506 (assign_hard_reg): Adjust calls of queue_update_cost. Add
508 (bucket_allocno_compare_func): Restore previous version.
510 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
512 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
513 double-word size when handling aggregate return values.
514 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
515 that homogeneous SFmode and DFmode aggregates are passed and returned
516 in general registers.
518 2020-02-21 Jakub Jelinek <jakub@redhat.com>
521 * opts.c (print_filtered_help): Translate help before appending
522 messages to it rather than after that.
524 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
526 PR rtl-optimization/PR92989
527 * lra-lives.c (process_bb_lives): Restore the original order
528 of the bb liveness update. Call make_hard_regno_dead for each
529 register clobbered at the start of an EH receiver.
531 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
534 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
535 self-recursively generated.
537 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
540 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
543 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
545 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
546 Document new target supports option.
548 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
550 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
551 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
552 * config/arm/iterators.md (MATMUL): New iterator.
553 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
554 (mmla_sfx): New attribute.
555 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
556 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
557 (UNSPEC_MATMUL_US): New.
559 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
561 * config/arm/arm.md: Prevent scalar shifts from being used when big
564 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
565 Richard Biener <rguenther@suse.de>
567 PR tree-optimization/93586
568 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
569 after mismatched array refs; do not sure type size information to
570 recover from unmatched referneces with !flag_strict_aliasing_p.
572 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
574 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
575 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
576 (scatter_store<mode>): Rename to ...
577 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
578 (scatter<mode>_exec): Delete. Move contents ...
579 (mask_scatter_store<mode>): ... here, and rename that to ...
580 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
581 Remove mode conversion.
582 (mask_gather_load<mode>): Rename to ...
583 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
584 Remove mode conversion.
585 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
587 2020-02-21 Martin Jambor <mjambor@suse.cz>
589 PR tree-optimization/93845
590 * tree-sra.c (verify_sra_access_forest): Only test access size of
593 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
595 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
596 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
597 (addv64di3_exec): Likewise.
598 (subv64di3): Likewise.
599 (subv64di3_exec): Likewise.
600 (addv64di3_zext): Likewise.
601 (addv64di3_zext_exec): Likewise.
602 (addv64di3_zext_dup): Likewise.
603 (addv64di3_zext_dup_exec): Likewise.
604 (addv64di3_zext_dup2): Likewise.
605 (addv64di3_zext_dup2_exec): Likewise.
606 (addv64di3_sext_dup2): Likewise.
607 (addv64di3_sext_dup2_exec): Likewise.
608 (<expander>v64di3): Likewise.
609 (<expander>v64di3_exec): Likewise.
610 (*<reduc_op>_dpp_shr_v64di): Likewise.
611 (*plus_carry_dpp_shr_v64di): Likewise.
612 * config/gcn/gcn.md (adddi3): Likewise.
613 (addptrdi3): Likewise.
614 (<expander>di3): Likewise.
616 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
618 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
620 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
622 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
623 support. Use aarch64_emit_mult instead of emitting multiplication
624 instructions directly.
625 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
626 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
628 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
630 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
631 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
632 instead of emitting multiplication instructions directly.
633 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
634 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
635 (@aarch64_frecps<mode>): New expanders.
637 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
639 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
640 on and produce uint64_ts rather than ints.
641 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
642 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
644 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
646 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
647 an unused xmsk register when handling approximate rsqrt.
649 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
651 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
652 flag_finite_math_only condition.
654 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
657 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
658 to destination operand for shufps alternative.
659 (*vec_extractv2si_1): Ditto.
661 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
664 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
667 2020-02-20 Martin Liska <mliska@suse.cz>
670 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
672 2020-02-20 Martin Liska <mliska@suse.cz>
675 * common/config/avr/avr-common.c: Remote trailing "|".
677 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
679 * collect2.c (maybe_run_lto_and_relink): Fix typo in
682 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
684 PR tree-optimization/93767
685 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
686 access-size bias from the offset calculations for negative strides.
688 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
690 * collect2.c (c_file, o_file): Make const again.
691 (ldout,lderrout, dump_ld_file): Remove.
692 (tool_cleanup): Avoid calling not signal-safe functions.
693 (maybe_run_lto_and_relink): Avoid possible signal handler
694 access to unintialzed memory (lto_o_files).
695 (main): Avoid leaking temp files in $TMPDIR.
696 Initialize c_file/o_file with concat, which avoids exposing
697 uninitialized memory to signal handler, which calls unlink(!).
698 Avoid calling maybe_unlink when the main function returns,
699 since the atexit handler is already doing this.
700 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
702 2020-02-19 Martin Jambor <mjambor@suse.cz>
704 PR tree-optimization/93776
705 * tree-sra.c (create_access): Do not create zero size accesses.
706 (get_access_for_expr): Do not search for zero sized accesses.
708 2020-02-19 Martin Jambor <mjambor@suse.cz>
710 PR tree-optimization/93667
711 * tree-sra.c (scalarizable_type_p): Return false if record fields
712 do not follow wach other.
714 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
716 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
717 rather than fmv.x.s/fmv.s.x.
719 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
721 * config/aarch64/aarch64-simd-builtins.def
722 (intrinsic_vec_smult_lo_): New.
723 (intrinsic_vec_umult_lo_): Likewise.
724 (vec_widen_smult_hi_): Likewise.
725 (vec_widen_umult_hi_): Likewise.
726 * config/aarch64/aarch64-simd.md
727 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
728 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
729 (vmull_high_s16): Likewise.
730 (vmull_high_s32): Likewise.
731 (vmull_high_u8): Likewise.
732 (vmull_high_u16): Likewise.
733 (vmull_high_u32): Likewise.
734 (vmull_s8): Likewise.
735 (vmull_s16): Likewise.
736 (vmull_s32): Likewise.
737 (vmull_u8): Likewise.
738 (vmull_u16): Likewise.
739 (vmull_u32): Likewise.
741 2020-02-18 Martin Liska <mliska@suse.cz>
743 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
744 bootstrap by missing removal of invalid sanity check.
746 2020-02-18 Martin Liska <mliska@suse.cz>
749 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
750 Always compare LHS of gimple_assign.
752 2020-02-18 Martin Liska <mliska@suse.cz>
755 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
756 and return type of functions.
757 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
758 Drop MALLOC attribute for void functions.
759 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
760 malloc_state for a new VOID clone.
762 2020-02-18 Martin Liska <mliska@suse.cz>
765 * common.opt: Add -fprofile-reproducibility.
766 * doc/invoke.texi: Document it.
767 * value-prof.c (dump_histogram_value):
768 Document and support behavior for counters[0]
769 being a negative value.
770 (get_nth_most_common_value): Handle negative
771 counters[0] in respect to flag_profile_reproducible.
773 2020-02-18 Jakub Jelinek <jakub@redhat.com>
776 * cgraph.c (verify_speculative_call): Use speculative_id instead of
777 speculative_uid in messages. Remove trailing whitespace from error
778 message. Use num_speculative_call_targets instead of
779 num_speculative_targets in a message.
780 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
781 edge messages and stmt instead of cal_stmt in reference message.
783 PR tree-optimization/93780
784 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
785 before calling build_vector_type.
786 (execute_update_addresses_taken): Likewise.
789 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
790 typo, functoin -> function.
791 * tree.c (free_lang_data_in_decl): Fix comment typo,
792 functoin -> function.
793 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
795 2020-02-17 David Malcolm <dmalcolm@redhat.com>
797 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
799 (print_option_information): Don't call get_option_url if URLs
802 2020-02-17 Alexandre Oliva <oliva@adacore.com>
804 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
805 handling of register_common-less targets.
807 2020-02-17 Martin Liska <mliska@suse.cz>
810 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
812 2020-02-17 Martin Liska <mliska@suse.cz>
815 * config/rs6000/rs6000.c (rs6000_option_override_internal):
818 2020-02-17 Martin Liska <mliska@suse.cz>
821 * config/rx/elf.opt: Fix typo.
823 2020-02-17 Richard Biener <rguenther@suse.de>
826 * opts-global.c (print_ignored_options): Use inform and
829 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
832 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
834 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
837 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
838 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
840 2020-02-15 Jason Merrill <jason@redhat.com>
842 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
844 2020-02-15 Jakub Jelinek <jakub@redhat.com>
846 PR tree-optimization/93744
847 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
848 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
849 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
850 sure @2 in the first and @1 in the other patterns has no side-effects.
852 2020-02-15 David Malcolm <dmalcolm@redhat.com>
853 Bernd Edlinger <bernd.edlinger@hotmail.de>
857 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
858 * configure.ac (--with-diagnostics-urls): New configuration
859 option, based on --with-diagnostics-color.
860 (DIAGNOSTICS_URLS_DEFAULT): New define.
861 * config.h: Regenerate.
862 * configure: Regenerate.
863 * diagnostic.c (diagnostic_urls_init): Handle -1 for
864 DIAGNOSTICS_URLS_DEFAULT from configure-time
865 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
866 and TERM_URLS environment variable.
867 * diagnostic-url.h (diagnostic_url_format): New enum type.
868 (diagnostic_urls_enabled_p): rename to...
869 (determine_url_format): ... this, and change return type.
870 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
871 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
872 the linux console, and mingw.
873 (diagnostic_urls_enabled_p): rename to...
874 (determine_url_format): ... this, and adjust.
875 * pretty-print.h (pretty_printer::show_urls): rename to...
876 (pretty_printer::url_format): ... this, and change to enum.
877 * pretty-print.c (pretty_printer::pretty_printer,
878 pp_begin_url, pp_end_url, test_urls): Adjust.
879 * doc/install.texi (--with-diagnostics-urls): Document the new
880 configuration option.
881 (--with-diagnostics-color): Document the existing interaction
882 with GCC_COLORS better.
883 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
884 vindex reference. Update description of defaults based on the above.
885 (-fdiagnostics-color): Update description of how -fdiagnostics-color
886 interacts with GCC_COLORS.
888 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
891 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
892 conjunction with TARGET_GNU_TLS in early return.
894 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
896 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
897 the mode is not wider than UNITS_PER_WORD.
899 2020-02-14 Martin Jambor <mjambor@suse.cz>
901 PR tree-optimization/93516
902 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
903 access of the same type as the parent.
904 (propagate_subaccesses_from_lhs): Likewise.
906 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
909 * config/i386/avx512vbmi2intrin.h
910 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
911 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
912 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
913 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
914 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
915 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
916 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
917 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
918 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
919 of lacking a closing parenthesis.
920 * config/i386/avx512vbmi2vlintrin.h
921 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
922 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
923 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
924 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
925 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
926 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
927 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
928 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
929 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
930 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
931 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
932 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
933 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
934 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
935 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
936 _mm_shldi_epi32, _mm_mask_shldi_epi32,
937 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
938 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
940 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
943 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
944 the target function entry.
946 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
948 * common/config/arc/arc-common.c (arc_option_optimization_table):
949 Disable if-conversion step when optimized for size.
951 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
953 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
954 R12-R15 are always in ARCOMPACT16_REGS register class.
955 * config/arc/arc.opt (mq-class): Deprecate.
956 * config/arc/constraint.md ("q"): Remove dependency on mq-class
958 * doc/invoke.texi (mq-class): Update text.
959 * common/config/arc/arc-common.c (arc_option_optimization_table):
962 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
964 * config/arc/arc.c (arc_insn_cost): New function.
965 (TARGET_INSN_COST): Define.
966 * config/arc/arc.md (cost): New attribute.
967 (add_n): Use arc_nonmemory_operand.
968 (ashlsi3_insn): Likewise, also update constraints.
969 (ashrsi3_insn): Likewise.
971 (add_shift): Likewise.
972 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
974 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
976 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
978 (umulsidi_600): Likewise.
980 2020-02-13 Jakub Jelinek <jakub@redhat.com>
983 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
984 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
985 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
986 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
987 pass __A to the builtin followed by __W instead of __A followed by
989 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
990 _mm512_mask_popcnt_epi64): Likewise.
991 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
992 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
993 _mm256_mask_popcnt_epi64): Likewise.
995 PR tree-optimization/93582
996 * fold-const.h (shift_bytes_in_array_left,
997 shift_bytes_in_array_right): Declare.
998 * fold-const.c (shift_bytes_in_array_left,
999 shift_bytes_in_array_right): New function, moved from
1000 gimple-ssa-store-merging.c, no longer static.
1001 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
1002 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
1003 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
1004 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
1005 shift_bytes_in_array.
1006 (verify_shift_bytes_in_array): Rename to ...
1007 (verify_shift_bytes_in_array_left): ... this. Use
1008 shift_bytes_in_array_left instead of shift_bytes_in_array.
1009 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
1010 instead of verify_shift_bytes_in_array.
1011 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
1012 / native_interpret_expr where the store covers all needed bits,
1013 punt on PDP-endian, otherwise allow all involved offsets and sizes
1014 not to be byte-aligned.
1017 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
1018 use const_0_to_255_operand predicate instead of immediate_operand.
1019 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
1020 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
1021 vgf2p8affineinvqb_<mode><mask_name>,
1022 vgf2p8affineqb_<mode><mask_name>): Drop mode from
1023 const_0_to_255_operand predicated operands.
1025 2020-02-12 Jeff Law <law@redhat.com>
1027 * config/h8300/h8300.md (comparison shortening peepholes): Use
1028 a mode iterator to merge the HImode and SImode peepholes.
1030 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1033 * real.c (is_even): Make static. Function comment fix.
1034 (is_halfway_below): Make static, don't assert R is not inf/nan,
1035 instead return false for those. Small formatting fixes.
1037 2020-02-12 Martin Sebor <msebor@redhat.com>
1040 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1041 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
1042 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1043 (strlen_check_and_optimize_call): Adjust callee name.
1045 2020-02-12 Jeff Law <law@redhat.com>
1047 * config/h8300/h8300.md (comparison shortening peepholes): Drop
1048 (and (xor)) variant. Combine other two into single peephole.
1050 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1052 PR rtl-optimization/93565
1053 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1055 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1057 * config/aarch64/aarch64-simd.md
1058 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1059 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1060 generating separate ADDV and zero_extend patterns.
1061 * config/aarch64/iterators.md (VDQV_E): New iterator.
1063 2020-02-12 Jeff Law <law@redhat.com>
1065 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1066 expanders, splits, etc.
1067 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1068 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1069 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1070 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1071 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1072 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1074 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1076 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1079 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1080 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
1081 TARGET_AVX512DQ from condition.
1082 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1083 instead of <mask_mode512bit_condition> in condition. If
1084 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1086 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1089 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
1092 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1094 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
1096 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1097 where strlen is more legible.
1098 (rs6000_builtin_vectorized_libmass): Ditto.
1099 (rs6000_print_options_internal): Ditto.
1101 2020-02-11 Martin Sebor <msebor@redhat.com>
1103 PR tree-optimization/93683
1104 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1106 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
1108 * config/rs6000/predicates.md (cint34_operand): Rename the
1109 -mprefixed-addr option to be -mprefixed.
1110 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1111 the -mprefixed-addr option to be -mprefixed.
1112 (OTHER_FUTURE_MASKS): Likewise.
1113 (POWERPC_MASKS): Likewise.
1114 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1115 the -mprefixed-addr option to be -mprefixed. Change error
1116 messages to refer to -mprefixed.
1117 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1119 (rs6000_legitimate_offset_address_p): Likewise.
1120 (rs6000_mode_dependent_address): Likewise.
1121 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1122 "-mprefixed" for target attributes and pragmas.
1123 (address_to_insn_form): Rename the -mprefixed-addr option to be
1125 (rs6000_adjust_insn_length): Likewise.
1126 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1127 -mprefixed-addr option to be -mprefixed.
1128 (ASM_OUTPUT_OPCODE): Likewise.
1129 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1130 -mprefixed-addr option to be -mprefixed.
1131 * config/rs6000/rs6000.opt (-mprefixed): Rename the
1132 -mprefixed-addr option to be prefixed. Change the option from
1133 being undocumented to being documented.
1134 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1135 -mprefixed option. Update the -mpcrel documentation to mention
1138 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
1140 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1141 including FIRST_PSEUDO_REGISTER - 1.
1142 * ira-color.c (print_hard_reg_set): Ditto.
1144 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1146 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
1147 (USTERNOP_QUALIFIERS): New define.
1148 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1149 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1150 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1151 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1152 * config/arm/arm_neon.h (vusdot_s32): New.
1153 (vusdot_lane_s32): New.
1154 (vusdotq_lane_s32): New.
1155 (vsudot_lane_s32): New.
1156 (vsudotq_lane_s32): New.
1157 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1158 * config/arm/iterators.md (DOTPROD_I8MM): New.
1159 (sup, opsuffix): Add <us/su>.
1160 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1161 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1163 2020-02-11 Richard Biener <rguenther@suse.de>
1165 PR tree-optimization/93661
1166 PR tree-optimization/93662
1167 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1169 * tree-sra.c (get_access_for_expr): Likewise.
1171 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1174 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1175 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1176 Change condition from TARGET_AVX2 to TARGET_AVX.
1178 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1181 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1182 argument of strncmp.
1184 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1186 Try to generate zero-based comparisons.
1187 * config/cris/cris.c (cris_reduce_compare): New function.
1188 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1189 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1190 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1192 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1195 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1196 in Thumb state and also as a destination in Arm state. Add T16
1199 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1201 * md.texi (Define Subst): Match closing paren in example.
1203 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1207 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1208 arguments of strncmp.
1210 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1213 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1214 but different source value.
1215 (adjust_callers_for_value_intersection): New function.
1216 (gather_edges_for_value): Adjust order of callers to let a
1217 non-self-recursive caller be the first element.
1218 (self_recursive_pass_through_p): Add a new parameter "simple", and
1219 check generalized self-recursive pass-through jump function.
1220 (self_recursive_agg_pass_through_p): Likewise.
1221 (find_more_scalar_values_for_callers_subset): Compute value from
1222 pass-through jump function for self-recursive.
1223 (intersect_with_plats): Cleanup previous implementation code for value
1224 itersection with self-recursive call edge.
1225 (intersect_with_agg_replacements): Likewise.
1226 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1227 function for self-recursive call edge. Cleanup previous implementation
1228 code for value intersection with self-recursive call edge.
1229 (decide_whether_version_node): Remove dead callers and adjust order
1230 to let a non-self-recursive caller be the first element.
1232 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1234 * recog.c: Move pass_split_before_sched2 code in front of
1235 pass_split_before_regstack.
1236 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1237 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1238 (rest_of_handle_split_before_sched2): Remove.
1239 (pass_split_before_sched2::execute): Unconditionally call
1241 (enable_split_before_sched2): New function.
1242 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1243 (pass_split_before_regstack::gate): Ditto.
1244 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1245 Update name check for renamed split4 pass.
1246 * config/sh/sh.c (register_sh_passes): Update pass insertion
1247 point for renamed split4 pass.
1249 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1251 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1252 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1253 copying them around between host and target.
1255 2020-02-08 Andrew Pinski <apinski@marvell.com>
1258 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1259 STRICT_ALIGNMENT also.
1261 2020-02-08 Jim Wilson <jimw@sifive.com>
1264 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1266 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1267 Jakub Jelinek <jakub@redhat.com>
1270 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1271 xmm16-xmm31 call-used even in 64-bit ms-abi.
1273 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1275 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1276 (simd_ummla, simd_usmmla): Likewise.
1277 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1278 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1279 (vusmmlaq_s32): New.
1281 2020-02-07 Richard Biener <rguenther@suse.de>
1284 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1285 skipping unreachable regions.
1286 (optimize_inline_calls): Skip folding stmts when we didn't
1289 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1292 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1293 Don't return aggregates with only SFmode and DFmode in SSE
1295 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1297 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1300 * config/rs6000/rs6000-logue.c
1301 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1302 if it fails, move rs into end_addr and retry. Add
1303 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1304 the insn pattern doesn't describe well what exactly happens to
1308 * config/i386/predicates.md (avx_identity_operand): Remove.
1309 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1310 (avx_<castmode><avxsizesuffix>_<castmode>,
1311 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1312 a VEC_CONCAT of the operand and UNSPEC_CAST.
1313 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1314 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1318 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1319 recog_data.insn if distance_non_agu_define changed it.
1321 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1324 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1325 we only had X-FORM (reg+reg) addressing for vectors. Also before
1326 ISA 3.0, we only had X-FORM addressing for scalars in the
1327 traditional Altivec registers.
1329 2020-02-06 <zhongyunde@huawei.com>
1330 Vladimir Makarov <vmakarov@redhat.com>
1332 PR rtl-optimization/93561
1333 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1334 hard register range.
1336 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1338 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1341 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1343 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1344 where the low and the high 32 bits are equal to each other specially,
1345 with an rldimi instruction.
1347 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1349 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1351 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1353 * config/arm/arm-tables.opt: Regenerate.
1355 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1358 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1359 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1360 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1362 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1364 PR rtl-optimization/87763
1365 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1367 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1369 * config/aarch64/aarch64-simd-builtins.def
1370 (bfmlaq): New built-in function.
1371 (bfmlalb): New built-in function.
1372 (bfmlalt): New built-in function.
1373 (bfmlalb_lane): New built-in function.
1374 (bfmlalt_lane): New built-in function.
1375 * config/aarch64/aarch64-simd.md
1376 (aarch64_bfmmlaqv4sf): New pattern.
1377 (aarch64_bfmlal<bt>v4sf): New pattern.
1378 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1379 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1380 (vbfmlalbq_f32): New intrinsic.
1381 (vbfmlaltq_f32): New intrinsic.
1382 (vbfmlalbq_lane_f32): New intrinsic.
1383 (vbfmlaltq_lane_f32): New intrinsic.
1384 (vbfmlalbq_laneq_f32): New intrinsic.
1385 (vbfmlaltq_laneq_f32): New intrinsic.
1386 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1387 (bt): New int attribute.
1389 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1391 * config/i386/i386.md (*pushtf): Emit "#" instead of
1392 calling gcc_unreachable in insn output.
1395 (*pushsf_rex64): Ditto for alternatives other than 1.
1396 (*pushsf): Ditto for alternatives other than 1.
1398 2020-02-06 Martin Liska <mliska@suse.cz>
1400 PR gcov-profile/91971
1401 PR gcov-profile/93466
1402 * coverage.c (coverage_init): Revert mangling of
1403 path into filename. It can lead to huge filename length.
1404 Creation of subfolders seem more natural.
1406 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1409 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1410 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1411 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1413 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1416 * config/i386/predicates.md (avx_identity_operand): New predicate.
1417 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1418 define_insn_and_split.
1421 * omp-low.c (use_pointer_for_field): For nested constructs, also
1422 look for map clauses on target construct.
1423 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1424 taskreg_nesting_level.
1427 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1428 shared clause, call omp_notice_variable on outer context if any.
1430 2020-02-05 Jason Merrill <jason@redhat.com>
1433 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1434 non-zero address even if weak and not yet defined.
1436 2020-02-05 Martin Sebor <msebor@redhat.com>
1438 PR tree-optimization/92765
1439 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1440 * tree-ssa-strlen.c (compute_string_length): Remove.
1441 (determine_min_objsize): Remove.
1442 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1443 Avoid using type size as the upper bound on string length.
1444 (handle_builtin_string_cmp): Add an argument. Adjust.
1445 (strlen_check_and_optimize_call): Pass additional argument to
1446 handle_builtin_string_cmp.
1448 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1450 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1451 (*pushdi2_rex64 peephole2): Unconditionally split after
1453 (*ashl<mode>3_doubleword): Ditto.
1454 (*<shift_insn><mode>3_doubleword): Ditto.
1456 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1459 * config/rs6000/rs6000.c (get_vector_offset): Fix
1461 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1463 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1465 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1468 (Special Functions for Debugging the Analyzer): Update description
1469 of __analyzer_dump_exploded_nodes.
1471 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1474 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1475 include sets and not clobbers in the vzeroupper pattern.
1476 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1477 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1478 (*avx_vzeroupper_1): New define_insn_and_split.
1481 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1482 don't run when !optimize.
1483 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1486 2020-02-05 Richard Biener <rguenther@suse.de>
1489 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1490 checks before matching calls.
1492 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1494 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1495 function comment typo.
1498 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1499 simd_clone_create failed when i == 0, adjust clone->nargs by
1502 2020-02-05 Martin Liska <mliska@suse.cz>
1505 * doc/invoke.texi: Document that one should
1506 not combine ASLR and -fpch.
1508 2020-02-04 Richard Biener <rguenther@suse.de>
1510 PR tree-optimization/93538
1511 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1513 2020-02-04 Richard Biener <rguenther@suse.de>
1515 PR tree-optimization/91123
1516 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1517 (vn_walk_cb_data::last_vuse): New member.
1518 (vn_walk_cb_data::saved_operands): Likewsie.
1519 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1520 (vn_walk_cb_data::push_partial_def): Use finish.
1521 (vn_reference_lookup_2): Update last_vuse and use finish if
1522 we've saved operands.
1523 (vn_reference_lookup_3): Use finish and update calls to
1524 push_partial_defs everywhere. When translating through
1525 memcpy or aggregate copies save off operands and alias-set.
1526 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1527 operation for redundant store removal.
1529 2020-02-04 Richard Biener <rguenther@suse.de>
1531 PR tree-optimization/92819
1532 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1533 generating more stmts than before.
1535 2020-02-04 Martin Liska <mliska@suse.cz>
1537 * config/arm/arm.c (arm_gen_far_branch): Move the function
1538 outside of selftests.
1540 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1542 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1543 function to adjust PC-relative vector addresses.
1544 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1545 handle vectors with PC-relative addresses.
1547 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1549 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1551 (hard_reg_and_mode_to_addr_mask): Delete.
1552 (rs6000_adjust_vec_address): If the original vector address
1553 was REG+REG or REG+OFFSET and the element is not zero, do the add
1554 of the elements in the original address before adding the offset
1555 for the vector element. Use address_to_insn_form to validate the
1556 address using the register being loaded, rather than guessing
1557 whether the address is a DS-FORM or DQ-FORM address.
1559 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1561 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1562 to calculate the offset in memory from the start of a vector of a
1563 particular element. Add code to keep the element number in
1564 bounds if the element number is variable.
1565 (rs6000_adjust_vec_address): Move calculation of offset of the
1566 vector element to get_vector_offset.
1567 (rs6000_split_vec_extract_var): Do not do the initial AND of
1568 element here, move the code to get_vector_offset.
1570 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1572 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1575 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1577 * config/rs6000/constraints.md: Improve documentation.
1579 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1582 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1583 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1585 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1587 * config.gcc: Remove "carrizo" support.
1588 * config/gcn/gcn-opts.h (processor_type): Likewise.
1589 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1590 * config/gcn/gcn.opt (gpu_type): Likewise.
1591 * config/gcn/t-omp-device: Likewise.
1593 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1596 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1597 * config/arm/arm.c (arm_gen_far_branch): New function
1599 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1601 2020-02-03 Julian Brown <julian@codesourcery.com>
1602 Tobias Burnus <tobias@codesourcery.com>
1604 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1606 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1609 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1610 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1613 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1615 PR rtl-optimization/91333
1616 * ira-color.c (struct allocno_color_data): Add member
1618 (init_allocno_threads): Set the member up.
1619 (bucket_allocno_compare_func): Add compare hard reg
1622 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1624 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1626 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1627 * config.in: Regenerated.
1628 * configure: Regenerated.
1629 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1630 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1631 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1633 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1635 * configure: Regenerate.
1637 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1639 PR rtl-optimization/91333
1640 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1641 reg preferences comparison up.
1643 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1645 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1646 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1647 aarch64-sve-builtins-base.h.
1648 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1649 aarch64-sve-builtins-base.cc.
1650 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1651 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1653 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1654 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1655 (svcvtnt): New functions.
1656 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1657 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1658 (svcvtnt): New functions.
1659 (svcvt): Add a form that converts f32 to bf16.
1660 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1661 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1663 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1664 Treat B as bfloat16_t.
1665 (ternary_bfloat_lane_base): New class.
1666 (ternary_bfloat_def): Likewise.
1667 (ternary_bfloat): New shape.
1668 (ternary_bfloat_lane_def): New class.
1669 (ternary_bfloat_lane): New shape.
1670 (ternary_bfloat_lanex2_def): New class.
1671 (ternary_bfloat_lanex2): New shape.
1672 (ternary_bfloat_opt_n_def): New class.
1673 (ternary_bfloat_opt_n): New shape.
1674 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1675 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1676 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1677 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1678 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1679 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1680 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1681 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1682 the pattern off the narrow mode instead of the wider one.
1683 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1684 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1685 (sve_fp_op): Handle them.
1686 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1687 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1689 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1691 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1692 * config/aarch64/aarch64-modes.def (BF): Move definition before
1693 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1694 (SVE_MODES): Handle BF modes.
1695 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1697 (aarch64_full_sve_mode): Likewise.
1698 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1700 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1701 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1702 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1703 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1705 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1707 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1708 (TYPES_all_data): Add bf16.
1709 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1710 (register_tuple_type): Increase buffer size.
1711 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1712 (bf16): New type suffix.
1713 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1714 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1715 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1716 Change type from all_data to all_arith.
1717 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1720 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1721 Matthew Malcomson <matthew.malcomson@arm.com>
1722 Richard Sandiford <richard.sandiford@arm.com>
1724 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1725 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1726 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1727 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1728 __ARM_FEATURE_MATMUL_FP64.
1729 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1730 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1731 be disabled at the same time.
1732 (f32mm): New extension.
1733 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1734 (AARCH64_FL_F64MM): Bump to the next bit up.
1735 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1736 (TARGET_SVE_F64MM): New macros.
1737 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1738 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1739 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1740 (UNSPEC_ZIP2Q): New unspeccs.
1741 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1742 (optab, sur, perm_insn): Handle the new unspecs.
1743 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1744 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1745 TARGET_SVE_F64MM instead of separate tests.
1746 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1747 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1748 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1749 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1750 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1751 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1752 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1753 (TYPES_s_signed): New macro.
1754 (TYPES_s_integer): Use it.
1755 (TYPES_d_float): New macro.
1756 (TYPES_d_data): Use it.
1757 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1758 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1759 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1760 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1761 (svmmla): New shape.
1762 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1763 template parameters.
1764 (ternary_resize2_lane_base): Likewise.
1765 (ternary_resize2_base): New class.
1766 (ternary_qq_lane_base): Likewise.
1767 (ternary_intq_uintq_lane_def): Likewise.
1768 (ternary_intq_uintq_lane): New shape.
1769 (ternary_intq_uintq_opt_n_def): New class
1770 (ternary_intq_uintq_opt_n): New shape.
1771 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1772 (ternary_uintq_intq_def): New class.
1773 (ternary_uintq_intq): New shape.
1774 (ternary_uintq_intq_lane_def): New class.
1775 (ternary_uintq_intq_lane): New shape.
1776 (ternary_uintq_intq_opt_n_def): New class.
1777 (ternary_uintq_intq_opt_n): New shape.
1778 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1779 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1780 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1781 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1783 (svdotprod_lane_impl): ...this new class.
1784 (svmmla_impl, svusdot_impl): New classes.
1785 (svdot_lane): Update to use svdotprod_lane_impl.
1786 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1787 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1789 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1790 function, with no types defined.
1791 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1792 AARCH64_FL_I8MM functions.
1793 (svmmla): New AARCH64_FL_F32MM function.
1794 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1795 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1796 AARCH64_FL_F64MM function.
1797 (REQUIRED_EXTENSIONS):
1799 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1801 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1804 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1806 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1807 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1808 (*movti_internal): Do not check for
1809 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1810 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1811 just after check for TARGET_AVX.
1812 (*movdf_internal): Ditto.
1813 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1814 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1815 * config/i386/sse.md (mov<mode>_internal): Only check
1816 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1817 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1818 (<sse>_andnot<mode>3<mask_name>): Move check for
1819 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1820 (<code><mode>3<mask_name>): Ditto.
1821 (*andnot<mode>3): Ditto.
1822 (*andnottf3): Ditto.
1823 (*<code><mode>3): Ditto.
1824 (*<code>tf3): Ditto.
1825 (*andnot<VI:mode>3): Remove
1826 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1827 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1828 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1829 (sse4_1_blendv<ssemodesuffix>): Ditto.
1830 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1831 Explain that tune applies to 128bit instructions only.
1833 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1835 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1836 to definition of hsa_kernel_description. Parse assembly to find SGPR
1837 and VGPR count of kernel and store in hsa_kernel_description.
1839 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1841 PR rtl-optimization/91838
1842 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1843 to truncate if allowed or reject combination.
1845 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1847 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1848 (find_inv_vars_cb): Likewise.
1850 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1852 * calls.c (special_function_p): Split out the check for DECL_NAME
1853 being non-NULL and fndecl being extern at file scope into a
1854 new maybe_special_function_p and call it. Drop check for fndecl
1855 being non-NULL that was after a usage of DECL_NAME (fndecl).
1856 * tree.h (maybe_special_function_p): New inline function.
1858 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1860 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1861 (mask_gather_load<mode>): ... here, and zero-initialize the
1863 (maskload<mode>di): Zero-initialize the destination.
1866 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1869 * doc/analyzer.texi (Limitations): Note that constraints on
1870 floating-point values are currently ignored.
1872 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1875 * symtab.c (symtab_node::noninterposable_alias): If localalias
1876 already exists, but is not usable, append numbers after it until
1877 a unique name is found. Formatting fix.
1880 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1883 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1885 * config/gcn/gcn.c (print_operand): Handle LTGT.
1886 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1888 2020-01-30 Richard Biener <rguenther@suse.de>
1890 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1891 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1893 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1895 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1896 without a DECL in .data.rel.ro.local.
1898 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1901 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1905 * config/i386/sse.md
1906 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1907 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1908 any_extend code iterator instead of always zero_extend.
1909 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1910 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1911 Use any_extend code iterator instead of always zero_extend.
1912 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1913 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1914 Use any_extend code iterator instead of always zero_extend.
1915 (*sse2_pmovmskb_ext): New define_insn.
1916 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1919 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1920 (*popcountsi2_zext_falsedep): New define_insn.
1922 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1924 * config.in: Regenerated.
1925 * configure: Regenerated.
1927 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1930 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1931 LLVM's assembler changed the default in version 9.
1933 2020-01-24 Jeff Law <law@redhat.com>
1935 PR tree-optimization/89689
1936 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1938 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1942 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1944 PR rtl-optimization/87763
1945 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1946 simplification to handle subregs as well as bare regs.
1947 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1949 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1952 * ira.c (ira): Revert use of simplified LRA algorithm.
1954 2020-01-29 Martin Jambor <mjambor@suse.cz>
1956 PR tree-optimization/92706
1957 * tree-sra.c (struct access): Fields first_link, last_link,
1958 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1959 next_rhs_queued and grp_rhs_queued respectively, new fields
1960 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1961 (struct assign_link): Field next renamed to next_rhs, new field
1962 next_lhs. Updated comment.
1963 (work_queue_head): Renamed to rhs_work_queue_head.
1964 (lhs_work_queue_head): New variable.
1965 (add_link_to_lhs): New function.
1966 (relink_to_new_repr): Also relink LHS lists.
1967 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1968 (add_access_to_lhs_work_queue): New function.
1969 (pop_access_from_work_queue): Renamed to
1970 pop_access_from_rhs_work_queue.
1971 (pop_access_from_lhs_work_queue): New function.
1972 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1974 (child_would_conflict_in_lacc): Renamed to
1975 child_would_conflict_in_acc. Adjusted parameter names.
1976 (create_artificial_child_access): New parameter set_grp_read, use it.
1977 (subtree_mark_written_and_enqueue): Renamed to
1978 subtree_mark_written_and_rhs_enqueue.
1979 (propagate_subaccesses_across_link): Renamed to
1980 propagate_subaccesses_from_rhs.
1981 (propagate_subaccesses_from_lhs): New function.
1982 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1985 2020-01-29 Martin Jambor <mjambor@suse.cz>
1987 PR tree-optimization/92706
1988 * tree-sra.c (struct access): Adjust comment of
1989 grp_total_scalarization.
1990 (find_access_in_subtree): Look for single children spanning an entire
1992 (scalarizable_type_p): Allow register accesses, adjust callers.
1993 (completely_scalarize): Remove function.
1994 (scalarize_elem): Likewise.
1995 (create_total_scalarization_access): Likewise.
1996 (sort_and_splice_var_accesses): Do not track total scalarization
1998 (analyze_access_subtree): New parameter totally, adjust to new meaning
1999 of grp_total_scalarization.
2000 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
2001 (can_totally_scalarize_forest_p): New function.
2002 (create_total_scalarization_access): Likewise.
2003 (create_total_access_and_reshape): Likewise.
2004 (total_should_skip_creating_access): Likewise.
2005 (totally_scalarize_subtree): Likewise.
2006 (analyze_all_variable_accesses): Perform total scalarization after
2007 subaccess propagation using the new functions above.
2008 (initialize_constant_pool_replacements): Output initializers by
2009 traversing the access tree.
2011 2020-01-29 Martin Jambor <mjambor@suse.cz>
2013 * tree-sra.c (verify_sra_access_forest): New function.
2014 (verify_all_sra_access_forests): Likewise.
2015 (create_artificial_child_access): Set parent.
2016 (analyze_all_variable_accesses): Call the verifier.
2018 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2020 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
2021 if called on indirect edge.
2022 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
2023 speculative call if needed.
2025 2020-01-29 Richard Biener <rguenther@suse.de>
2027 PR tree-optimization/93428
2028 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
2029 permutation when the load node is created.
2030 (vect_analyze_slp_instance): Re-use it here.
2032 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2034 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2036 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
2038 PR rtl-optimization/93272
2039 * ira-lives.c (process_out_of_region_eh_regs): New function.
2040 (process_bb_node_lives): Call it.
2042 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2044 * coverage.c (read_counts_file): Make error message lowercase.
2046 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2048 * profile-count.c (profile_quality_display_names): Fix ordering.
2050 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2053 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2054 hash only when edge is first within the sequence.
2055 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2056 (symbol_table::create_edge): Do not set target_prob.
2057 (cgraph_edge::remove_caller): Watch for speculative calls when updating
2059 (cgraph_edge::make_speculative): Drop target_prob parameter.
2060 (cgraph_edge::speculative_call_info): Remove.
2061 (cgraph_edge::first_speculative_call_target): New member function.
2062 (update_call_stmt_hash_for_removing_direct_edge): New function.
2063 (cgraph_edge::resolve_speculation): Rewrite to new API.
2064 (cgraph_edge::speculative_call_for_target): New member function.
2065 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2066 multiple speculation targets.
2067 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2069 (verify_speculative_call): Verify that targets form an interval.
2070 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2071 (cgraph_edge::first_speculative_call_target): New member function.
2072 (cgraph_edge::next_speculative_call_target): New member function.
2073 (cgraph_edge::speculative_call_target_ref): New member function.
2074 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2075 (cgraph_edge): Remove target_prob.
2076 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2077 Fix handling of speculative calls.
2078 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2079 * ipa-fnsummary.c (analyze_function_body): Likewise.
2080 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2081 * ipa-profile.c (dump_histogram): Fix formating.
2082 (ipa_profile_generate_summary): Watch for overflows.
2083 (ipa_profile): Do not require probablity to be 1/2; update to new API.
2084 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2085 (update_indirect_edges_after_inlining): Update to new API.
2086 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2088 * profile-count.h: (profile_probability::adjusted): New.
2089 * tree-inline.c (copy_bb): Update to new speculative call API; fix
2090 updating of profile.
2091 * value-prof.c (gimple_ic_transform): Rename to ...
2092 (dump_ic_profile): ... this one; update dumping.
2093 (stream_in_histogram_value): Fix formating.
2094 (gimple_value_profile_transformations): Update.
2096 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2099 * config/i386/i386.md (*movoi_internal_avx): Remove
2100 TARGET_SSE_TYPELESS_STORES check.
2101 (*movti_internal): Prefer TARGET_AVX over
2102 TARGET_SSE_TYPELESS_STORES.
2103 (*movtf_internal): Likewise.
2104 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2105 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
2106 from TARGET_SSE_TYPELESS_STORES.
2108 2020-01-28 David Malcolm <dmalcolm@redhat.com>
2110 * diagnostic-core.h (warning_at): Rename overload to...
2111 (warning_meta): ...this.
2112 (emit_diagnostic_valist): Delete decl of overload taking
2113 diagnostic_metadata.
2114 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2115 (warning_at): Rename overload taking diagnostic_metadata to...
2116 (warning_meta): ...this.
2118 2020-01-28 Richard Biener <rguenther@suse.de>
2120 PR tree-optimization/93439
2121 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2122 * tree-cfg.c (move_sese_region_to_fn): ... here.
2123 (verify_types_in_gimple_reference): Verify used cliques are
2126 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2129 * config/i386/i386-options.c (set_ix86_tune_features): Add an
2130 argument of a pointer to struct gcc_options and pass it to
2131 parse_mtune_ctrl_str.
2132 (ix86_function_specific_restore): Pass opts to
2133 set_ix86_tune_features.
2134 (ix86_option_override_internal): Likewise.
2135 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2136 gcc_options and use it for x_ix86_tune_ctrl_string.
2138 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2140 PR rtl-optimization/87763
2141 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2142 simplification to handle subregs as well as bare regs.
2143 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2145 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2147 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2148 for reduction chains that (now) include a call.
2150 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2152 PR tree-optimization/92822
2153 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2154 out the don't-care elements of a vector whose significant elements
2155 are duplicates, make the don't-care elements duplicates too.
2157 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2159 PR tree-optimization/93434
2160 * tree-predcom.c (split_data_refs_to_components): Record which
2161 components have had aliasing loads removed. Prevent store-store
2162 commoning for all such components.
2164 2020-01-28 Jakub Jelinek <jakub@redhat.com>
2167 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2168 -1 or is_vshift is true, use new_vector with number of elts npatterns
2169 rather than new_unary_operation.
2171 PR tree-optimization/93454
2172 * gimple-fold.c (fold_array_ctor_reference): Perform
2173 elt_size.to_uhwi () just once, instead of calling it in every
2174 iteration. Punt if that value is above size of the temporary
2175 buffer. Decrease third native_encode_expr argument when
2176 bufoff + elt_sz is above size of buf.
2178 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2180 * config/mips/mips.c (mips_declare_object_name)
2181 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2183 2020-01-27 Martin Liska <mliska@suse.cz>
2185 PR gcov-profile/93403
2186 * tree-profile.c (gimple_init_gcov_profiler): Generate
2187 both __gcov_indirect_call_profiler_v4 and
2188 __gcov_indirect_call_profiler_v4_atomic.
2190 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2193 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2195 (@aarch64_split_simd_mov<mode>): Use it.
2196 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2197 Leave the vec_extract patterns to handle 2-element vectors.
2198 (aarch64_simd_mov_from_<mode>high): Likewise.
2199 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2200 (vec_extractv2dfv1df): Likewise.
2202 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2204 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2205 jump conditions for *compare_condjump<GPI:mode>.
2207 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2210 * digraph.cc (test_edge::test_edge): Specify template for base
2213 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2215 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2217 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2219 * config/arc/arc-protos.h (gen_mlo): Remove.
2220 (gen_mhi): Likewise.
2221 * config/arc/arc.c (AUX_MULHI): Define.
2222 (arc_must_save_reister): Special handling for r58/59.
2223 (arc_compute_frame_size): Consider mlo/mhi registers.
2224 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2226 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2227 mlo/mhi name selection.
2228 (arc_restore_callee_saves): Don't early restore blink when ISR.
2229 (arc_expand_prologue): Add mlo/mhi saving.
2230 (arc_expand_epilogue): Add mlo/mhi restoring.
2233 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2234 numbering when MUL64 option is used.
2235 (DWARF2_FRAME_REG_OUT): Define.
2236 * config/arc/arc.md (arc600_stall): New pattern.
2237 (VUNSPEC_ARC_ARC600_STALL): Define.
2238 (mulsi64): Use correct mlo/mhi registers.
2239 (mulsi_600): Clean it up.
2240 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2242 (mhi_operand): Likewise.
2244 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2245 Petro Karashchenko <petro.karashchenko@ring.com>
2247 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2248 attributes if needed.
2249 (prepare_move_operands): Generate special unspec instruction for
2251 (arc_isuncached_mem_p): Propagate uncached attribute to each
2253 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2254 (VUNSPEC_ARC_STDI): Likewise.
2255 (ALLI): New mode iterator.
2256 (mALLI): New mode attribute.
2257 (lddi): New instruction pattern.
2259 (stdidi_split): Split instruction for architectures which are not
2260 supporting ll64 option.
2261 (lddidi_split): Likewise.
2263 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2265 PR rtl-optimization/92989
2266 * lra-lives.c (process_bb_lives): Update the live-in set before
2267 processing additional clobbers.
2269 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2271 PR rtl-optimization/93170
2272 * cselib.c (cselib_invalidate_regno_val): New function, split out
2274 (cselib_invalidate_regno): ...here.
2275 (cselib_invalidated_by_call_p): New function.
2276 (cselib_process_insn): Iterate over all the hard-register entries in
2277 REG_VALUES and invalidate any that cross call-clobbered registers.
2279 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2281 * dojump.c (split_comparison): Use HONOR_NANS rather than
2282 HONOR_SNANS when splitting LTGT.
2284 2020-01-27 Martin Liska <mliska@suse.cz>
2287 * opts.c (print_filtered_help): Exclude language-specific
2288 options from --help=common unless enabled in all FEs.
2290 2020-01-27 Martin Liska <mliska@suse.cz>
2292 * opts.c (print_help): Exclude params from
2293 all except --help=param.
2295 2020-01-27 Martin Liska <mliska@suse.cz>
2298 * config/i386/i386-features.c (make_resolver_func):
2299 Align the code with ppc64 target implementation.
2300 Do not generate a unique name for resolver function.
2302 2020-01-27 Richard Biener <rguenther@suse.de>
2304 PR tree-optimization/93397
2305 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2306 converted reduction chain SLP graph adjustment.
2308 2020-01-26 Marek Polacek <polacek@redhat.com>
2311 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2314 2020-01-26 Jason Merrill <jason@redhat.com>
2317 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2320 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2322 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2323 (rx_setmem): Likewise.
2325 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2328 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2329 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2330 drop <di> from constraint of last operand.
2333 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2334 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2335 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2337 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2340 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2343 2020-01-24 Jeff Law <law@redhat.com>
2345 PR tree-optimization/92788
2346 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2349 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2352 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2353 *avx_vperm_broadcast_<mode>,
2354 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2355 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2356 Move before avx2_perm<mode>/avx512f_perm<mode>.
2359 * simplify-rtx.c (simplify_const_unary_operation,
2360 simplify_const_binary_operation): Punt for mode precision above
2361 MAX_BITSIZE_MODE_ANY_INT.
2363 2020-01-24 Andrew Pinski <apinski@marvell.com>
2365 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2368 2020-01-24 Jeff Law <law@redhat.com>
2371 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2372 for REGs. Call output_operand_lossage to get more reasonable
2375 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2377 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2378 gcn_fp_compare_operator.
2379 (vec_cmpu<mode>di): Use gcn_compare_operator.
2380 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2381 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2382 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2383 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2384 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2385 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2386 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2387 gcn_fp_compare_operator.
2388 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2389 gcn_fp_compare_operator.
2390 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2391 gcn_fp_compare_operator.
2392 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2393 gcn_fp_compare_operator.
2395 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2397 * doc/install.texi (Cross-Compiler-Specific Options): Document
2398 `--with-toolexeclibdir' option.
2400 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2402 * target.def (flags_regnum): Also mention effect on delay slot filling.
2403 * doc/tm.texi: Regenerate.
2405 2020-01-23 Jeff Law <law@redhat.com>
2407 PR translation/90162
2408 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2410 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2413 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2416 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2418 PR rtl-optimization/93402
2419 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2422 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2424 * config.in: Regenerated.
2425 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2426 for TARGET_LIBC_GNUSTACK.
2427 * configure: Regenerated.
2428 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2429 found to be 2.31 or greater.
2431 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2433 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2435 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2436 (mips_asm_file_end): New function. Delegate to
2437 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2438 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2440 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2443 * config/i386/i386-modes.def (POImode): New mode.
2444 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2445 * config/i386/i386.md (DPWI): New mode attribute.
2446 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2448 (QPWI): ... this. Use POI instead of OI for TImode.
2449 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2450 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2453 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2456 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2458 (speculation_tracker_rev): New pattern.
2459 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2460 Use speculation_tracker_rev to track the inverse condition.
2462 2020-01-23 Richard Biener <rguenther@suse.de>
2464 PR tree-optimization/93381
2465 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2466 alias-set of the def as argument and record the first one.
2467 (vn_walk_cb_data::first_set): New member.
2468 (vn_reference_lookup_3): Pass the alias-set of the current def
2469 to push_partial_def. Fix alias-set used in the aggregate copy
2471 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2472 * real.c (clear_significand_below): Fix out-of-bound access.
2474 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2477 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2478 New define_insn patterns.
2480 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2482 * doc/sourcebuild.texi (check-function-bodies): Add an
2483 optional target/xfail selector.
2485 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2487 PR rtl-optimization/93124
2488 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2489 bare USE and CLOBBER insns.
2491 2020-01-22 Andrew Pinski <apinski@marvell.com>
2493 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2495 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2498 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2499 diagnostic_manager into "ana" namespace.
2500 * selftest-run-tests.c (selftest::run_tests): Update for move of
2501 selftest::run_analyzer_selftests to
2502 ana::selftest::run_analyzer_selftests.
2504 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2506 * cfgexpand.c (union_stack_vars): Update the size.
2508 2020-01-22 Richard Biener <rguenther@suse.de>
2510 PR tree-optimization/93381
2511 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2512 throughout, handle all conversions the same.
2514 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2517 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2518 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2519 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2520 Call force_reg on high_in2 unconditionally.
2522 2020-01-22 Martin Liska <mliska@suse.cz>
2524 PR tree-optimization/92924
2525 * profile.c (compute_value_histograms): Divide
2528 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2531 * output.h (assemble_name_resolve): Declare.
2532 * varasm.c (assemble_name_resolve): New function.
2533 (assemble_name): Use it.
2534 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2536 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2538 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2539 update_web_docs_git instead of update_web_docs_svn.
2541 2020-01-21 Andrew Pinski <apinski@marvell.com>
2544 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2545 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2546 (*tlsgd_small_<mode>): Likewise.
2547 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2548 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2549 register. Convert that register back to dest using convert_mode.
2551 2020-01-21 Jim Wilson <jimw@sifive.com>
2553 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2556 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2557 Uros Bizjak <ubizjak@gmail.com>
2560 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2562 (legitimize_tls_address): Do GNU2 TLS address computation in
2563 ptr_mode and zero-extend result to Pmode.
2564 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2565 :P with :PTR and Pmode with ptr_mode.
2566 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2567 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2568 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2570 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2573 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2574 the last two operands are CONST_INT_P before using them as such.
2576 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2578 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2579 to get the integer element types.
2581 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2583 * config/aarch64/aarch64-sve-builtins.h
2584 (function_expander::convert_to_pmode): Declare.
2585 * config/aarch64/aarch64-sve-builtins.cc
2586 (function_expander::convert_to_pmode): New function.
2587 (function_expander::get_contiguous_base): Use it.
2588 (function_expander::prepare_gather_address_operands): Likewise.
2589 * config/aarch64/aarch64-sve-builtins-sve2.cc
2590 (svwhilerw_svwhilewr_impl::expand): Likewise.
2592 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2595 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2596 cfun->machine->label_is_assembled.
2597 (aarch64_print_patchable_function_entry): New.
2598 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2599 * config/aarch64/aarch64.h (struct machine_function): New field,
2602 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2605 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2608 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2611 * cgraph.c (cgraph_edge::resolve_speculation,
2612 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2613 call_stmt_site_hash.
2615 2020-01-21 Martin Liska <mliska@suse.cz>
2617 * config/rs6000/rs6000.c (common_mode_defined): Remove
2620 2020-01-21 Richard Biener <rguenther@suse.de>
2622 PR tree-optimization/92328
2623 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2624 type when value-numbering same-sized store by inserting a
2626 (eliminate_dom_walker::eliminate_stmt): When eliminating
2627 a redundant store handle bit-reinterpretation of the same value.
2629 2020-01-21 Andrew Pinski <apinski@marvel.com>
2632 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2634 (prepare_block_for_update): This. Use a worklist instead of
2637 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2639 * gcc/config/arm/arm.c (clear_operation_p):
2640 Initialise last_regno, skip first iteration
2641 based on the first_set value and use ints instead
2642 of the unnecessary HOST_WIDE_INTs.
2644 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2647 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2648 compare_mode other than SFmode or DFmode.
2650 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2653 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2654 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2655 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2657 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2659 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2661 2020-01-20 Andrew Pinski <apinski@marvell.com>
2664 * targhooks.c (default_print_patchable_function_entry): Use
2665 output_asm_insn to emit the nop instruction.
2667 2020-01-20 Fangrui Song <maskray@google.com>
2670 * targhooks.c (default_print_patchable_function_entry): Align to
2673 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2676 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2677 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2678 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2679 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2680 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2681 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2682 Remove the {q} suffix from lea.
2683 (*tls_dynamic_gnu2_call_64): Renamed to ...
2684 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2685 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2686 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2687 Pass Pmode to gen_tls_dynamic_gnu2_64.
2689 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2691 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2693 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2695 * config/aarch64/aarch64-sve-builtins-base.cc
2696 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2698 2020-01-20 Richard Biener <rguenther@suse.de>
2701 * dwarf2out.c (prune_unused_types): Unconditionally mark
2702 called function DIEs.
2704 2020-01-20 Martin Liska <mliska@suse.cz>
2706 PR tree-optimization/93199
2707 * tree-eh.c (struct leh_state): Add
2708 new field outer_non_cleanup.
2709 (cleanup_is_dead_in): Pass leh_state instead
2710 of eh_region. Add a checking that state->outer_non_cleanup
2711 points to outer non-clean up region.
2712 (lower_try_finally): Record outer_non_cleanup
2714 (lower_catch): Likewise.
2715 (lower_eh_filter): Likewise.
2716 (lower_eh_must_not_throw): Likewise.
2717 (lower_cleanup): Likewise.
2719 2020-01-20 Richard Biener <rguenther@suse.de>
2721 PR tree-optimization/93094
2722 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2723 (vect_transform_loop): Likewise.
2724 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2725 loop_vectorized_call to vect_transform_loop.
2726 * tree-vect-loop.c (vect_transform_loop): Pass down
2727 loop_vectorized_call to vect_loop_versioning.
2728 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2729 the earlier discovered loop_vectorized_call.
2731 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2733 * doc/contribute.texi: Update for SVN -> Git transition.
2734 * doc/install.texi: Likewise.
2736 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2739 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2740 speculative targets.
2741 (verify_speculative_call): New function
2742 (cgraph_node::verify_node): Use it.
2743 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2746 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2749 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2750 (cgraph_edge::make_direct): Remove all indirect targets.
2751 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2752 (cgraph_node::verify_node): Verify that only one call_stmt or
2753 lto_stmt_uid is set.
2754 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2756 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2757 (lto_output_ref): Simplify streaming of stmt.
2758 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2760 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2762 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2763 Mark parameter unused.
2765 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2767 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2769 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2771 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2773 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2775 * Makefile.in: Add coroutine-passes.o.
2776 * builtin-types.def (BT_CONST_SIZE): New.
2777 (BT_FN_BOOL_PTR): New.
2778 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2779 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2780 * coroutine-builtins.def: New file.
2781 * coroutine-passes.cc: New file.
2782 * function.h (struct GTY function): Add a bit to indicate that the
2783 function is a coroutine component.
2784 * internal-fn.c (expand_CO_FRAME): New.
2785 (expand_CO_YIELD): New.
2786 (expand_CO_SUSPN): New.
2787 (expand_CO_ACTOR): New.
2788 * internal-fn.def (CO_ACTOR): New.
2792 * passes.def: Add pass_coroutine_lower_builtins,
2793 pass_coroutine_early_expand_ifns.
2794 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2795 (make_pass_coroutine_early_expand_ifns): New.
2796 * doc/invoke.texi: Document the fcoroutines command line
2799 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2801 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2804 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2805 after checking the argument is a REG. Don't use REGNO (reg)
2806 again to set last_regno, reuse regno variable instead.
2808 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2810 * doc/analyzer.texi (Limitations): Add note about NaN.
2812 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2813 Sudakshina Das <sudi.das@arm.com>
2815 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2816 and valid immediate.
2817 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2818 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2819 * config/arm/constraints.md (Pg): New.
2820 * config/arm/predicates.md (long_shift_imm): New.
2821 (arm_reg_or_long_shift_imm): Likewise.
2822 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2823 (thumb2_lsll): Likewise.
2826 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2827 Sudakshina Das <sudi.das@arm.com>
2829 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2830 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2831 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2832 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2833 * config/arm/thumb2.md (thumb2_asrl): New.
2834 (thumb2_lsll): Likewise.
2836 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2838 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2841 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2843 * gdbinit.in (help-gcc-hooks): New command.
2844 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2845 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2848 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2850 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2851 correct target macro.
2853 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2855 * config/aarch64/aarch64-protos.h
2856 (aarch64_sve_ld1ro_operand_p): New.
2857 * config/aarch64/aarch64-sve-builtins-base.cc
2858 (class load_replicate): New.
2859 (class svld1ro_impl): New.
2860 (class svld1rq_impl): Change to inherit from load_replicate.
2861 (svld1ro): New sve intrinsic function base.
2862 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2863 New DEF_SVE_FUNCTION.
2864 * config/aarch64/aarch64-sve-builtins-base.h
2865 (svld1ro): New decl.
2866 * config/aarch64/aarch64-sve-builtins.cc
2867 (function_expander::add_mem_operand): Modify assert to allow
2869 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2871 * config/aarch64/aarch64.c
2872 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2873 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2874 (aarch64_sve_ld1ro_operand_p): New.
2875 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2876 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2877 * config/aarch64/predicates.md
2878 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2880 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2882 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2883 Introduce this ACLE specified predefined macro.
2884 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2885 (fp): Disabling this disables f64mm.
2886 (simd): Disabling this disables f64mm.
2887 (fp16): Disabling this disables f64mm.
2888 (sve): Disabling this disables f64mm.
2889 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2890 (AARCH64_ISA_F64MM): New.
2891 (TARGET_F64MM): New.
2892 * doc/invoke.texi (f64mm): Document new option.
2894 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2896 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2897 (neoversen1_tunings): Likewise.
2899 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2902 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2903 Add assert to ensure prolog has been emitted.
2904 (aarch64_split_atomic_op): Likewise.
2905 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2906 Use epilogue_completed rather than reload_completed.
2907 (aarch64_atomic_exchange<mode>): Likewise.
2908 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2909 (atomic_nand<mode>): Likewise.
2910 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2911 (atomic_fetch_nand<mode>): Likewise.
2912 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2913 (atomic_nand_fetch<mode>): Likewise.
2915 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2918 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2920 (REVERSE_CONDITION): Delete.
2921 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2922 (CCFP_CCFPE): Likewise.
2923 (e): New mode attribute.
2924 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2925 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2926 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2927 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2928 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2929 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2930 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2931 name of generator from gen_ccmpdi to gen_ccmpccdi.
2932 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2933 the previous comparison but aren't able to, use the new ccmp_rev
2936 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2938 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2939 than testing directly for INTEGER_CST.
2940 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2942 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2944 PR tree-optimization/93292
2945 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2946 get_vectype_for_scalar_type returns NULL.
2948 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2950 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2951 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2953 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2955 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2957 * params.opt: (max-predicted-iterations): Set bounds.
2958 * predict.c (real_almost_one, real_br_prob_base,
2959 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2960 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2961 probabilities; do not truncate to reg_br_prob_bases.
2962 (estimate_loops_at_level): Pass max_cyclic_prob.
2963 (estimate_loops): Compute max_cyclic_prob.
2964 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2966 * profile-count.c (profile_probability::to_sreal): New.
2967 * profile-count.h (class sreal): Move up in file.
2968 (profile_probability::to_sreal): Declare.
2970 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2973 (arm_invalid_conversion): New function for target hook.
2974 (arm_invalid_unary_op): New function for target hook.
2975 (arm_invalid_binary_op): New function for target hook.
2977 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2979 * config.gcc: Add arm_bf16.h.
2980 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2981 (arm_simd_builtin_std_type): Add BFmode.
2982 (arm_init_simd_builtin_types): Define element types for vector types.
2983 (arm_init_bf16_types): New function.
2984 (arm_init_builtins): Add arm_init_bf16_types function call.
2985 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2986 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2987 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2988 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2989 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2990 (arm_mangle_type): Add __bf16.
2991 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2992 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2993 arm_bf16_ptr_type_node.
2994 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2995 define_split between ARM registers.
2996 * config/arm/arm_bf16.h: New file.
2997 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2998 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
3000 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
3001 * config/arm/vfp.md: Add BFmode to movhf patterns.
3003 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
3004 Andre Vieira <andre.simoesdiasvieira@arm.com>
3006 * config/arm/arm-cpus.in (mve, mve_float): New features.
3007 (dsp, mve, mve.fp): New options.
3008 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
3009 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
3010 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
3012 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3013 Thomas Preud'homme <thomas.preudhomme@arm.com>
3015 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
3017 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
3018 error for using -mcmse when targeting Armv8.1-M Mainline.
3020 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3021 Thomas Preud'homme <thomas.preudhomme@arm.com>
3023 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
3024 address in r4 when targeting Armv8.1-M Mainline.
3025 (nonsecure_call_value_internal): Likewise.
3026 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
3027 a register match_operand again. Emit BLXNS when targeting
3029 (nonsecure_call_value_reg_thumb2): Likewise.
3031 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3032 Thomas Preud'homme <thomas.preudhomme@arm.com>
3034 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
3035 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3036 variable as true when floating-point ABI is not hard. Replace
3037 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3038 Generate VLSTM and VLLDM instruction respectively before and
3039 after a function call to cmse_nonsecure_call function.
3040 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3041 (VUNSPEC_VLLDM): Likewise.
3042 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3043 (lazy_load_multiple_insn): Likewise.
3045 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3046 Thomas Preud'homme <thomas.preudhomme@arm.com>
3048 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3049 (arm_emit_vfp_multi_reg_pop): Likewise.
3050 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3051 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
3052 restore callee-saved VFP registers.
3054 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3055 Thomas Preud'homme <thomas.preudhomme@arm.com>
3057 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3058 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3059 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
3060 callee-saved GPRs as well as clear ip register before doing a nonsecure
3061 call then restore callee-saved GPRs after it when targeting
3063 (arm_reorg): Adapt to function rename.
3065 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3066 Thomas Preud'homme <thomas.preudhomme@arm.com>
3068 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3069 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3070 clear_vfp_multiple pattern based on a new vfp parameter.
3071 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3072 targeting Armv8.1-M Mainline.
3073 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3074 unconditionally when targeting Armv8.1-M Mainline architecture. Check
3075 whether VFP registers are available before looking call_used_regs for a
3077 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3078 of prototype of clear_operation_p.
3079 (clear_vfp_multiple_operation): New predicate.
3080 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3081 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3083 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3084 Thomas Preud'homme <thomas.preudhomme@arm.com>
3086 * config/arm/arm-protos.h (clear_operation_p): Declare.
3087 * config/arm/arm.c (clear_operation_p): New function.
3088 (cmse_clear_registers): Generate clear_multiple instruction pattern if
3089 targeting Armv8.1-M Mainline or successor.
3090 (output_return_instruction): Only output APSR register clearing if
3091 Armv8.1-M Mainline instructions not available.
3092 (thumb_exit): Likewise.
3093 * config/arm/predicates.md (clear_multiple_operation): New predicate.
3094 * config/arm/thumb2.md (clear_apsr): New define_insn.
3095 (clear_multiple): Likewise.
3096 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3098 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3099 Thomas Preud'homme <thomas.preudhomme@arm.com>
3101 * config/arm/arm.c (fp_sysreg_names): Declare and define.
3102 (use_return_insn): Also return false for Armv8.1-M Mainline.
3103 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3104 Mainline instructions are available.
3105 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3106 when targeting Armv8.1-M Mainline Security Extensions.
3107 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3108 Mainline entry function.
3109 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3110 targeting Armv8.1-M Mainline or successor.
3111 (arm_expand_epilogue): Fix indentation of caller-saved register
3112 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
3114 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3115 (FP_SYSREGS): Likewise.
3116 (enum vfp_sysregs_encoding): Define enum.
3117 (fp_sysreg_names): Declare.
3118 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3119 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3120 (pop_fpsysreg_insn): Likewise.
3122 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3123 Thomas Preud'homme <thomas.preudhomme@arm.com>
3125 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3126 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3127 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3128 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3129 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3130 (ARMv8_1m_main): New feature group.
3131 (armv8.1-m.main): New architecture.
3132 * config/arm/arm-tables.opt: Regenerate.
3133 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3134 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3135 (arm_options_perform_arch_sanity_checks): Error out when targeting
3136 Armv8.1-M Mainline Security Extensions.
3137 * config/arm/arm.h (arm_arch8_1m_main): Declare.
3139 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3141 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3142 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3143 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3144 aarch64_bfdot_laneq): New.
3145 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3146 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3147 vbfdotq_laneq_f32): New.
3148 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3149 VBFMLA_W, VBF): New.
3150 (isquadop): Add V4BF, V8BF.
3152 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3154 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3155 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3156 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3157 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3158 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3159 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3160 usdot_laneq, sudot_lane,sudot_laneq): New.
3161 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3162 (aarch64_<sur>dot_lane): New.
3163 * config/aarch64/arm_neon.h (vusdot_s32): New.
3165 (vusdot_lane_s32): New.
3166 (vsudot_lane_s32): New.
3167 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3168 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3170 2020-01-16 Martin Liska <mliska@suse.cz>
3172 * value-prof.c (dump_histogram_value): Fix
3173 obvious spacing issue.
3175 2020-01-16 Andrew Pinski <apinski@marvell.com>
3177 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3178 !storage_order_barrier_p.
3180 2020-01-16 Andrew Pinski <apinski@marvell.com>
3182 * sched-int.h (_dep): Add unused bit-field field for the padding.
3183 * sched-deps.c (init_dep_1): Init unused field.
3185 2020-01-16 Andrew Pinski <apinski@marvell.com>
3187 * optabs.h (create_expand_operand): Initialize target field also.
3189 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3191 PR tree-optimization/92429
3192 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3193 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3195 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3198 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3200 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3201 aarch64_sve_int_mode to each mode.
3203 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3205 * doc/analyzer.texi (Overview): Add note about
3206 -fdump-ipa-analyzer.
3208 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3210 PR tree-optimization/93231
3211 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3212 input_type is unsigned. Use tree_to_shwi for shift constant.
3213 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3214 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3215 inputs more efficiently.
3217 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3219 * config/i386/i386.md (*movsf_internal): Do not require
3220 SSE2 ISA for alternatives 14 and 15.
3222 2020-01-15 Richard Biener <rguenther@suse.de>
3225 * tree-eh.c (sink_clobbers): If we already visited the destination
3226 block do not defer insertion.
3227 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3228 the purpose of defered insertion.
3230 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3232 * BASE-VER: Bump to 10.0.1.
3234 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3236 PR tree-optimization/93247
3237 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3238 type of the stmt that we're going to vectorize.
3240 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3242 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3243 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3246 2020-01-15 Martin Liska <mliska@suse.cz>
3248 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3249 2 calls of streamer_read_hwi in a function call.
3251 2020-01-15 Richard Biener <rguenther@suse.de>
3253 * alias.c (record_alias_subset): Avoid redundant work when
3254 subset is already recorded.
3256 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3258 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3259 the analyzer options provide CWE identifiers.
3261 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3263 * tree-diagnostic-path.cc (path_summary::event_range::print):
3264 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3265 using get_pure_location.
3267 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3269 PR tree-optimization/93262
3270 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3271 perform head trimming only if the last argument is constant,
3272 either all ones, or larger or equal to head trim, in the latter
3273 case decrease the last argument by head_trim.
3275 PR tree-optimization/93249
3276 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3277 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3278 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3279 perform head trim unless we can prove there are no '\0' chars
3280 from the source among the first head_trim chars.
3282 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3284 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3286 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3289 * config/i386/sse.md
3290 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3291 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3292 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3293 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3294 just a single alternative instead of two, make operands 1 and 2
3297 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3300 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3303 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3305 * Makefile.in (lang_opt_files): Add analyzer.opt.
3306 (ANALYZER_OBJS): New.
3307 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3308 tristate.o and ANALYZER_OBJS.
3309 (TEXI_GCCINT_FILES): Add analyzer.texi.
3310 * common.opt (-fanalyzer): New driver option.
3311 * config.in: Regenerate.
3312 * configure: Regenerate.
3313 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3314 (gccdepdir): Also create depdir for "analyzer" subdir.
3315 * digraph.cc: New file.
3316 * digraph.h: New file.
3317 * doc/analyzer.texi: New file.
3318 * doc/gccint.texi ("Static Analyzer") New menu item.
3319 (analyzer.texi): Include it.
3320 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3321 ("Warning Options"): Add static analysis warnings to the list.
3322 (-Wno-analyzer-double-fclose): New option.
3323 (-Wno-analyzer-double-free): New option.
3324 (-Wno-analyzer-exposure-through-output-file): New option.
3325 (-Wno-analyzer-file-leak): New option.
3326 (-Wno-analyzer-free-of-non-heap): New option.
3327 (-Wno-analyzer-malloc-leak): New option.
3328 (-Wno-analyzer-possible-null-argument): New option.
3329 (-Wno-analyzer-possible-null-dereference): New option.
3330 (-Wno-analyzer-null-argument): New option.
3331 (-Wno-analyzer-null-dereference): New option.
3332 (-Wno-analyzer-stale-setjmp-buffer): New option.
3333 (-Wno-analyzer-tainted-array-index): New option.
3334 (-Wno-analyzer-use-after-free): New option.
3335 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3336 (-Wno-analyzer-use-of-uninitialized-value): New option.
3337 (-Wanalyzer-too-complex): New option.
3338 (-fanalyzer-call-summaries): New warning.
3339 (-fanalyzer-checker=): New warning.
3340 (-fanalyzer-fine-grained): New warning.
3341 (-fno-analyzer-state-merge): New warning.
3342 (-fno-analyzer-state-purge): New warning.
3343 (-fanalyzer-transitivity): New warning.
3344 (-fanalyzer-verbose-edges): New warning.
3345 (-fanalyzer-verbose-state-changes): New warning.
3346 (-fanalyzer-verbosity=): New warning.
3347 (-fdump-analyzer): New warning.
3348 (-fdump-analyzer-callgraph): New warning.
3349 (-fdump-analyzer-exploded-graph): New warning.
3350 (-fdump-analyzer-exploded-nodes): New warning.
3351 (-fdump-analyzer-exploded-nodes-2): New warning.
3352 (-fdump-analyzer-exploded-nodes-3): New warning.
3353 (-fdump-analyzer-supergraph): New warning.
3354 * doc/sourcebuild.texi (dg-require-dot): New.
3355 (dg-check-dot): New.
3356 * gdbinit.in (break-on-saved-diagnostic): New command.
3357 * graphviz.cc: New file.
3358 * graphviz.h: New file.
3359 * ordered-hash-map-tests.cc: New file.
3360 * ordered-hash-map.h: New file.
3361 * passes.def (pass_analyzer): Add before
3362 pass_ipa_whole_program_visibility.
3363 * selftest-run-tests.c (selftest::run_tests): Call
3364 selftest::ordered_hash_map_tests_cc_tests.
3365 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3367 * shortest-paths.h: New file.
3368 * timevar.def (TV_ANALYZER): New timevar.
3369 (TV_ANALYZER_SUPERGRAPH): Likewise.
3370 (TV_ANALYZER_STATE_PURGE): Likewise.
3371 (TV_ANALYZER_PLAN): Likewise.
3372 (TV_ANALYZER_SCC): Likewise.
3373 (TV_ANALYZER_WORKLIST): Likewise.
3374 (TV_ANALYZER_DUMP): Likewise.
3375 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3376 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3377 * tree-pass.h (make_pass_analyzer): New decl.
3378 * tristate.cc: New file.
3379 * tristate.h: New file.
3381 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3384 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3385 alternatives 9 and 10.
3387 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3389 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3390 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3391 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3392 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3393 (selftest::hash_map_tests_c_tests): Call it.
3394 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3395 New static constant, using the value of = H::empty_zero_p.
3396 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3397 from default_hash_traits <Value>.
3398 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3400 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3401 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3402 calls to mark_empty with !Descriptor::empty_zero_p.
3403 (hash_table::empty_slow): Conditionalize the memset call with a
3404 check that Descriptor::empty_zero_p; otherwise, loop through the
3405 entries calling mark_empty on them.
3406 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3407 (pointer_hash::empty_zero_p): Likewise.
3408 (pair_hash::empty_zero_p): Likewise.
3409 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3411 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3412 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3413 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3414 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3415 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3416 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3417 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3418 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3420 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3423 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3425 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3426 fix typo on return value.
3428 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3431 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3433 (cgraph_edge::make_speculative): Add param for setting speculative_id
3435 (cgraph_edge::speculative_call_info): Update comments and find reference
3436 by speculative_id for multiple indirect targets.
3437 (cgraph_edge::resolve_speculation): Decrease the speculations
3438 for indirect edge, drop it's speculative if not direct target
3439 left. Update comments.
3440 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3441 (cgraph_node::dump): Print num_speculative_call_targets.
3442 (cgraph_node::verify_node): Don't report error if speculative
3443 edge not include statement.
3444 (cgraph_edge::num_speculative_call_targets_p): New function.
3445 * cgraph.h (int common_target_id): Remove.
3446 (int common_target_probability): Remove.
3447 (num_speculative_call_targets): New variable.
3448 (make_speculative): Add param for setting speculative_id.
3449 (cgraph_edge::num_speculative_call_targets_p): New declare.
3450 (target_prob): New variable.
3451 (speculative_id): New variable.
3452 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3453 call summaries for multiple speculative call targets.
3454 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3455 * ipa-profile.c (struct speculative_call_target): New struct.
3456 (class speculative_call_summary): New class.
3457 (class speculative_call_summaries): New class.
3458 (call_sums): New variable.
3459 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3460 (ipa_profile_write_edge_summary): New function.
3461 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3462 (ipa_profile_dump_all_summaries): New function.
3463 (ipa_profile_read_edge_summary): New function.
3464 (ipa_profile_read_summary_section): New function.
3465 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3466 (ipa_profile): Generate num_speculative_call_targets from
3468 * ipa-ref.h (speculative_id): New variable.
3469 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3470 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3471 common_target_probability. Stream out speculative_id and
3472 num_speculative_call_targets.
3473 (input_edge): Likewise.
3474 * predict.c (dump_prediction): Remove edges count assert to be
3476 * symtab.c (symtab_node::create_reference): Init speculative_id.
3477 (symtab_node::clone_references): Clone speculative_id.
3478 (symtab_node::clone_referring): Clone speculative_id.
3479 (symtab_node::clone_reference): Clone speculative_id.
3480 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3481 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3482 if indirect call contains multiple speculative targets.
3483 * value-prof.h (check_ic_target): Remove.
3484 * value-prof.c (gimple_value_profile_transformations):
3485 Use void function gimple_ic_transform.
3486 * value-prof.c (gimple_ic_transform): Handle topn case.
3487 Fix comment typos. Change it to a void function.
3489 2020-01-13 Andrew Pinski <apinski@marvell.com>
3491 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3492 (octeontx2t98): New define.
3493 (octeontx2t96): New define.
3494 (octeontx2t93): New define.
3495 (octeontx2f95): New define.
3496 (octeontx2f95n): New define.
3497 (octeontx2f95mm): New define.
3498 * config/aarch64/aarch64-tune.md: Regenerate.
3499 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3501 2020-01-13 Jason Merrill <jason@redhat.com>
3503 PR c++/33799 - destroy return value if local cleanup throws.
3504 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3506 2020-01-13 Martin Liska <mliska@suse.cz>
3508 * ipa-cp.c (get_max_overall_size): Use newly
3509 renamed param param_ipa_cp_unit_growth.
3510 * params.opt: Remove legacy param name.
3512 2020-01-13 Martin Sebor <msebor@redhat.com>
3514 PR tree-optimization/93213
3515 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3516 stores to be eliminated.
3518 2020-01-13 Martin Liska <mliska@suse.cz>
3520 * opts.c (print_help): Do not print CL_PARAM
3521 and CL_WARNING for CL_OPTIMIZATION.
3523 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3526 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3527 depending on optimization settings.
3529 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3531 PR tree-optimization/90838
3532 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3533 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3534 argument rather than to initialize temporary for targets that
3535 don't use the mode argument at all. Initialize ctzval to avoid
3538 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3540 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3541 * tree-core.h: Document it.
3542 * gimplify.c (gimplify_omp_workshare): Set it.
3543 * omp-low.c (lower_omp_target): Use it.
3544 * tree-pretty-print.c (dump_omp_clause): Print it.
3546 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3547 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3549 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3551 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3552 * common.opt (fdiagnostics-path-format=): New option.
3553 (diagnostic_path_format): New enum.
3554 (fdiagnostics-show-path-depths): New option.
3555 * coretypes.h (diagnostic_event_id_t): New forward decl.
3556 * diagnostic-color.c (color_dict): Add "path".
3557 * diagnostic-event-id.h: New file.
3558 * diagnostic-format-json.cc (json_from_expanded_location): Make
3560 (json_end_diagnostic): Call context->make_json_for_path if it
3561 exists and the diagnostic has a path.
3562 (diagnostic_output_format_init): Clear context->print_path.
3563 * diagnostic-path.h: New file.
3564 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3565 when printing a run of events in a diagnostic_path so that they
3566 all get the same color.
3567 (layout::m_diagnostic_path_p): New field.
3568 (layout::layout): Initialize it.
3569 (layout::print_any_labels): Don't colorize the label text for an
3570 event in a diagnostic_path.
3571 (gcc_rich_location::add_location_if_nearby): Add
3572 "restrict_to_current_line_spans" and "label" params. Pass the
3573 former to layout.maybe_add_location_range; pass the latter
3574 when calling add_range.
3575 * diagnostic.c: Include "diagnostic-path.h".
3576 (diagnostic_initialize): Initialize context->path_format and
3577 context->show_path_depths.
3578 (diagnostic_show_any_path): New function.
3579 (diagnostic_path::interprocedural_p): New function.
3580 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3581 (simple_diagnostic_path::num_events): New function.
3582 (simple_diagnostic_path::get_event): New function.
3583 (simple_diagnostic_path::add_event): New function.
3584 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3585 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3586 (debug): New overload taking a diagnostic_path *.
3587 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3588 * diagnostic.h (enum diagnostic_path_format): New enum.
3589 (json::value): New forward decl.
3590 (diagnostic_context::path_format): New field.
3591 (diagnostic_context::show_path_depths): New field.
3592 (diagnostic_context::print_path): New callback field.
3593 (diagnostic_context::make_json_for_path): New callback field.
3594 (diagnostic_show_any_path): New decl.
3595 (json_from_expanded_location): New decl.
3596 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3597 (-fdiagnostics-show-path-depths): New option.
3598 (-fdiagnostics-color): Add "path" to description of default
3599 GCC_COLORS; describe it.
3600 (-fdiagnostics-format=json): Document how diagnostic paths are
3601 represented in the JSON output format.
3602 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3603 Add optional params "restrict_to_current_line_spans" and "label".
3604 * opts.c (common_handle_option): Handle
3605 OPT_fdiagnostics_path_format_ and
3606 OPT_fdiagnostics_show_path_depths.
3607 * pretty-print.c: Include "diagnostic-event-id.h".
3608 (pp_format): Implement "%@" format code for printing
3609 diagnostic_event_id_t *.
3610 (selftest::test_pp_format): Add tests for "%@".
3611 * selftest-run-tests.c (selftest::run_tests): Call
3612 selftest::tree_diagnostic_path_cc_tests.
3613 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3614 * toplev.c (general_init): Initialize global_dc->path_format and
3615 global_dc->show_path_depths.
3616 * tree-diagnostic-path.cc: New file.
3617 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3618 non-static. Drop "diagnostic" param in favor of storing the
3619 original value of "where" and re-using it.
3620 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3621 maybe_unwind_expanded_macro_loc.
3622 (tree_diagnostics_defaults): Initialize context->print_path and
3623 context->make_json_for_path.
3624 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3626 (default_tree_make_json_for_path): New decl.
3627 (maybe_unwind_expanded_macro_loc): New decl.
3629 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3631 PR tree-optimization/93210
3632 * fold-const.h (native_encode_initializer,
3633 can_native_interpret_type_p): Declare.
3634 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3636 (native_encode_initializer): New function, moved from dwarf2out.c.
3637 Adjust to native_encode_expr compatible arguments, including dry-run
3638 and partial extraction modes. Don't handle STRING_CST.
3639 (can_native_interpret_type_p): No longer static.
3640 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3641 offset / BITS_PER_UNIT fits into int and don't call it if
3642 can_native_interpret_type_p fails. If suboff is NULL and for
3643 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3644 native_encode_initializer.
3645 (fold_const_aggregate_ref_1): Formatting fix.
3646 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3647 (tree_add_const_value_attribute): Adjust caller.
3649 PR tree-optimization/90838
3650 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3651 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3652 CTZ_DEFINED_VALUE_AT_ZERO.
3654 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3657 * lra-constraints.c (match_reload): Permit input operands have the
3658 same mode as output while other input operands have a different
3661 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3663 PR tree-optimization/90838
3664 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3665 (check_ctz_string): Likewise.
3666 (optimize_count_trailing_zeroes): Likewise.
3667 (simplify_count_trailing_zeroes): Likewise.
3668 (pass_forwprop::execute): Try ctz simplification.
3669 * match.pd: Add matching for ctz idioms.
3671 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3673 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3675 (aarch64_invalid_unary_op): New function for target hook.
3676 (aarch64_invalid_binary_op): New function for target hook.
3678 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3680 * config.gcc: Add arm_bf16.h.
3681 * config/aarch64/aarch64-builtins.c
3682 (aarch64_simd_builtin_std_type): Add BFmode.
3683 (aarch64_init_simd_builtin_types): Define element types for vector
3685 (aarch64_init_bf16_types): New function.
3686 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3687 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3689 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3690 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3692 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3693 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3694 * config/aarch64/aarch64.c
3695 (aarch64_classify_vector_mode): Add support for BF types.
3696 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3697 (aarch64_vq_mode): Add support for BF types.
3698 (aarch64_simd_container_mode): Add support for BF types.
3699 (aarch64_mangle_type): Add support for BF scalar type.
3700 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3701 * config/aarch64/arm_bf16.h: New file.
3702 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3703 * config/aarch64/iterators.md: Add BF types to mode attributes.
3704 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3706 2020-01-10 Jason Merrill <jason@redhat.com>
3708 PR c++/93173 - incorrect tree sharing.
3709 * gimplify.c (copy_if_shared): No longer static.
3710 * gimplify.h: Declare it.
3712 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3714 * doc/invoke.texi (-msve-vector-bits=): Document that
3715 -msve-vector-bits=128 now generates VL-specific code for
3716 little-endian targets.
3717 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3718 build_vector_type_for_mode to construct the data vector types.
3719 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3720 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3721 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3722 for 128-bit vectors.
3724 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3726 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3729 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3731 * config/aarch64/aarch64-builtins.c
3732 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3733 rather than checking the number of elements and the element mode.
3735 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3737 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3738 get_related_vectype_for_scalar_type rather than build_vector_type
3739 to create the index type for a conditional reduction.
3741 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3743 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3744 for any type of gather or scatter, including strided accesses.
3746 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3748 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3751 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3753 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3755 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3756 parameter and its use to reset DR_OFFSET's.
3757 (vect_transform_loop): Remove orig_drs_init argument.
3758 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3759 member of dr_vec_info rather than the offset of the associated
3760 data_reference's innermost_loop_behavior.
3761 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3762 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3763 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3764 get_dr_vinfo_offset.
3765 (vectorizable_store): Likewise.
3766 (vectorizable_load): Likewise.
3768 2020-01-10 Richard Biener <rguenther@suse.de>
3770 * gimple-ssa-store-merging
3771 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3773 2020-01-10 Martin Liska <mliska@suse.cz>
3776 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3777 encapsulation that was there before r280040.
3779 2020-01-10 Richard Biener <rguenther@suse.de>
3782 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3783 sequences to avoid walking them again for secondary opportunities.
3784 (pass_lower_eh_dispatch::execute): Instead actually insert
3787 2020-01-10 Richard Biener <rguenther@suse.de>
3790 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3791 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3792 avoid quadraticness.
3794 2020-01-10 Martin Jambor <mjambor@suse.cz>
3796 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3797 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3798 to get param_ipa_sra_max_replacements.
3799 (param_splitting_across_edge): Pass the caller to
3800 pull_accesses_from_callee.
3802 2020-01-10 Martin Jambor <mjambor@suse.cz>
3804 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3805 * ipa-cp.c (max_new_size): Removed.
3806 (orig_overall_size): New variable.
3807 (get_max_overall_size): New function.
3808 (estimate_local_effects): Use it. Adjust dump.
3809 (decide_about_value): Likewise.
3810 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3811 orig_overall_size. Adjust dump.
3812 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3814 2020-01-10 Martin Jambor <mjambor@suse.cz>
3816 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3817 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3818 instead of param_ipa_max_agg_items.
3819 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3820 optimization info for the callee.
3822 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3824 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3825 markers if debug_inline_points is false.
3827 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3829 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3831 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3832 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3833 aarch64-sve-builtins-sve2.h.
3834 (aarch64-sve-builtins-sve2.o): New rule.
3835 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3836 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3837 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3838 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3839 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3840 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3842 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3843 instructions that are handled here.
3844 (@cond_asrd<mode>): Generalize to...
3845 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3846 (*cond_asrd<mode>_2): Generalize to...
3847 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3848 (*cond_asrd<mode>_z): Generalize to...
3849 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3850 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3851 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3852 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3853 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3855 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3856 (@aarch64_scatter_stnt<mode>): Likewise.
3857 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3858 (@aarch64_mul_lane_<mode>): Likewise.
3859 (@aarch64_sve_suqadd<mode>_const): Likewise.
3860 (*<sur>h<addsub><mode>): Generalize to...
3861 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3863 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3864 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3865 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3866 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3867 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3868 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3869 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3870 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3871 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3872 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3873 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3874 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3875 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3876 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3877 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3878 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3879 (@aarch64_sve2_xar<mode>): Likewise.
3880 (@aarch64_sve2_bcax<mode>): Likewise.
3881 (*aarch64_sve2_eor3<mode>): Rename to...
3882 (@aarch64_sve2_eor3<mode>): ...this.
3883 (@aarch64_sve2_bsl<mode>): New expander.
3884 (@aarch64_sve2_nbsl<mode>): Likewise.
3885 (@aarch64_sve2_bsl1n<mode>): Likewise.
3886 (@aarch64_sve2_bsl2n<mode>): Likewise.
3887 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3888 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3889 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3890 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3891 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3892 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3893 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3894 (<su>mull<bt><Vwide>): Generalize to...
3895 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3897 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3898 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3899 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3900 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3901 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3902 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3903 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3904 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3905 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3906 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3907 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3908 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3909 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3910 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3911 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3912 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3913 (<SHRNB:r>shrnb<mode>): Generalize to...
3914 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3916 (<SHRNT:r>shrnt<mode>): Generalize to...
3917 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3919 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3920 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3921 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3922 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3923 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3924 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3925 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3926 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3927 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3928 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3929 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3930 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3931 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3932 (@aarch64_sve2_cvtnt<mode>): Likewise.
3933 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3934 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3935 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3936 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3937 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3938 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3939 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3940 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3941 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3942 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3943 (@aarch64_sve2_pmul<mode>): Likewise.
3944 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3945 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3946 (@aarch64_sve2_tbl2<mode>): Likewise.
3947 (@aarch64_sve2_tbx<mode>): Likewise.
3948 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3949 (@aarch64_sve2_histcnt<mode>): Likewise.
3950 (@aarch64_sve2_histseg<mode>): Likewise.
3951 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3952 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3953 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3954 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3955 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3956 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3957 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3958 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3959 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3960 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3961 (SVE2_PMULL_PAIR_I): New mode iterators.
3962 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3963 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3964 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3965 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3966 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3967 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3968 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3969 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3970 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3971 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3972 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3973 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3974 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3975 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3976 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3977 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3978 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3979 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3980 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3981 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3982 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3983 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3984 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3985 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3986 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3987 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3988 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3989 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3990 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3991 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3992 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3994 (VNARROW, Ventype): New mode attributes.
3995 (Vewtype): Handle VNx2DI. Fix typo in comment.
3996 (VDOUBLE): New mode attribute.
3997 (sve_lane_con): Handle VNx8HI.
3998 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3999 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
4000 (sve_int_op, sve_int_op_rev): Handle the above codes.
4001 (sve_pred_int_rhs2_operand): Likewise.
4002 (MULLBT, SHRNB, SHRNT): Delete.
4003 (SVE_INT_SHIFT_IMM): New int iterator.
4004 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
4005 and UNSPEC_WHILEHS for TARGET_SVE2.
4006 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
4007 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
4008 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
4009 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
4010 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
4011 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
4012 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
4013 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
4014 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
4015 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
4016 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
4017 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
4018 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
4019 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
4020 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
4021 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
4022 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
4023 (optab): Handle the new unspecs.
4024 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
4026 (lr): Handle the new unspecs.
4028 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
4029 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
4030 (sve_int_qsub_op): New int attributes.
4031 (sve_fp_op, rot): Handle the new unspecs.
4032 * config/aarch64/aarch64-sve-builtins.h
4033 (function_resolver::require_matching_pointer_type): Declare.
4034 (function_resolver::resolve_unary): Add an optional boolean argument.
4035 (function_resolver::finish_opt_n_resolution): Add an optional
4036 type_suffix_index argument.
4037 (gimple_folder::redirect_call): Declare.
4038 (gimple_expander::prepare_gather_address_operands): Add an optional
4040 * config/aarch64/aarch64-sve-builtins.cc: Include
4041 aarch64-sve-builtins-sve2.h.
4042 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4043 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4044 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4045 (TYPES_hsd_integer): Use TYPES_hsd_signed.
4046 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4047 (TYPES_s_unsigned): Likewise.
4048 (TYPES_s_integer): Use TYPES_s_unsigned.
4049 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4050 (TYPES_sd_integer): Use them.
4051 (TYPES_d_unsigned): New macro.
4052 (TYPES_d_integer): Use it.
4053 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4054 (TYPES_cvt_narrow): Likewise.
4055 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4056 (preds_mx): New variable.
4057 (function_builder::add_overloaded_function): Allow the new feature
4058 set to be more restrictive than the original one.
4059 (function_resolver::infer_pointer_type): Remove qualifiers from
4060 the pointer type before printing it.
4061 (function_resolver::require_matching_pointer_type): New function.
4062 (function_resolver::resolve_sv_displacement): Handle functions
4063 that don't support 32-bit vector indices or svint32_t vector offsets.
4064 (function_resolver::finish_opt_n_resolution): Take the inferred type
4065 as a separate argument.
4066 (function_resolver::resolve_unary): Optionally treat all forms in
4067 the same way as normal merging functions.
4068 (gimple_folder::redirect_call): New function.
4069 (function_expander::prepare_gather_address_operands): Add an argument
4070 that says whether scaled forms are available. If they aren't,
4071 handle scaling of vector indices and don't add the extension and
4073 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4074 fall back to using cond_* instead.
4075 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4076 Split out the member variables into...
4077 (rtx_code_function_base): ...this new base class.
4078 (rtx_code_function_rotated): Inherit rtx_code_function_base.
4079 (unspec_based_function): Split out the member variables into...
4080 (unspec_based_function_base): ...this new base class.
4081 (unspec_based_function_rotated): Inherit unspec_based_function_base.
4082 (unspec_based_function_exact_insn): New class.
4083 (unspec_based_add_function, unspec_based_add_lane_function)
4084 (unspec_based_lane_function, unspec_based_pred_function)
4085 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4086 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4087 (unspec_based_sub_function, unspec_based_sub_lane_function): New
4089 (unspec_based_fused_function): New class.
4090 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4091 (unspec_based_fused_lane_function): New class.
4092 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4094 (CODE_FOR_MODE1): New macro.
4095 (fixed_insn_function): New class.
4096 (while_comparison): Likewise.
4097 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4098 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4099 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4100 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4101 (load_gather_sv_restricted, shift_left_imm_long): Declare.
4102 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4103 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4104 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4105 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4106 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4107 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4108 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4109 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4110 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4111 Also add an initial argument for unary_convert_narrowt, regardless
4112 of the predication type.
4113 (build_32_64): Allow loads and stores to specify MODE_none.
4114 (build_sv_index64, build_sv_uint_offset): New functions.
4115 (long_type_suffix): New function.
4116 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4117 (binary_imm_long_base, load_gather_sv_base): Likewise.
4118 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4119 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4120 (unary_narrowb_base, unary_narrowt_base): Likewise.
4121 (binary_long_lane_def, binary_long_lane): New shape.
4122 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4123 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4124 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4125 (binary_to_uint_def, binary_to_uint): Likewise.
4126 (binary_wide_def, binary_wide): Likewise.
4127 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4128 (compare_def, compare): Likewise.
4129 (compare_ptr_def, compare_ptr): Likewise.
4130 (load_ext_gather_index_restricted_def,
4131 load_ext_gather_index_restricted): Likewise.
4132 (load_ext_gather_offset_restricted_def,
4133 load_ext_gather_offset_restricted): Likewise.
4134 (load_gather_sv_def): Inherit from load_gather_sv_base.
4135 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4136 (shift_left_imm_def, shift_left_imm): Likewise.
4137 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4138 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4139 (store_scatter_index_restricted_def,
4140 store_scatter_index_restricted): Likewise.
4141 (store_scatter_offset_restricted_def,
4142 store_scatter_offset_restricted): Likewise.
4143 (tbl_tuple_def, tbl_tuple): Likewise.
4144 (ternary_long_lane_def, ternary_long_lane): Likewise.
4145 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4146 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4147 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4148 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4149 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4150 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4151 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4152 (ternary_uint_def, ternary_uint): Likewise.
4153 (unary_convert): Fix typo in comment.
4154 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4155 (unary_long_def, unary_long): Likewise.
4156 (unary_narrowb_def, unary_narrowb): Likewise.
4157 (unary_narrowt_def, unary_narrowt): Likewise.
4158 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4159 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4160 (unary_to_int_def, unary_to_int): Likewise.
4161 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4162 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4163 (svasrd_impl): Delete.
4164 (svcadd_impl::expand): Handle integer operations too.
4165 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4166 new functions to derive the unspec numbers.
4167 (svmla_svmls_lane_impl): Replace with...
4168 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
4169 integer operations too.
4170 (svwhile_impl): Rename to...
4171 (svwhilelx_impl): ...this and inherit from while_comparison.
4172 (svasrd): Use unspec_based_function.
4173 (svmla_lane): Use svmla_lane_impl.
4174 (svmls_lane): Use svmls_lane_impl.
4175 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4176 (svwhilele, svwhilelt): Use svwhilelx_impl.
4177 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4178 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4179 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4180 * config/aarch64/aarch64-sve-builtins.def: Include
4181 aarch64-sve-builtins-sve2.def.
4183 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4185 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4186 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4187 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4188 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4189 immediates as well as vector ones.
4190 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4191 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4192 (aarch64_sve_qsub_immediate): Update calls accordingly.
4194 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4196 * config/aarch64/aarch64-sve2.md: Add banner comments.
4197 (<su>mulh<r>s<mode>3): Move further up file.
4198 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4199 (*aarch64_sve2_sra<mode>): Move further down file.
4200 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4202 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4204 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4206 (while_optab_cmp): Handle them.
4207 * config/aarch64/aarch64-sve.md
4208 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4209 and add a "@" marker.
4210 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4211 instead of gen_aarch64_sve2_while_ptest.
4212 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4214 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4216 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4217 (UNSPEC_WHILELE): ...this.
4218 (UNSPEC_WHILE_LO): Rename to...
4219 (UNSPEC_WHILELO): ...this.
4220 (UNSPEC_WHILE_LS): Rename to...
4221 (UNSPEC_WHILELS): ...this.
4222 (UNSPEC_WHILE_LT): Rename to...
4223 (UNSPEC_WHILELT): ...this.
4224 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4225 (cmp_op, while_optab_cmp): Likewise.
4226 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4227 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4228 (svwhilelt): Likewise.
4230 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4232 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4233 (unary_to_uint): Define.
4234 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4235 (unary_count): Rename to...
4236 (unary_to_uint_def, unary_to_uint): ...this.
4237 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4239 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4241 * config/aarch64/aarch64-sve-builtins-functions.h
4242 (code_for_mode_function): New class.
4243 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4244 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4245 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4246 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4247 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4249 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4251 * config/aarch64/iterators.md (addsub): New code attribute.
4252 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4254 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4255 in the asm string and attributes. Fix indentation.
4256 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4258 (@aarch64_sve_<optab><mode>): ...this.
4259 * config/aarch64/aarch64-sve-builtins.h
4260 (function_expander::expand_signed_unpred_op): Delete.
4261 * config/aarch64/aarch64-sve-builtins.cc
4262 (function_expander::expand_signed_unpred_op): Likewise.
4263 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4264 try using code_for_aarch64_sve instead.
4265 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4266 (svqsub_impl): Likewise.
4267 (svqadd, svqsub): Use rtx_code_function instead.
4269 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4271 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4272 (HADDSUB, sur, addsub): Remove them.
4274 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4276 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4277 functions the same way, rather than singling out those that
4278 aren't mapped directly to optabs.
4280 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4282 * target.def (compatible_vector_types_p): New target hook.
4283 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4284 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4285 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4286 * doc/tm.texi: Regenerate.
4287 * gimple-expr.c: Include target.h.
4288 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4289 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4291 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4292 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4293 Use the original predicate if it already has a suitable type.
4295 2020-01-09 Martin Jambor <mjambor@suse.cz>
4297 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4298 resolve_speculation and redirect_call_stmt_to_callee static. Change
4299 return type of set_call_stmt to cgraph_edge *.
4300 * auto-profile.c (afdo_indirect_call): Adjust call to
4301 redirect_call_stmt_to_callee.
4302 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4303 make the this pointer explicit, adjust self-recursive calls and the
4304 call top make_direct. Return the resulting edge.
4305 (cgraph_edge::remove): Make this pointer explicit.
4306 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4307 (cgraph_edge::make_direct): Likewise, adjust call to
4308 resolve_speculation.
4309 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4310 call to set_call_stmt.
4311 (cgraph_update_edges_for_call_stmt_node): Update call to
4312 set_call_stmt and remove.
4313 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4314 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4315 (cgraph_node::create_edge_including_clones): Moved "first" definition
4316 of edge to the block where it was used. Adjusted calls to
4318 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4319 cgraph_edge::remove.
4320 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4321 make_direct and redirect_call_stmt_to_callee.
4322 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4323 resolve_speculation and make_direct.
4324 * ipa-inline-transform.c (inline_transform): Adjust call to
4325 redirect_call_stmt_to_callee.
4326 (check_speculations_1):: Adjust call to resolve_speculation.
4327 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4328 resolve-speculation.
4329 (inline_small_functions): Adjust call to resolve_speculation.
4330 (ipa_inline): Likewise.
4331 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4333 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4334 safe with regards to edge removal, adjust calls to
4335 redirect_call_stmt_to_callee.
4336 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4337 and redirect_call_stmt_to_callee.
4338 * multiple_target.c (create_dispatcher_calls): Adjust call to
4339 redirect_call_stmt_to_callee
4340 (redirect_to_specific_clone): Likewise.
4341 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4342 Adjust calls to cgraph_edge::remove.
4343 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4344 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4345 (expand_call_inline): Adjust call to cgraph_edge::remove.
4347 2020-01-09 Martin Liska <mliska@suse.cz>
4349 * params.opt: Set Optimization for
4350 param_max_speculative_devirt_maydefs.
4352 2020-01-09 Martin Sebor <msebor@redhat.com>
4356 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4358 2020-01-09 Martin Liska <mliska@suse.cz>
4360 * auto-profile.c (auto_profile): Use opt_for_fn
4362 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4363 (propagate_vals_across_arith_jfunc): Likewise.
4364 (hint_time_bonus): Likewise.
4365 (incorporate_penalties): Likewise.
4366 (good_cloning_opportunity_p): Likewise.
4367 (perform_estimation_of_a_value): Likewise.
4368 (estimate_local_effects): Likewise.
4369 (ipcp_propagate_stage): Likewise.
4370 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4371 (set_switch_stmt_execution_predicate): Likewise.
4372 (analyze_function_body): Likewise.
4373 * ipa-inline-analysis.c (offline_size): Likewise.
4374 * ipa-inline.c (early_inliner): Likewise.
4375 * ipa-prop.c (ipa_analyze_node): Likewise.
4376 (ipcp_transform_function): Likewise.
4377 * ipa-sra.c (process_scan_results): Likewise.
4378 (ipa_sra_summarize_function): Likewise.
4379 * params.opt: Rename ipcp-unit-growth to
4380 ipa-cp-unit-growth. Add Optimization for various
4381 IPA-related parameters.
4383 2020-01-09 Richard Biener <rguenther@suse.de>
4386 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4388 2020-01-09 Richard Biener <rguenther@suse.de>
4390 PR tree-optimization/93040
4391 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4393 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4395 * common/config/avr/avr-common.c (avr_option_optimization_table)
4396 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4398 2020-01-09 Martin Liska <mliska@suse.cz>
4400 * cgraphclones.c (symbol_table::materialize_all_clones):
4401 Use cgraph_node::dump_name.
4403 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4406 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4407 output_operand_lossage instead of gcc_unreachable.
4408 * doc/md.texi (riscv f constraint): Fix typo.
4411 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4412 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4413 CONST_SCALAR_INT_P instead of CONST_INT_P.
4414 (*subv<mode>4_1): Rename to ...
4415 (subv<mode>4_1): ... this.
4416 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4417 define_insn_and_split patterns.
4418 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4421 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4423 * vec.c (class selftest::count_dtor): New class.
4424 (selftest::test_auto_delete_vec): New test.
4425 (selftest::vec_c_tests): Call it.
4426 * vec.h (class auto_delete_vec): New class template.
4427 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4429 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4431 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4433 2020-01-08 Jim Wilson <jimw@sifive.com>
4435 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4436 use of TLS_MODEL_LOCAL_EXEC when not pic.
4438 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4440 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4443 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4446 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4447 *stack_protect_set_3 peephole2): Also check that the second
4448 insns source is general_operand.
4451 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4452 predicate for output operand instead of register_operand.
4453 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4454 memory destination and non-memory operands[2].
4456 2020-01-08 Martin Liska <mliska@suse.cz>
4458 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4459 ::dump_asm_name instead of (::name or ::asm_name).
4460 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4461 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4462 (analyze_functions): Likewise.
4463 (expand_all_functions): Likewise.
4464 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4465 (propagate_bits_across_jump_function): Likewise.
4466 (dump_profile_updates): Likewise.
4467 (ipcp_store_bits_results): Likewise.
4468 (ipcp_store_vr_results): Likewise.
4469 * ipa-devirt.c (dump_targets): Likewise.
4470 * ipa-fnsummary.c (analyze_function_body): Likewise.
4471 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4472 (process_hsa_functions): Likewise.
4473 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4474 (set_alias_uids): Likewise.
4475 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4476 * ipa-inline.c (recursive_inlining): Likewise.
4477 (inline_to_all_callers_1): Likewise.
4478 (ipa_inline): Likewise.
4479 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4480 (ipa_propagate_frequency): Likewise.
4481 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4482 (remove_described_reference): Likewise.
4483 * ipa-pure-const.c (worse_state): Likewise.
4484 (check_retval_uses): Likewise.
4485 (analyze_function): Likewise.
4486 (propagate_pure_const): Likewise.
4487 (propagate_nothrow): Likewise.
4488 (dump_malloc_lattice): Likewise.
4489 (propagate_malloc): Likewise.
4490 (pass_local_pure_const::execute): Likewise.
4491 * ipa-visibility.c (optimize_weakref): Likewise.
4492 (function_and_variable_visibility): Likewise.
4493 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4494 (ipa_discover_variable_flags): Likewise.
4495 * lto-streamer-out.c (output_function): Likewise.
4496 (output_constructor): Likewise.
4497 * tree-inline.c (copy_bb): Likewise.
4498 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4499 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4501 2020-01-08 Richard Biener <rguenther@suse.de>
4504 * tree-eh.c (sink_clobbers): Update virtual operands for
4505 the first and last stmt only. Add a dry-run capability.
4506 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4507 after CFG manipulations and in RPO order to catch all
4508 secondary opportunities reliably.
4510 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4513 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4515 2019-01-08 Richard Biener <rguenther@suse.de>
4518 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4519 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4520 virtual operand, also updating SSA use.
4521 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4522 Update stmt after resetting virtual operand.
4523 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4524 * gimple-iterator.c (gsi_remove): When not removing the stmt
4525 permanently do not delink immediate uses or mark the stmt modified.
4527 2020-01-08 Martin Liska <mliska@suse.cz>
4529 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4530 (ipa_call_context::estimate_size_and_time): Likewise.
4531 (inline_analyze_function): Likewise.
4533 2020-01-08 Martin Liska <mliska@suse.cz>
4535 * cgraph.c (cgraph_node::dump): Use systematically
4538 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4540 Add -nodevicespecs option for avr.
4543 * config/avr/avr.opt (-nodevicespecs): New driver option.
4544 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4545 "-specs=device-specs/..." if that option is not set.
4546 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4548 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4550 Implement 64-bit double functions for avr.
4553 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4554 --with-double-comparison.
4555 * doc/install.texi: Document them.
4556 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4557 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4558 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4559 * doc/invoke.texi (AVR Built-in Macros): Document them.
4560 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4561 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4562 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4564 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4567 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4568 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4569 when only building rm-profile multilibs.
4571 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4574 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4575 lattice for a value to check.
4576 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4577 finite propagation in self-recursive scc.
4579 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4581 * ipa-inline.c (caller_growth_limits): Restore the AND.
4583 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4585 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4586 (VEC_ALLREG_ALT): New iterator.
4587 (VEC_ALLREG_INT_MODE): New iterator.
4588 (VCMP_MODE): New iterator.
4589 (VCMP_MODE_INT): New iterator.
4590 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4591 (vec_cmp<u>v64qidi): New define_expand.
4592 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4593 (vec_cmpu<mode>di_exec): New define_expand.
4594 (vec_cmp<u>v64qidi_exec): New define_expand.
4595 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4596 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4597 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4598 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4599 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4600 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4601 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4602 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4603 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4604 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4606 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4607 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4609 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4611 * config/gcn/constraints.md (DA): Update description and match.
4613 (Db): New constraint.
4614 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4616 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4617 Implement 'Db' mixed immediate type.
4618 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4619 (addcv64si3_dup<exec_vcc>): Delete.
4620 (subcv64si3<exec_vcc>): Rework constraints.
4621 (addv64di3): Rework constraints.
4622 (addv64di3_exec): Rework constraints.
4623 (subv64di3): Rework constraints.
4624 (addv64di3_dup): Delete.
4625 (addv64di3_dup_exec): Delete.
4626 (addv64di3_zext): Rework constraints.
4627 (addv64di3_zext_exec): Rework constraints.
4628 (addv64di3_zext_dup): Rework constraints.
4629 (addv64di3_zext_dup_exec): Rework constraints.
4630 (addv64di3_zext_dup2): Rework constraints.
4631 (addv64di3_zext_dup2_exec): Rework constraints.
4632 (addv64di3_sext_dup2): Rework constraints.
4633 (addv64di3_sext_dup2_exec): Rework constraints.
4635 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4637 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4638 existing target checks.
4640 2020-01-07 Richard Biener <rguenther@suse.de>
4642 * doc/install.texi: Bump minimal supported MPC version.
4644 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4646 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4647 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4648 * langhooks.c: Include stor-layout.h.
4649 (lhd_simulate_enum_decl): New function.
4650 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4651 handle_arm_sve_h for the LTO frontend.
4652 (register_vector_type): Cope with null returns from pushdecl.
4654 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4656 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4657 (aarch64_sve::nvectors_if_data_type): Replace with...
4658 (aarch64_sve::builtin_type_p): ...this.
4659 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4660 (find_vector_type): Delete.
4661 (add_sve_type_attribute): New function.
4662 (lookup_sve_type_attribute): Likewise.
4663 (register_builtin_types): Add an "SVE type" attribute to each type.
4664 (register_tuple_type): Likewise.
4665 (svbool_type_p, nvectors_if_data_type): Delete.
4666 (mangle_builtin_type): Use lookup_sve_type_attribute.
4667 (builtin_type_p): Likewise. Add an overload that returns the
4668 number of constituent vector and predicate registers.
4669 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4670 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4671 instead of aarch64_sve_argument_p.
4672 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4673 (aarch64_pass_by_reference): Likewise.
4674 (aarch64_function_value_1): Likewise.
4675 (aarch64_return_in_memory): Likewise.
4676 (aarch64_layout_arg): Likewise.
4678 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4680 PR tree-optimization/93156
4681 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4682 least significant bit is always clear.
4684 PR tree-optimization/93118
4685 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4686 simplifier with two intermediate conversions.
4688 2020-01-07 Martin Liska <mliska@suse.cz>
4690 * params.opt: Add Optimization for various parameters.
4692 2020-01-07 Martin Liska <mliska@suse.cz>
4695 * doc/extend.texi: Explain cloning for target_clone
4698 2020-01-07 Martin Liska <mliska@suse.cz>
4700 PR tree-optimization/92860
4701 * common.opt: Make in Optimization option
4702 as it is affected by -O0, which is an Optimization
4704 * tree-inline.c (tree_inlinable_function_p):
4705 Use opt_for_fn for warn_inline.
4706 (expand_call_inline): Likewise.
4708 2020-01-07 Martin Liska <mliska@suse.cz>
4710 PR tree-optimization/92860
4711 * common.opt: Make flag_ree as optimization
4714 2020-01-07 Martin Liska <mliska@suse.cz>
4716 PR optimization/92860
4717 * params.opt: Mark param_min_crossjump_insns with Optimization
4720 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4722 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4723 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4725 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4727 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4728 helper function to return the valid addressing formats for a given
4729 hard register and mode.
4730 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4732 * config/rs6000/constraints.md (Q constraint): Update
4734 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4737 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4738 Use 'Q' for doing vector extract from memory.
4739 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4741 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4742 doing vector extract from memory.
4743 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4744 extract from memory.
4746 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4747 for the offset being 34-bits when -mcpu=future is used.
4749 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4751 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4752 instead of cmpib_comparison_operator in cmpib patterns.
4753 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4754 of cmpib_comparison_operator. Revise comment.
4756 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4758 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4759 in an IFN_DIV_POW2 node to be equal.
4761 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4763 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4764 (vect_check_scalar_mask): ...this.
4765 (vectorizable_store, vectorizable_load): Update call accordingly.
4766 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4767 argument in calls to conditional internal functions.
4769 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4771 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4772 '0' matching inputs.
4773 (subv64di3_exec): Likewise.
4775 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4777 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4778 * doc/md.texi (movstr): Likewise.
4780 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4782 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4785 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4787 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4789 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4790 to a temporary file and use move-if-change to update the real
4791 file where necessary.
4793 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4795 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4796 rather than Upa for CPY /M.
4798 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4800 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4803 2020-01-06 Martin Liska <mliska@suse.cz>
4805 PR tree-optimization/92860
4806 * params.opt: Mark param_max_combine_insns with Optimization
4809 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4812 * config/i386/i386.md (SWIDWI): New mode iterator.
4813 (DWI, dwi): Add TImode variants.
4814 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4815 <general_hilo_operand> instead of <general_operand>. Use
4816 CONST_SCALAR_INT_P instead of CONST_INT_P.
4817 (*addv<mode>4_1): Rename to ...
4818 (addv<mode>4_1): ... this.
4819 (QWI): New mode attribute.
4820 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4821 define_insn_and_split patterns.
4822 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4824 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4825 <general_hilo_operand> instead of <general_operand>.
4826 (*addcarry<mode>_1): New define_insn.
4827 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4829 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4831 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4832 Use "call" instead of "set".
4834 2020-01-03 Martin Jambor <mjambor@suse.cz>
4837 * ipa-cp.c (print_all_lattices): Skip functions without info.
4839 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4842 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4843 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4844 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4845 for 'e' simd clones.
4848 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4850 (mprefer-vector-width=): Add Save.
4851 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4852 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4853 (ix86_debug_options, ix86_function_specific_print): Adjust
4854 ix86_target_string callers.
4855 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4856 (ix86_valid_target_attribute_tree): Likewise.
4857 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4858 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4859 ix86_target_string caller.
4862 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4863 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4864 instead of gen_int_shift_amount + convert_modes.
4866 PR rtl-optimization/93088
4867 * loop-iv.c (find_single_def_src): Punt after looking through
4868 128 reg copies for regs with single definitions. Move definitions
4871 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4873 * config/arm/arm-c.c (arm_cpu_builtins): Define
4874 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4875 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4876 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4877 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4878 * config/arm/arm-tables.opt: Regenerated.
4879 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4880 arm_arch_i8mm and arm_arch_bf16 when enabled.
4881 * config/arm/arm.h (TARGET_I8MM): New macro.
4882 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4883 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4884 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4885 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4886 (v8_6_a_simd_variants): New.
4887 (v8_*_a_simd_variants): Add i8mm and bf16.
4888 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4890 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4893 * predict.c (compute_function_frequency): Don't call
4894 warn_function_cold on functions that already have cold attribute.
4896 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4899 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4900 COMDAT group function labels in .data.rel.ro.local section.
4901 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4904 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4905 comparison_operator in B and S integer comparisons. Likewise, use
4906 ordered_comparison_operator instead of cmpib_comparison_operator in
4908 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4910 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4912 Update copyright years.
4914 * gcc.c (process_command): Update copyright notice dates.
4915 * gcov-dump.c (print_version): Ditto.
4916 * gcov.c (print_version): Ditto.
4917 * gcov-tool.c (print_version): Ditto.
4918 * gengtype.c (create_file): Ditto.
4919 * doc/cpp.texi: Bump @copying's copyright year.
4920 * doc/cppinternals.texi: Ditto.
4921 * doc/gcc.texi: Ditto.
4922 * doc/gccint.texi: Ditto.
4923 * doc/gcov.texi: Ditto.
4924 * doc/install.texi: Ditto.
4925 * doc/invoke.texi: Ditto.
4927 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4929 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4932 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4934 PR tree-optimization/93098
4935 * match.pd (popcount): For shift amounts, use integer_onep
4936 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4937 tests. Make sure that precision is power of two larger than or equal
4938 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4939 instead of ULL suffixed constants. Formatting fixes.
4941 Copyright (C) 2020 Free Software Foundation, Inc.
4943 Copying and distribution of this file, with or without modification,
4944 are permitted in any medium without royalty provided the copyright
4945 notice and this notice are preserved.