1 2020-02-27 Richard Biener <rguenther@suse.de>
3 PR tree-optimization/93508
4 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
5 non-_CHK variants. Valueize their length arguments.
7 2020-02-27 Richard Biener <rguenther@suse.de>
9 PR tree-optimization/93953
10 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
11 to the hash-map entry.
13 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
15 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
17 2020-02-27 Mark Williams <mwilliams@fb.com>
19 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
20 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
21 -ffile-prefix-map and -fmacro-prefix-map.
22 * lto-streamer-out.c: Include file-prefix-map.h.
23 (lto_output_location): Remap the file part of locations.
25 2020-02-27 Jakub Jelinek <jakub@redhat.com>
28 * gimplify.c (gimplify_init_constructor): Don't promote readonly
29 DECL_REGISTER variables to TREE_STATIC.
31 PR tree-optimization/93582
32 PR tree-optimization/93945
33 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
34 non-zero INTEGER_CST second argument and ref->offset or ref->size
35 not a multiple of BITS_PER_UNIT.
37 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
39 * doc/install.texi (Binaries): Update description of BullFreeware.
41 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
45 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
46 C++ Language Options, Warning Options, and Static Analyzer
47 Options lists. Document negative form of options enabled by
48 default. Move some things around to more accurately sort
50 (C++ Dialect Options, Warning Options, Static Analyzer
51 Options): Document negative form of options when enabled by
52 default. Move some things around to more accurately sort
53 warnings by category. Add some missing index entries.
56 2020-02-26 Carl Love <cel@us.ibm.com>
59 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
60 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
61 for the vector unsigned short arguments. It is also listed as the
62 name of the built-in for arguments vector unsigned short,
63 vector unsigned int and vector unsigned long long built-ins. The
64 name of the builtins for these arguments should be:
65 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
66 __builtin_crypto_vpmsumd respectively.
68 2020-02-26 Richard Biener <rguenther@suse.de>
70 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
73 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
76 * optabs-tree.c (supportable_convert_operation): Reject types with
79 2020-02-26 David Malcolm <dmalcolm@redhat.com>
81 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
83 2020-02-26 Jakub Jelinek <jakub@redhat.com>
85 PR tree-optimization/93820
86 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
87 argument to ALL_INTEGER_CST_P boolean.
88 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
89 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
90 adjacent INTEGER_CST store into merged_store->only_constants like
93 2020-02-25 Jakub Jelinek <jakub@redhat.com>
96 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
98 * cfghooks.c (verify_flow_info): Likewise.
99 * predict.c (combine_predictions_for_bb): Likewise.
100 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
101 sucessor -> successor.
102 (find_traces_1_round): Fix comment typo, destinarion -> destination.
103 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
105 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
106 message typo, sucessors -> successors.
108 2020-02-25 Martin Sebor <msebor@redhat.com>
110 * doc/extend.texi (attribute access): Correct an example.
112 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
114 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
116 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
118 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
119 (VD): Enable for V4BF.
121 (VQ): Enable for V8BF.
124 (VDBL, Vdbl): Add V4BF.
125 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
126 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
127 (bfloat16x8x2_t): Likewise.
128 (bfloat16x4x3_t): Likewise.
129 (bfloat16x8x3_t): Likewise.
130 (bfloat16x4x4_t): Likewise.
131 (bfloat16x8x4_t): Likewise.
132 (vcombine_bf16): New.
133 (vld1_bf16, vld1_bf16_x2): New.
134 (vld1_bf16_x3, vld1_bf16_x4): New.
135 (vld1q_bf16, vld1q_bf16_x2): New.
136 (vld1q_bf16_x3, vld1q_bf16_x4): New.
137 (vld1_lane_bf16): New.
138 (vld1q_lane_bf16): New.
139 (vld1_dup_bf16): New.
140 (vld1q_dup_bf16): New.
143 (vld2_dup_bf16): New.
144 (vld2q_dup_bf16): New.
147 (vld3_dup_bf16): New.
148 (vld3q_dup_bf16): New.
151 (vld4_dup_bf16): New.
152 (vld4q_dup_bf16): New.
153 (vst1_bf16, vst1_bf16_x2): New.
154 (vst1_bf16_x3, vst1_bf16_x4): New.
155 (vst1q_bf16, vst1q_bf16_x2): New.
156 (vst1q_bf16_x3, vst1q_bf16_x4): New.
157 (vst1_lane_bf16): New.
158 (vst1q_lane_bf16): New.
166 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
168 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
169 (VALL_F16): Likewise.
170 (VALLDI_F16): Likewise.
173 (vswap_width_name): Likewise.
174 (VSWAP_WIDTH): Likewise.
178 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
179 (vget_lane_bf16, vgetq_lane_bf16): New.
181 (vdup_n_bf16, vdupq_n_bf16): New.
182 (vdup_lane_bf16, vdup_laneq_bf16): New.
183 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
184 (vduph_lane_bf16, vduph_laneq_bf16): New.
185 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
186 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
187 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
188 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
189 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
190 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
191 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
192 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
193 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
194 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
195 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
196 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
197 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
198 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
199 (vreinterpretq_bf16_p128): New.
200 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
201 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
202 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
203 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
204 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
205 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
206 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
207 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
208 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
209 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
210 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
211 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
212 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
213 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
214 (vreinterpretq_p128_bf16): New.
216 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
218 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
219 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
220 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
221 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
222 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
223 * config/arm/iterators.md (VSF2BF): New attribute.
224 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
225 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
226 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
228 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
230 * config/arm/arm.md (required_for_purecode): New attribute.
231 (enabled): Handle required_for_purecode.
232 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
233 work with -mpure-code.
235 2020-02-25 Jakub Jelinek <jakub@redhat.com>
237 PR rtl-optimization/93908
238 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
241 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
243 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
245 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
247 * doc/install.texi (--enable-checking): Adjust wording.
249 2020-02-25 Richard Biener <rguenther@suse.de>
251 PR tree-optimization/93868
252 * tree-vect-slp.c (slp_copy_subtree): New function.
253 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
254 re-arranging stmts in it.
256 2020-02-25 Jakub Jelinek <jakub@redhat.com>
259 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
260 dummy function and remove it at the end.
263 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
264 paramter -> parameter.
265 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
266 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
268 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
270 * doc/install.texi (--enable-checking): Properly document current
272 (--enable-stage1-checking): Minor clarification about bootstrap.
274 2020-02-24 David Malcolm <dmalcolm@redhat.com>
277 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
278 -fanalyzer-checker=taint is also required.
279 (-fanalyzer-checker=): Note that providing this option enables the
280 given checker, and doing so may be required for checkers that are
283 2020-02-24 David Malcolm <dmalcolm@redhat.com>
285 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
286 significant control flow events; add a "3" which shows all
287 control flow events; the old "3" becomes "4".
289 2020-02-24 Jakub Jelinek <jakub@redhat.com>
291 PR tree-optimization/93582
292 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
293 pd.offset and pd.size to be counted in bits rather than bytes, add
294 support for maxsizei that is not a multiple of BITS_PER_UNIT and
295 handle bitfield stores and loads.
296 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
297 uncomparable quantities - bytes vs. bits. Allow push_partial_def
298 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
299 pd.offset/pd.size to be counted in bits rather than bytes.
300 Formatting fix. Rename shadowed len variable to buflen.
302 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
303 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
306 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
307 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
308 * opts-common.c (parse_options_from_collect_gcc_options): New function.
309 (prepend_xassembler_to_collect_as_options): Likewise.
310 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
311 (prepend_xassembler_to_collect_as_options): Likewise.
312 * lto-opts.c (lto_write_options): Stream assembler options
313 in COLLECT_AS_OPTIONS.
314 * lto-wrapper.c (xassembler_options_error): New static variable.
315 (get_options_from_collect_gcc_options): Move parsing options code to
316 parse_options_from_collect_gcc_options and call it.
317 (merge_and_complain): Validate -Xassembler options.
318 (append_compiler_options): Handle OPT_Xassembler.
319 (run_gcc): Append command line -Xassembler options to
321 * doc/invoke.texi: Add documentation about using Xassembler
324 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
326 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
328 (riscv_rtx_costs): Update cost model for LTGT.
330 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
332 PR rtl-optimization/93564
333 * ira-color.c (struct update_cost_queue_elem): New member start.
334 (queue_update_cost, get_next_update_cost): Add new arg start.
335 (allocnos_conflict_p): New function.
336 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
337 Add checking conflicts with allocnos_conflict_p.
338 (update_costs_from_prefs, restore_costs_from_copies): Adjust
339 update_costs_from_allocno calls.
340 (update_conflict_hard_regno_costs): Add checking conflicts with
341 allocnos_conflict_p. Adjust calls of queue_update_cost and
342 get_next_update_cost.
343 (assign_hard_reg): Adjust calls of queue_update_cost. Add
345 (bucket_allocno_compare_func): Restore previous version.
347 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
349 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
350 double-word size when handling aggregate return values.
351 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
352 that homogeneous SFmode and DFmode aggregates are passed and returned
353 in general registers.
355 2020-02-21 Jakub Jelinek <jakub@redhat.com>
358 * opts.c (print_filtered_help): Translate help before appending
359 messages to it rather than after that.
361 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
363 PR rtl-optimization/PR92989
364 * lra-lives.c (process_bb_lives): Restore the original order
365 of the bb liveness update. Call make_hard_regno_dead for each
366 register clobbered at the start of an EH receiver.
368 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
371 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
372 self-recursively generated.
374 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
377 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
380 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
382 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
383 Document new target supports option.
385 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
387 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
388 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
389 * config/arm/iterators.md (MATMUL): New iterator.
390 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
391 (mmla_sfx): New attribute.
392 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
393 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
394 (UNSPEC_MATMUL_US): New.
396 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
398 * config/arm/arm.md: Prevent scalar shifts from being used when big
401 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
402 Richard Biener <rguenther@suse.de>
404 PR tree-optimization/93586
405 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
406 after mismatched array refs; do not sure type size information to
407 recover from unmatched referneces with !flag_strict_aliasing_p.
409 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
411 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
412 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
413 (scatter_store<mode>): Rename to ...
414 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
415 (scatter<mode>_exec): Delete. Move contents ...
416 (mask_scatter_store<mode>): ... here, and rename that to ...
417 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
418 Remove mode conversion.
419 (mask_gather_load<mode>): Rename to ...
420 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
421 Remove mode conversion.
422 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
424 2020-02-21 Martin Jambor <mjambor@suse.cz>
426 PR tree-optimization/93845
427 * tree-sra.c (verify_sra_access_forest): Only test access size of
430 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
432 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
433 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
434 (addv64di3_exec): Likewise.
435 (subv64di3): Likewise.
436 (subv64di3_exec): Likewise.
437 (addv64di3_zext): Likewise.
438 (addv64di3_zext_exec): Likewise.
439 (addv64di3_zext_dup): Likewise.
440 (addv64di3_zext_dup_exec): Likewise.
441 (addv64di3_zext_dup2): Likewise.
442 (addv64di3_zext_dup2_exec): Likewise.
443 (addv64di3_sext_dup2): Likewise.
444 (addv64di3_sext_dup2_exec): Likewise.
445 (<expander>v64di3): Likewise.
446 (<expander>v64di3_exec): Likewise.
447 (*<reduc_op>_dpp_shr_v64di): Likewise.
448 (*plus_carry_dpp_shr_v64di): Likewise.
449 * config/gcn/gcn.md (adddi3): Likewise.
450 (addptrdi3): Likewise.
451 (<expander>di3): Likewise.
453 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
455 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
457 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
459 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
460 support. Use aarch64_emit_mult instead of emitting multiplication
461 instructions directly.
462 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
463 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
465 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
467 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
468 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
469 instead of emitting multiplication instructions directly.
470 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
471 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
472 (@aarch64_frecps<mode>): New expanders.
474 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
476 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
477 on and produce uint64_ts rather than ints.
478 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
479 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
481 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
483 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
484 an unused xmsk register when handling approximate rsqrt.
486 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
488 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
489 flag_finite_math_only condition.
491 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
494 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
495 to destination operand for shufps alternative.
496 (*vec_extractv2si_1): Ditto.
498 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
501 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
504 2020-02-20 Martin Liska <mliska@suse.cz>
507 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
509 2020-02-20 Martin Liska <mliska@suse.cz>
512 * common/config/avr/avr-common.c: Remote trailing "|".
514 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
516 * collect2.c (maybe_run_lto_and_relink): Fix typo in
519 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
521 PR tree-optimization/93767
522 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
523 access-size bias from the offset calculations for negative strides.
525 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
527 * collect2.c (c_file, o_file): Make const again.
528 (ldout,lderrout, dump_ld_file): Remove.
529 (tool_cleanup): Avoid calling not signal-safe functions.
530 (maybe_run_lto_and_relink): Avoid possible signal handler
531 access to unintialzed memory (lto_o_files).
532 (main): Avoid leaking temp files in $TMPDIR.
533 Initialize c_file/o_file with concat, which avoids exposing
534 uninitialized memory to signal handler, which calls unlink(!).
535 Avoid calling maybe_unlink when the main function returns,
536 since the atexit handler is already doing this.
537 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
539 2020-02-19 Martin Jambor <mjambor@suse.cz>
541 PR tree-optimization/93776
542 * tree-sra.c (create_access): Do not create zero size accesses.
543 (get_access_for_expr): Do not search for zero sized accesses.
545 2020-02-19 Martin Jambor <mjambor@suse.cz>
547 PR tree-optimization/93667
548 * tree-sra.c (scalarizable_type_p): Return false if record fields
549 do not follow wach other.
551 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
553 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
554 rather than fmv.x.s/fmv.s.x.
556 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
558 * config/aarch64/aarch64-simd-builtins.def
559 (intrinsic_vec_smult_lo_): New.
560 (intrinsic_vec_umult_lo_): Likewise.
561 (vec_widen_smult_hi_): Likewise.
562 (vec_widen_umult_hi_): Likewise.
563 * config/aarch64/aarch64-simd.md
564 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
565 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
566 (vmull_high_s16): Likewise.
567 (vmull_high_s32): Likewise.
568 (vmull_high_u8): Likewise.
569 (vmull_high_u16): Likewise.
570 (vmull_high_u32): Likewise.
571 (vmull_s8): Likewise.
572 (vmull_s16): Likewise.
573 (vmull_s32): Likewise.
574 (vmull_u8): Likewise.
575 (vmull_u16): Likewise.
576 (vmull_u32): Likewise.
578 2020-02-18 Martin Liska <mliska@suse.cz>
580 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
581 bootstrap by missing removal of invalid sanity check.
583 2020-02-18 Martin Liska <mliska@suse.cz>
586 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
587 Always compare LHS of gimple_assign.
589 2020-02-18 Martin Liska <mliska@suse.cz>
592 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
593 and return type of functions.
594 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
595 Drop MALLOC attribute for void functions.
596 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
597 malloc_state for a new VOID clone.
599 2020-02-18 Martin Liska <mliska@suse.cz>
602 * common.opt: Add -fprofile-reproducibility.
603 * doc/invoke.texi: Document it.
604 * value-prof.c (dump_histogram_value):
605 Document and support behavior for counters[0]
606 being a negative value.
607 (get_nth_most_common_value): Handle negative
608 counters[0] in respect to flag_profile_reproducible.
610 2020-02-18 Jakub Jelinek <jakub@redhat.com>
613 * cgraph.c (verify_speculative_call): Use speculative_id instead of
614 speculative_uid in messages. Remove trailing whitespace from error
615 message. Use num_speculative_call_targets instead of
616 num_speculative_targets in a message.
617 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
618 edge messages and stmt instead of cal_stmt in reference message.
620 PR tree-optimization/93780
621 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
622 before calling build_vector_type.
623 (execute_update_addresses_taken): Likewise.
626 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
627 typo, functoin -> function.
628 * tree.c (free_lang_data_in_decl): Fix comment typo,
629 functoin -> function.
630 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
632 2020-02-17 David Malcolm <dmalcolm@redhat.com>
634 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
636 (print_option_information): Don't call get_option_url if URLs
639 2020-02-17 Alexandre Oliva <oliva@adacore.com>
641 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
642 handling of register_common-less targets.
644 2020-02-17 Martin Liska <mliska@suse.cz>
647 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
649 2020-02-17 Martin Liska <mliska@suse.cz>
652 * config/rs6000/rs6000.c (rs6000_option_override_internal):
655 2020-02-17 Martin Liska <mliska@suse.cz>
658 * config/rx/elf.opt: Fix typo.
660 2020-02-17 Richard Biener <rguenther@suse.de>
663 * opts-global.c (print_ignored_options): Use inform and
666 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
669 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
671 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
674 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
675 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
677 2020-02-15 Jason Merrill <jason@redhat.com>
679 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
681 2020-02-15 Jakub Jelinek <jakub@redhat.com>
683 PR tree-optimization/93744
684 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
685 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
686 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
687 sure @2 in the first and @1 in the other patterns has no side-effects.
689 2020-02-15 David Malcolm <dmalcolm@redhat.com>
690 Bernd Edlinger <bernd.edlinger@hotmail.de>
694 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
695 * configure.ac (--with-diagnostics-urls): New configuration
696 option, based on --with-diagnostics-color.
697 (DIAGNOSTICS_URLS_DEFAULT): New define.
698 * config.h: Regenerate.
699 * configure: Regenerate.
700 * diagnostic.c (diagnostic_urls_init): Handle -1 for
701 DIAGNOSTICS_URLS_DEFAULT from configure-time
702 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
703 and TERM_URLS environment variable.
704 * diagnostic-url.h (diagnostic_url_format): New enum type.
705 (diagnostic_urls_enabled_p): rename to...
706 (determine_url_format): ... this, and change return type.
707 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
708 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
709 the linux console, and mingw.
710 (diagnostic_urls_enabled_p): rename to...
711 (determine_url_format): ... this, and adjust.
712 * pretty-print.h (pretty_printer::show_urls): rename to...
713 (pretty_printer::url_format): ... this, and change to enum.
714 * pretty-print.c (pretty_printer::pretty_printer,
715 pp_begin_url, pp_end_url, test_urls): Adjust.
716 * doc/install.texi (--with-diagnostics-urls): Document the new
717 configuration option.
718 (--with-diagnostics-color): Document the existing interaction
719 with GCC_COLORS better.
720 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
721 vindex reference. Update description of defaults based on the above.
722 (-fdiagnostics-color): Update description of how -fdiagnostics-color
723 interacts with GCC_COLORS.
725 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
728 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
729 conjunction with TARGET_GNU_TLS in early return.
731 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
733 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
734 the mode is not wider than UNITS_PER_WORD.
736 2020-02-14 Martin Jambor <mjambor@suse.cz>
738 PR tree-optimization/93516
739 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
740 access of the same type as the parent.
741 (propagate_subaccesses_from_lhs): Likewise.
743 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
746 * config/i386/avx512vbmi2intrin.h
747 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
748 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
749 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
750 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
751 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
752 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
753 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
754 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
755 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
756 of lacking a closing parenthesis.
757 * config/i386/avx512vbmi2vlintrin.h
758 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
759 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
760 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
761 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
762 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
763 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
764 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
765 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
766 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
767 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
768 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
769 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
770 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
771 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
772 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
773 _mm_shldi_epi32, _mm_mask_shldi_epi32,
774 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
775 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
777 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
780 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
781 the target function entry.
783 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
785 * common/config/arc/arc-common.c (arc_option_optimization_table):
786 Disable if-conversion step when optimized for size.
788 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
790 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
791 R12-R15 are always in ARCOMPACT16_REGS register class.
792 * config/arc/arc.opt (mq-class): Deprecate.
793 * config/arc/constraint.md ("q"): Remove dependency on mq-class
795 * doc/invoke.texi (mq-class): Update text.
796 * common/config/arc/arc-common.c (arc_option_optimization_table):
799 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
801 * config/arc/arc.c (arc_insn_cost): New function.
802 (TARGET_INSN_COST): Define.
803 * config/arc/arc.md (cost): New attribute.
804 (add_n): Use arc_nonmemory_operand.
805 (ashlsi3_insn): Likewise, also update constraints.
806 (ashrsi3_insn): Likewise.
808 (add_shift): Likewise.
809 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
811 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
813 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
815 (umulsidi_600): Likewise.
817 2020-02-13 Jakub Jelinek <jakub@redhat.com>
820 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
821 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
822 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
823 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
824 pass __A to the builtin followed by __W instead of __A followed by
826 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
827 _mm512_mask_popcnt_epi64): Likewise.
828 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
829 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
830 _mm256_mask_popcnt_epi64): Likewise.
832 PR tree-optimization/93582
833 * fold-const.h (shift_bytes_in_array_left,
834 shift_bytes_in_array_right): Declare.
835 * fold-const.c (shift_bytes_in_array_left,
836 shift_bytes_in_array_right): New function, moved from
837 gimple-ssa-store-merging.c, no longer static.
838 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
839 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
840 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
841 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
842 shift_bytes_in_array.
843 (verify_shift_bytes_in_array): Rename to ...
844 (verify_shift_bytes_in_array_left): ... this. Use
845 shift_bytes_in_array_left instead of shift_bytes_in_array.
846 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
847 instead of verify_shift_bytes_in_array.
848 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
849 / native_interpret_expr where the store covers all needed bits,
850 punt on PDP-endian, otherwise allow all involved offsets and sizes
851 not to be byte-aligned.
854 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
855 use const_0_to_255_operand predicate instead of immediate_operand.
856 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
857 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
858 vgf2p8affineinvqb_<mode><mask_name>,
859 vgf2p8affineqb_<mode><mask_name>): Drop mode from
860 const_0_to_255_operand predicated operands.
862 2020-02-12 Jeff Law <law@redhat.com>
864 * config/h8300/h8300.md (comparison shortening peepholes): Use
865 a mode iterator to merge the HImode and SImode peepholes.
867 2020-02-12 Jakub Jelinek <jakub@redhat.com>
870 * real.c (is_even): Make static. Function comment fix.
871 (is_halfway_below): Make static, don't assert R is not inf/nan,
872 instead return false for those. Small formatting fixes.
874 2020-02-12 Martin Sebor <msebor@redhat.com>
877 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
878 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
879 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
880 (strlen_check_and_optimize_call): Adjust callee name.
882 2020-02-12 Jeff Law <law@redhat.com>
884 * config/h8300/h8300.md (comparison shortening peepholes): Drop
885 (and (xor)) variant. Combine other two into single peephole.
887 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
889 PR rtl-optimization/93565
890 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
892 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
894 * config/aarch64/aarch64-simd.md
895 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
896 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
897 generating separate ADDV and zero_extend patterns.
898 * config/aarch64/iterators.md (VDQV_E): New iterator.
900 2020-02-12 Jeff Law <law@redhat.com>
902 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
903 expanders, splits, etc.
904 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
905 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
906 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
907 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
908 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
909 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
911 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
913 2020-02-12 Jakub Jelinek <jakub@redhat.com>
916 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
917 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
918 TARGET_AVX512DQ from condition.
919 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
920 instead of <mask_mode512bit_condition> in condition. If
921 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
923 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
926 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
929 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
931 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
933 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
934 where strlen is more legible.
935 (rs6000_builtin_vectorized_libmass): Ditto.
936 (rs6000_print_options_internal): Ditto.
938 2020-02-11 Martin Sebor <msebor@redhat.com>
940 PR tree-optimization/93683
941 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
943 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
945 * config/rs6000/predicates.md (cint34_operand): Rename the
946 -mprefixed-addr option to be -mprefixed.
947 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
948 the -mprefixed-addr option to be -mprefixed.
949 (OTHER_FUTURE_MASKS): Likewise.
950 (POWERPC_MASKS): Likewise.
951 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
952 the -mprefixed-addr option to be -mprefixed. Change error
953 messages to refer to -mprefixed.
954 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
956 (rs6000_legitimate_offset_address_p): Likewise.
957 (rs6000_mode_dependent_address): Likewise.
958 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
959 "-mprefixed" for target attributes and pragmas.
960 (address_to_insn_form): Rename the -mprefixed-addr option to be
962 (rs6000_adjust_insn_length): Likewise.
963 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
964 -mprefixed-addr option to be -mprefixed.
965 (ASM_OUTPUT_OPCODE): Likewise.
966 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
967 -mprefixed-addr option to be -mprefixed.
968 * config/rs6000/rs6000.opt (-mprefixed): Rename the
969 -mprefixed-addr option to be prefixed. Change the option from
970 being undocumented to being documented.
971 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
972 -mprefixed option. Update the -mpcrel documentation to mention
975 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
977 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
978 including FIRST_PSEUDO_REGISTER - 1.
979 * ira-color.c (print_hard_reg_set): Ditto.
981 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
983 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
984 (USTERNOP_QUALIFIERS): New define.
985 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
986 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
987 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
988 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
989 * config/arm/arm_neon.h (vusdot_s32): New.
990 (vusdot_lane_s32): New.
991 (vusdotq_lane_s32): New.
992 (vsudot_lane_s32): New.
993 (vsudotq_lane_s32): New.
994 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
995 * config/arm/iterators.md (DOTPROD_I8MM): New.
996 (sup, opsuffix): Add <us/su>.
997 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
998 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1000 2020-02-11 Richard Biener <rguenther@suse.de>
1002 PR tree-optimization/93661
1003 PR tree-optimization/93662
1004 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1006 * tree-sra.c (get_access_for_expr): Likewise.
1008 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1011 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1012 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1013 Change condition from TARGET_AVX2 to TARGET_AVX.
1015 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1018 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1019 argument of strncmp.
1021 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1023 Try to generate zero-based comparisons.
1024 * config/cris/cris.c (cris_reduce_compare): New function.
1025 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1026 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1027 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1029 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1032 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1033 in Thumb state and also as a destination in Arm state. Add T16
1036 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1038 * md.texi (Define Subst): Match closing paren in example.
1040 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1044 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1045 arguments of strncmp.
1047 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1050 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1051 but different source value.
1052 (adjust_callers_for_value_intersection): New function.
1053 (gather_edges_for_value): Adjust order of callers to let a
1054 non-self-recursive caller be the first element.
1055 (self_recursive_pass_through_p): Add a new parameter "simple", and
1056 check generalized self-recursive pass-through jump function.
1057 (self_recursive_agg_pass_through_p): Likewise.
1058 (find_more_scalar_values_for_callers_subset): Compute value from
1059 pass-through jump function for self-recursive.
1060 (intersect_with_plats): Cleanup previous implementation code for value
1061 itersection with self-recursive call edge.
1062 (intersect_with_agg_replacements): Likewise.
1063 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1064 function for self-recursive call edge. Cleanup previous implementation
1065 code for value intersection with self-recursive call edge.
1066 (decide_whether_version_node): Remove dead callers and adjust order
1067 to let a non-self-recursive caller be the first element.
1069 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1071 * recog.c: Move pass_split_before_sched2 code in front of
1072 pass_split_before_regstack.
1073 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1074 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1075 (rest_of_handle_split_before_sched2): Remove.
1076 (pass_split_before_sched2::execute): Unconditionally call
1078 (enable_split_before_sched2): New function.
1079 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1080 (pass_split_before_regstack::gate): Ditto.
1081 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1082 Update name check for renamed split4 pass.
1083 * config/sh/sh.c (register_sh_passes): Update pass insertion
1084 point for renamed split4 pass.
1086 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1088 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1089 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1090 copying them around between host and target.
1092 2020-02-08 Andrew Pinski <apinski@marvell.com>
1095 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1096 STRICT_ALIGNMENT also.
1098 2020-02-08 Jim Wilson <jimw@sifive.com>
1101 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1103 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1104 Jakub Jelinek <jakub@redhat.com>
1107 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1108 xmm16-xmm31 call-used even in 64-bit ms-abi.
1110 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1112 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1113 (simd_ummla, simd_usmmla): Likewise.
1114 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1115 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1116 (vusmmlaq_s32): New.
1118 2020-02-07 Richard Biener <rguenther@suse.de>
1121 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1122 skipping unreachable regions.
1123 (optimize_inline_calls): Skip folding stmts when we didn't
1126 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1129 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1130 Don't return aggregates with only SFmode and DFmode in SSE
1132 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1134 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1137 * config/rs6000/rs6000-logue.c
1138 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1139 if it fails, move rs into end_addr and retry. Add
1140 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1141 the insn pattern doesn't describe well what exactly happens to
1145 * config/i386/predicates.md (avx_identity_operand): Remove.
1146 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1147 (avx_<castmode><avxsizesuffix>_<castmode>,
1148 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1149 a VEC_CONCAT of the operand and UNSPEC_CAST.
1150 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1151 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1155 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1156 recog_data.insn if distance_non_agu_define changed it.
1158 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1161 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1162 we only had X-FORM (reg+reg) addressing for vectors. Also before
1163 ISA 3.0, we only had X-FORM addressing for scalars in the
1164 traditional Altivec registers.
1166 2020-02-06 <zhongyunde@huawei.com>
1167 Vladimir Makarov <vmakarov@redhat.com>
1169 PR rtl-optimization/93561
1170 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1171 hard register range.
1173 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1175 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1178 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1180 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1181 where the low and the high 32 bits are equal to each other specially,
1182 with an rldimi instruction.
1184 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1186 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1188 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1190 * config/arm/arm-tables.opt: Regenerate.
1192 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1195 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1196 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1197 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1199 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1201 PR rtl-optimization/87763
1202 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1204 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1206 * config/aarch64/aarch64-simd-builtins.def
1207 (bfmlaq): New built-in function.
1208 (bfmlalb): New built-in function.
1209 (bfmlalt): New built-in function.
1210 (bfmlalb_lane): New built-in function.
1211 (bfmlalt_lane): New built-in function.
1212 * config/aarch64/aarch64-simd.md
1213 (aarch64_bfmmlaqv4sf): New pattern.
1214 (aarch64_bfmlal<bt>v4sf): New pattern.
1215 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1216 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1217 (vbfmlalbq_f32): New intrinsic.
1218 (vbfmlaltq_f32): New intrinsic.
1219 (vbfmlalbq_lane_f32): New intrinsic.
1220 (vbfmlaltq_lane_f32): New intrinsic.
1221 (vbfmlalbq_laneq_f32): New intrinsic.
1222 (vbfmlaltq_laneq_f32): New intrinsic.
1223 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1224 (bt): New int attribute.
1226 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1228 * config/i386/i386.md (*pushtf): Emit "#" instead of
1229 calling gcc_unreachable in insn output.
1232 (*pushsf_rex64): Ditto for alternatives other than 1.
1233 (*pushsf): Ditto for alternatives other than 1.
1235 2020-02-06 Martin Liska <mliska@suse.cz>
1237 PR gcov-profile/91971
1238 PR gcov-profile/93466
1239 * coverage.c (coverage_init): Revert mangling of
1240 path into filename. It can lead to huge filename length.
1241 Creation of subfolders seem more natural.
1243 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1246 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1247 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1248 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1250 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1253 * config/i386/predicates.md (avx_identity_operand): New predicate.
1254 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1255 define_insn_and_split.
1258 * omp-low.c (use_pointer_for_field): For nested constructs, also
1259 look for map clauses on target construct.
1260 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1261 taskreg_nesting_level.
1264 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1265 shared clause, call omp_notice_variable on outer context if any.
1267 2020-02-05 Jason Merrill <jason@redhat.com>
1270 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1271 non-zero address even if weak and not yet defined.
1273 2020-02-05 Martin Sebor <msebor@redhat.com>
1275 PR tree-optimization/92765
1276 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1277 * tree-ssa-strlen.c (compute_string_length): Remove.
1278 (determine_min_objsize): Remove.
1279 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1280 Avoid using type size as the upper bound on string length.
1281 (handle_builtin_string_cmp): Add an argument. Adjust.
1282 (strlen_check_and_optimize_call): Pass additional argument to
1283 handle_builtin_string_cmp.
1285 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1287 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1288 (*pushdi2_rex64 peephole2): Unconditionally split after
1290 (*ashl<mode>3_doubleword): Ditto.
1291 (*<shift_insn><mode>3_doubleword): Ditto.
1293 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1296 * config/rs6000/rs6000.c (get_vector_offset): Fix
1298 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1300 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1302 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1305 (Special Functions for Debugging the Analyzer): Update description
1306 of __analyzer_dump_exploded_nodes.
1308 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1311 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1312 include sets and not clobbers in the vzeroupper pattern.
1313 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1314 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1315 (*avx_vzeroupper_1): New define_insn_and_split.
1318 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1319 don't run when !optimize.
1320 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1323 2020-02-05 Richard Biener <rguenther@suse.de>
1326 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1327 checks before matching calls.
1329 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1331 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1332 function comment typo.
1335 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1336 simd_clone_create failed when i == 0, adjust clone->nargs by
1339 2020-02-05 Martin Liska <mliska@suse.cz>
1342 * doc/invoke.texi: Document that one should
1343 not combine ASLR and -fpch.
1345 2020-02-04 Richard Biener <rguenther@suse.de>
1347 PR tree-optimization/93538
1348 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1350 2020-02-04 Richard Biener <rguenther@suse.de>
1352 PR tree-optimization/91123
1353 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1354 (vn_walk_cb_data::last_vuse): New member.
1355 (vn_walk_cb_data::saved_operands): Likewsie.
1356 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1357 (vn_walk_cb_data::push_partial_def): Use finish.
1358 (vn_reference_lookup_2): Update last_vuse and use finish if
1359 we've saved operands.
1360 (vn_reference_lookup_3): Use finish and update calls to
1361 push_partial_defs everywhere. When translating through
1362 memcpy or aggregate copies save off operands and alias-set.
1363 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1364 operation for redundant store removal.
1366 2020-02-04 Richard Biener <rguenther@suse.de>
1368 PR tree-optimization/92819
1369 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1370 generating more stmts than before.
1372 2020-02-04 Martin Liska <mliska@suse.cz>
1374 * config/arm/arm.c (arm_gen_far_branch): Move the function
1375 outside of selftests.
1377 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1379 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1380 function to adjust PC-relative vector addresses.
1381 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1382 handle vectors with PC-relative addresses.
1384 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1386 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1388 (hard_reg_and_mode_to_addr_mask): Delete.
1389 (rs6000_adjust_vec_address): If the original vector address
1390 was REG+REG or REG+OFFSET and the element is not zero, do the add
1391 of the elements in the original address before adding the offset
1392 for the vector element. Use address_to_insn_form to validate the
1393 address using the register being loaded, rather than guessing
1394 whether the address is a DS-FORM or DQ-FORM address.
1396 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1398 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1399 to calculate the offset in memory from the start of a vector of a
1400 particular element. Add code to keep the element number in
1401 bounds if the element number is variable.
1402 (rs6000_adjust_vec_address): Move calculation of offset of the
1403 vector element to get_vector_offset.
1404 (rs6000_split_vec_extract_var): Do not do the initial AND of
1405 element here, move the code to get_vector_offset.
1407 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1409 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1412 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1414 * config/rs6000/constraints.md: Improve documentation.
1416 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1419 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1420 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1422 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1424 * config.gcc: Remove "carrizo" support.
1425 * config/gcn/gcn-opts.h (processor_type): Likewise.
1426 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1427 * config/gcn/gcn.opt (gpu_type): Likewise.
1428 * config/gcn/t-omp-device: Likewise.
1430 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1433 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1434 * config/arm/arm.c (arm_gen_far_branch): New function
1436 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1438 2020-02-03 Julian Brown <julian@codesourcery.com>
1439 Tobias Burnus <tobias@codesourcery.com>
1441 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1443 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1446 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1447 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1450 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1452 PR rtl-optimization/91333
1453 * ira-color.c (struct allocno_color_data): Add member
1455 (init_allocno_threads): Set the member up.
1456 (bucket_allocno_compare_func): Add compare hard reg
1459 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1461 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1463 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1464 * config.in: Regenerated.
1465 * configure: Regenerated.
1466 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1467 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1468 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1470 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1472 * configure: Regenerate.
1474 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1476 PR rtl-optimization/91333
1477 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1478 reg preferences comparison up.
1480 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1482 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1483 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1484 aarch64-sve-builtins-base.h.
1485 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1486 aarch64-sve-builtins-base.cc.
1487 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1488 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1490 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1491 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1492 (svcvtnt): New functions.
1493 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1494 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1495 (svcvtnt): New functions.
1496 (svcvt): Add a form that converts f32 to bf16.
1497 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1498 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1500 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1501 Treat B as bfloat16_t.
1502 (ternary_bfloat_lane_base): New class.
1503 (ternary_bfloat_def): Likewise.
1504 (ternary_bfloat): New shape.
1505 (ternary_bfloat_lane_def): New class.
1506 (ternary_bfloat_lane): New shape.
1507 (ternary_bfloat_lanex2_def): New class.
1508 (ternary_bfloat_lanex2): New shape.
1509 (ternary_bfloat_opt_n_def): New class.
1510 (ternary_bfloat_opt_n): New shape.
1511 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1512 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1513 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1514 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1515 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1516 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1517 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1518 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1519 the pattern off the narrow mode instead of the wider one.
1520 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1521 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1522 (sve_fp_op): Handle them.
1523 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1524 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1526 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1528 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1529 * config/aarch64/aarch64-modes.def (BF): Move definition before
1530 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1531 (SVE_MODES): Handle BF modes.
1532 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1534 (aarch64_full_sve_mode): Likewise.
1535 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1537 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1538 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1539 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1540 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1542 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1544 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1545 (TYPES_all_data): Add bf16.
1546 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1547 (register_tuple_type): Increase buffer size.
1548 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1549 (bf16): New type suffix.
1550 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1551 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1552 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1553 Change type from all_data to all_arith.
1554 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1557 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1558 Matthew Malcomson <matthew.malcomson@arm.com>
1559 Richard Sandiford <richard.sandiford@arm.com>
1561 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1562 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1563 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1564 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1565 __ARM_FEATURE_MATMUL_FP64.
1566 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1567 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1568 be disabled at the same time.
1569 (f32mm): New extension.
1570 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1571 (AARCH64_FL_F64MM): Bump to the next bit up.
1572 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1573 (TARGET_SVE_F64MM): New macros.
1574 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1575 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1576 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1577 (UNSPEC_ZIP2Q): New unspeccs.
1578 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1579 (optab, sur, perm_insn): Handle the new unspecs.
1580 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1581 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1582 TARGET_SVE_F64MM instead of separate tests.
1583 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1584 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1585 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1586 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1587 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1588 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1589 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1590 (TYPES_s_signed): New macro.
1591 (TYPES_s_integer): Use it.
1592 (TYPES_d_float): New macro.
1593 (TYPES_d_data): Use it.
1594 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1595 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1596 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1597 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1598 (svmmla): New shape.
1599 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1600 template parameters.
1601 (ternary_resize2_lane_base): Likewise.
1602 (ternary_resize2_base): New class.
1603 (ternary_qq_lane_base): Likewise.
1604 (ternary_intq_uintq_lane_def): Likewise.
1605 (ternary_intq_uintq_lane): New shape.
1606 (ternary_intq_uintq_opt_n_def): New class
1607 (ternary_intq_uintq_opt_n): New shape.
1608 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1609 (ternary_uintq_intq_def): New class.
1610 (ternary_uintq_intq): New shape.
1611 (ternary_uintq_intq_lane_def): New class.
1612 (ternary_uintq_intq_lane): New shape.
1613 (ternary_uintq_intq_opt_n_def): New class.
1614 (ternary_uintq_intq_opt_n): New shape.
1615 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1616 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1617 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1618 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1620 (svdotprod_lane_impl): ...this new class.
1621 (svmmla_impl, svusdot_impl): New classes.
1622 (svdot_lane): Update to use svdotprod_lane_impl.
1623 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1624 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1626 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1627 function, with no types defined.
1628 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1629 AARCH64_FL_I8MM functions.
1630 (svmmla): New AARCH64_FL_F32MM function.
1631 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1632 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1633 AARCH64_FL_F64MM function.
1634 (REQUIRED_EXTENSIONS):
1636 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1638 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1641 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1643 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1644 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1645 (*movti_internal): Do not check for
1646 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1647 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1648 just after check for TARGET_AVX.
1649 (*movdf_internal): Ditto.
1650 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1651 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1652 * config/i386/sse.md (mov<mode>_internal): Only check
1653 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1654 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1655 (<sse>_andnot<mode>3<mask_name>): Move check for
1656 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1657 (<code><mode>3<mask_name>): Ditto.
1658 (*andnot<mode>3): Ditto.
1659 (*andnottf3): Ditto.
1660 (*<code><mode>3): Ditto.
1661 (*<code>tf3): Ditto.
1662 (*andnot<VI:mode>3): Remove
1663 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1664 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1665 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1666 (sse4_1_blendv<ssemodesuffix>): Ditto.
1667 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1668 Explain that tune applies to 128bit instructions only.
1670 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1672 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1673 to definition of hsa_kernel_description. Parse assembly to find SGPR
1674 and VGPR count of kernel and store in hsa_kernel_description.
1676 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1678 PR rtl-optimization/91838
1679 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1680 to truncate if allowed or reject combination.
1682 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1684 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1685 (find_inv_vars_cb): Likewise.
1687 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1689 * calls.c (special_function_p): Split out the check for DECL_NAME
1690 being non-NULL and fndecl being extern at file scope into a
1691 new maybe_special_function_p and call it. Drop check for fndecl
1692 being non-NULL that was after a usage of DECL_NAME (fndecl).
1693 * tree.h (maybe_special_function_p): New inline function.
1695 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1697 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1698 (mask_gather_load<mode>): ... here, and zero-initialize the
1700 (maskload<mode>di): Zero-initialize the destination.
1703 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1706 * doc/analyzer.texi (Limitations): Note that constraints on
1707 floating-point values are currently ignored.
1709 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1712 * symtab.c (symtab_node::noninterposable_alias): If localalias
1713 already exists, but is not usable, append numbers after it until
1714 a unique name is found. Formatting fix.
1717 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1720 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1722 * config/gcn/gcn.c (print_operand): Handle LTGT.
1723 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1725 2020-01-30 Richard Biener <rguenther@suse.de>
1727 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1728 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1730 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1732 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1733 without a DECL in .data.rel.ro.local.
1735 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1738 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1742 * config/i386/sse.md
1743 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1744 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1745 any_extend code iterator instead of always zero_extend.
1746 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1747 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1748 Use any_extend code iterator instead of always zero_extend.
1749 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1750 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1751 Use any_extend code iterator instead of always zero_extend.
1752 (*sse2_pmovmskb_ext): New define_insn.
1753 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1756 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1757 (*popcountsi2_zext_falsedep): New define_insn.
1759 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1761 * config.in: Regenerated.
1762 * configure: Regenerated.
1764 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1767 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1768 LLVM's assembler changed the default in version 9.
1770 2020-01-24 Jeff Law <law@redhat.com>
1772 PR tree-optimization/89689
1773 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1775 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1779 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1781 PR rtl-optimization/87763
1782 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1783 simplification to handle subregs as well as bare regs.
1784 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1786 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1789 * ira.c (ira): Revert use of simplified LRA algorithm.
1791 2020-01-29 Martin Jambor <mjambor@suse.cz>
1793 PR tree-optimization/92706
1794 * tree-sra.c (struct access): Fields first_link, last_link,
1795 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1796 next_rhs_queued and grp_rhs_queued respectively, new fields
1797 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1798 (struct assign_link): Field next renamed to next_rhs, new field
1799 next_lhs. Updated comment.
1800 (work_queue_head): Renamed to rhs_work_queue_head.
1801 (lhs_work_queue_head): New variable.
1802 (add_link_to_lhs): New function.
1803 (relink_to_new_repr): Also relink LHS lists.
1804 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1805 (add_access_to_lhs_work_queue): New function.
1806 (pop_access_from_work_queue): Renamed to
1807 pop_access_from_rhs_work_queue.
1808 (pop_access_from_lhs_work_queue): New function.
1809 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1811 (child_would_conflict_in_lacc): Renamed to
1812 child_would_conflict_in_acc. Adjusted parameter names.
1813 (create_artificial_child_access): New parameter set_grp_read, use it.
1814 (subtree_mark_written_and_enqueue): Renamed to
1815 subtree_mark_written_and_rhs_enqueue.
1816 (propagate_subaccesses_across_link): Renamed to
1817 propagate_subaccesses_from_rhs.
1818 (propagate_subaccesses_from_lhs): New function.
1819 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1822 2020-01-29 Martin Jambor <mjambor@suse.cz>
1824 PR tree-optimization/92706
1825 * tree-sra.c (struct access): Adjust comment of
1826 grp_total_scalarization.
1827 (find_access_in_subtree): Look for single children spanning an entire
1829 (scalarizable_type_p): Allow register accesses, adjust callers.
1830 (completely_scalarize): Remove function.
1831 (scalarize_elem): Likewise.
1832 (create_total_scalarization_access): Likewise.
1833 (sort_and_splice_var_accesses): Do not track total scalarization
1835 (analyze_access_subtree): New parameter totally, adjust to new meaning
1836 of grp_total_scalarization.
1837 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1838 (can_totally_scalarize_forest_p): New function.
1839 (create_total_scalarization_access): Likewise.
1840 (create_total_access_and_reshape): Likewise.
1841 (total_should_skip_creating_access): Likewise.
1842 (totally_scalarize_subtree): Likewise.
1843 (analyze_all_variable_accesses): Perform total scalarization after
1844 subaccess propagation using the new functions above.
1845 (initialize_constant_pool_replacements): Output initializers by
1846 traversing the access tree.
1848 2020-01-29 Martin Jambor <mjambor@suse.cz>
1850 * tree-sra.c (verify_sra_access_forest): New function.
1851 (verify_all_sra_access_forests): Likewise.
1852 (create_artificial_child_access): Set parent.
1853 (analyze_all_variable_accesses): Call the verifier.
1855 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1857 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1858 if called on indirect edge.
1859 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1860 speculative call if needed.
1862 2020-01-29 Richard Biener <rguenther@suse.de>
1864 PR tree-optimization/93428
1865 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1866 permutation when the load node is created.
1867 (vect_analyze_slp_instance): Re-use it here.
1869 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1871 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1873 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1875 PR rtl-optimization/93272
1876 * ira-lives.c (process_out_of_region_eh_regs): New function.
1877 (process_bb_node_lives): Call it.
1879 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1881 * coverage.c (read_counts_file): Make error message lowercase.
1883 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1885 * profile-count.c (profile_quality_display_names): Fix ordering.
1887 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1890 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1891 hash only when edge is first within the sequence.
1892 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1893 (symbol_table::create_edge): Do not set target_prob.
1894 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1896 (cgraph_edge::make_speculative): Drop target_prob parameter.
1897 (cgraph_edge::speculative_call_info): Remove.
1898 (cgraph_edge::first_speculative_call_target): New member function.
1899 (update_call_stmt_hash_for_removing_direct_edge): New function.
1900 (cgraph_edge::resolve_speculation): Rewrite to new API.
1901 (cgraph_edge::speculative_call_for_target): New member function.
1902 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1903 multiple speculation targets.
1904 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1906 (verify_speculative_call): Verify that targets form an interval.
1907 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1908 (cgraph_edge::first_speculative_call_target): New member function.
1909 (cgraph_edge::next_speculative_call_target): New member function.
1910 (cgraph_edge::speculative_call_target_ref): New member function.
1911 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1912 (cgraph_edge): Remove target_prob.
1913 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1914 Fix handling of speculative calls.
1915 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1916 * ipa-fnsummary.c (analyze_function_body): Likewise.
1917 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1918 * ipa-profile.c (dump_histogram): Fix formating.
1919 (ipa_profile_generate_summary): Watch for overflows.
1920 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1921 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1922 (update_indirect_edges_after_inlining): Update to new API.
1923 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1925 * profile-count.h: (profile_probability::adjusted): New.
1926 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1927 updating of profile.
1928 * value-prof.c (gimple_ic_transform): Rename to ...
1929 (dump_ic_profile): ... this one; update dumping.
1930 (stream_in_histogram_value): Fix formating.
1931 (gimple_value_profile_transformations): Update.
1933 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1936 * config/i386/i386.md (*movoi_internal_avx): Remove
1937 TARGET_SSE_TYPELESS_STORES check.
1938 (*movti_internal): Prefer TARGET_AVX over
1939 TARGET_SSE_TYPELESS_STORES.
1940 (*movtf_internal): Likewise.
1941 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1942 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1943 from TARGET_SSE_TYPELESS_STORES.
1945 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1947 * diagnostic-core.h (warning_at): Rename overload to...
1948 (warning_meta): ...this.
1949 (emit_diagnostic_valist): Delete decl of overload taking
1950 diagnostic_metadata.
1951 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1952 (warning_at): Rename overload taking diagnostic_metadata to...
1953 (warning_meta): ...this.
1955 2020-01-28 Richard Biener <rguenther@suse.de>
1957 PR tree-optimization/93439
1958 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1959 * tree-cfg.c (move_sese_region_to_fn): ... here.
1960 (verify_types_in_gimple_reference): Verify used cliques are
1963 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1966 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1967 argument of a pointer to struct gcc_options and pass it to
1968 parse_mtune_ctrl_str.
1969 (ix86_function_specific_restore): Pass opts to
1970 set_ix86_tune_features.
1971 (ix86_option_override_internal): Likewise.
1972 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1973 gcc_options and use it for x_ix86_tune_ctrl_string.
1975 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1977 PR rtl-optimization/87763
1978 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1979 simplification to handle subregs as well as bare regs.
1980 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1982 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1984 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1985 for reduction chains that (now) include a call.
1987 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1989 PR tree-optimization/92822
1990 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1991 out the don't-care elements of a vector whose significant elements
1992 are duplicates, make the don't-care elements duplicates too.
1994 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1996 PR tree-optimization/93434
1997 * tree-predcom.c (split_data_refs_to_components): Record which
1998 components have had aliasing loads removed. Prevent store-store
1999 commoning for all such components.
2001 2020-01-28 Jakub Jelinek <jakub@redhat.com>
2004 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2005 -1 or is_vshift is true, use new_vector with number of elts npatterns
2006 rather than new_unary_operation.
2008 PR tree-optimization/93454
2009 * gimple-fold.c (fold_array_ctor_reference): Perform
2010 elt_size.to_uhwi () just once, instead of calling it in every
2011 iteration. Punt if that value is above size of the temporary
2012 buffer. Decrease third native_encode_expr argument when
2013 bufoff + elt_sz is above size of buf.
2015 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2017 * config/mips/mips.c (mips_declare_object_name)
2018 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2020 2020-01-27 Martin Liska <mliska@suse.cz>
2022 PR gcov-profile/93403
2023 * tree-profile.c (gimple_init_gcov_profiler): Generate
2024 both __gcov_indirect_call_profiler_v4 and
2025 __gcov_indirect_call_profiler_v4_atomic.
2027 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2030 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2032 (@aarch64_split_simd_mov<mode>): Use it.
2033 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2034 Leave the vec_extract patterns to handle 2-element vectors.
2035 (aarch64_simd_mov_from_<mode>high): Likewise.
2036 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2037 (vec_extractv2dfv1df): Likewise.
2039 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2041 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2042 jump conditions for *compare_condjump<GPI:mode>.
2044 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2047 * digraph.cc (test_edge::test_edge): Specify template for base
2050 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2052 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2054 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2056 * config/arc/arc-protos.h (gen_mlo): Remove.
2057 (gen_mhi): Likewise.
2058 * config/arc/arc.c (AUX_MULHI): Define.
2059 (arc_must_save_reister): Special handling for r58/59.
2060 (arc_compute_frame_size): Consider mlo/mhi registers.
2061 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2063 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2064 mlo/mhi name selection.
2065 (arc_restore_callee_saves): Don't early restore blink when ISR.
2066 (arc_expand_prologue): Add mlo/mhi saving.
2067 (arc_expand_epilogue): Add mlo/mhi restoring.
2070 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2071 numbering when MUL64 option is used.
2072 (DWARF2_FRAME_REG_OUT): Define.
2073 * config/arc/arc.md (arc600_stall): New pattern.
2074 (VUNSPEC_ARC_ARC600_STALL): Define.
2075 (mulsi64): Use correct mlo/mhi registers.
2076 (mulsi_600): Clean it up.
2077 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2079 (mhi_operand): Likewise.
2081 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2082 Petro Karashchenko <petro.karashchenko@ring.com>
2084 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2085 attributes if needed.
2086 (prepare_move_operands): Generate special unspec instruction for
2088 (arc_isuncached_mem_p): Propagate uncached attribute to each
2090 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2091 (VUNSPEC_ARC_STDI): Likewise.
2092 (ALLI): New mode iterator.
2093 (mALLI): New mode attribute.
2094 (lddi): New instruction pattern.
2096 (stdidi_split): Split instruction for architectures which are not
2097 supporting ll64 option.
2098 (lddidi_split): Likewise.
2100 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2102 PR rtl-optimization/92989
2103 * lra-lives.c (process_bb_lives): Update the live-in set before
2104 processing additional clobbers.
2106 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2108 PR rtl-optimization/93170
2109 * cselib.c (cselib_invalidate_regno_val): New function, split out
2111 (cselib_invalidate_regno): ...here.
2112 (cselib_invalidated_by_call_p): New function.
2113 (cselib_process_insn): Iterate over all the hard-register entries in
2114 REG_VALUES and invalidate any that cross call-clobbered registers.
2116 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2118 * dojump.c (split_comparison): Use HONOR_NANS rather than
2119 HONOR_SNANS when splitting LTGT.
2121 2020-01-27 Martin Liska <mliska@suse.cz>
2124 * opts.c (print_filtered_help): Exclude language-specific
2125 options from --help=common unless enabled in all FEs.
2127 2020-01-27 Martin Liska <mliska@suse.cz>
2129 * opts.c (print_help): Exclude params from
2130 all except --help=param.
2132 2020-01-27 Martin Liska <mliska@suse.cz>
2135 * config/i386/i386-features.c (make_resolver_func):
2136 Align the code with ppc64 target implementation.
2137 Do not generate a unique name for resolver function.
2139 2020-01-27 Richard Biener <rguenther@suse.de>
2141 PR tree-optimization/93397
2142 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2143 converted reduction chain SLP graph adjustment.
2145 2020-01-26 Marek Polacek <polacek@redhat.com>
2148 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2151 2020-01-26 Jason Merrill <jason@redhat.com>
2154 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2157 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2159 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2160 (rx_setmem): Likewise.
2162 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2165 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2166 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2167 drop <di> from constraint of last operand.
2170 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2171 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2172 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2174 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2177 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2180 2020-01-24 Jeff Law <law@redhat.com>
2182 PR tree-optimization/92788
2183 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2186 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2189 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2190 *avx_vperm_broadcast_<mode>,
2191 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2192 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2193 Move before avx2_perm<mode>/avx512f_perm<mode>.
2196 * simplify-rtx.c (simplify_const_unary_operation,
2197 simplify_const_binary_operation): Punt for mode precision above
2198 MAX_BITSIZE_MODE_ANY_INT.
2200 2020-01-24 Andrew Pinski <apinski@marvell.com>
2202 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2205 2020-01-24 Jeff Law <law@redhat.com>
2208 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2209 for REGs. Call output_operand_lossage to get more reasonable
2212 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2214 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2215 gcn_fp_compare_operator.
2216 (vec_cmpu<mode>di): Use gcn_compare_operator.
2217 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2218 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2219 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2220 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2221 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2222 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2223 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2224 gcn_fp_compare_operator.
2225 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2226 gcn_fp_compare_operator.
2227 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2228 gcn_fp_compare_operator.
2229 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2230 gcn_fp_compare_operator.
2232 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2234 * doc/install.texi (Cross-Compiler-Specific Options): Document
2235 `--with-toolexeclibdir' option.
2237 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2239 * target.def (flags_regnum): Also mention effect on delay slot filling.
2240 * doc/tm.texi: Regenerate.
2242 2020-01-23 Jeff Law <law@redhat.com>
2244 PR translation/90162
2245 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2247 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2250 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2253 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2255 PR rtl-optimization/93402
2256 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2259 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2261 * config.in: Regenerated.
2262 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2263 for TARGET_LIBC_GNUSTACK.
2264 * configure: Regenerated.
2265 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2266 found to be 2.31 or greater.
2268 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2270 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2272 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2273 (mips_asm_file_end): New function. Delegate to
2274 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2275 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2277 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2280 * config/i386/i386-modes.def (POImode): New mode.
2281 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2282 * config/i386/i386.md (DPWI): New mode attribute.
2283 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2285 (QPWI): ... this. Use POI instead of OI for TImode.
2286 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2287 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2290 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2293 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2295 (speculation_tracker_rev): New pattern.
2296 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2297 Use speculation_tracker_rev to track the inverse condition.
2299 2020-01-23 Richard Biener <rguenther@suse.de>
2301 PR tree-optimization/93381
2302 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2303 alias-set of the def as argument and record the first one.
2304 (vn_walk_cb_data::first_set): New member.
2305 (vn_reference_lookup_3): Pass the alias-set of the current def
2306 to push_partial_def. Fix alias-set used in the aggregate copy
2308 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2309 * real.c (clear_significand_below): Fix out-of-bound access.
2311 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2314 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2315 New define_insn patterns.
2317 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2319 * doc/sourcebuild.texi (check-function-bodies): Add an
2320 optional target/xfail selector.
2322 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2324 PR rtl-optimization/93124
2325 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2326 bare USE and CLOBBER insns.
2328 2020-01-22 Andrew Pinski <apinski@marvell.com>
2330 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2332 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2335 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2336 diagnostic_manager into "ana" namespace.
2337 * selftest-run-tests.c (selftest::run_tests): Update for move of
2338 selftest::run_analyzer_selftests to
2339 ana::selftest::run_analyzer_selftests.
2341 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2343 * cfgexpand.c (union_stack_vars): Update the size.
2345 2020-01-22 Richard Biener <rguenther@suse.de>
2347 PR tree-optimization/93381
2348 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2349 throughout, handle all conversions the same.
2351 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2354 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2355 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2356 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2357 Call force_reg on high_in2 unconditionally.
2359 2020-01-22 Martin Liska <mliska@suse.cz>
2361 PR tree-optimization/92924
2362 * profile.c (compute_value_histograms): Divide
2365 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2368 * output.h (assemble_name_resolve): Declare.
2369 * varasm.c (assemble_name_resolve): New function.
2370 (assemble_name): Use it.
2371 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2373 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2375 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2376 update_web_docs_git instead of update_web_docs_svn.
2378 2020-01-21 Andrew Pinski <apinski@marvell.com>
2381 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2382 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2383 (*tlsgd_small_<mode>): Likewise.
2384 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2385 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2386 register. Convert that register back to dest using convert_mode.
2388 2020-01-21 Jim Wilson <jimw@sifive.com>
2390 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2393 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2394 Uros Bizjak <ubizjak@gmail.com>
2397 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2399 (legitimize_tls_address): Do GNU2 TLS address computation in
2400 ptr_mode and zero-extend result to Pmode.
2401 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2402 :P with :PTR and Pmode with ptr_mode.
2403 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2404 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2405 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2407 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2410 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2411 the last two operands are CONST_INT_P before using them as such.
2413 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2415 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2416 to get the integer element types.
2418 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2420 * config/aarch64/aarch64-sve-builtins.h
2421 (function_expander::convert_to_pmode): Declare.
2422 * config/aarch64/aarch64-sve-builtins.cc
2423 (function_expander::convert_to_pmode): New function.
2424 (function_expander::get_contiguous_base): Use it.
2425 (function_expander::prepare_gather_address_operands): Likewise.
2426 * config/aarch64/aarch64-sve-builtins-sve2.cc
2427 (svwhilerw_svwhilewr_impl::expand): Likewise.
2429 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2432 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2433 cfun->machine->label_is_assembled.
2434 (aarch64_print_patchable_function_entry): New.
2435 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2436 * config/aarch64/aarch64.h (struct machine_function): New field,
2439 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2442 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2445 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2448 * cgraph.c (cgraph_edge::resolve_speculation,
2449 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2450 call_stmt_site_hash.
2452 2020-01-21 Martin Liska <mliska@suse.cz>
2454 * config/rs6000/rs6000.c (common_mode_defined): Remove
2457 2020-01-21 Richard Biener <rguenther@suse.de>
2459 PR tree-optimization/92328
2460 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2461 type when value-numbering same-sized store by inserting a
2463 (eliminate_dom_walker::eliminate_stmt): When eliminating
2464 a redundant store handle bit-reinterpretation of the same value.
2466 2020-01-21 Andrew Pinski <apinski@marvel.com>
2469 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2471 (prepare_block_for_update): This. Use a worklist instead of
2474 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2476 * gcc/config/arm/arm.c (clear_operation_p):
2477 Initialise last_regno, skip first iteration
2478 based on the first_set value and use ints instead
2479 of the unnecessary HOST_WIDE_INTs.
2481 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2484 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2485 compare_mode other than SFmode or DFmode.
2487 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2490 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2491 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2492 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2494 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2496 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2498 2020-01-20 Andrew Pinski <apinski@marvell.com>
2501 * targhooks.c (default_print_patchable_function_entry): Use
2502 output_asm_insn to emit the nop instruction.
2504 2020-01-20 Fangrui Song <maskray@google.com>
2507 * targhooks.c (default_print_patchable_function_entry): Align to
2510 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2513 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2514 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2515 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2516 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2517 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2518 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2519 Remove the {q} suffix from lea.
2520 (*tls_dynamic_gnu2_call_64): Renamed to ...
2521 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2522 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2523 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2524 Pass Pmode to gen_tls_dynamic_gnu2_64.
2526 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2528 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2530 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2532 * config/aarch64/aarch64-sve-builtins-base.cc
2533 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2535 2020-01-20 Richard Biener <rguenther@suse.de>
2538 * dwarf2out.c (prune_unused_types): Unconditionally mark
2539 called function DIEs.
2541 2020-01-20 Martin Liska <mliska@suse.cz>
2543 PR tree-optimization/93199
2544 * tree-eh.c (struct leh_state): Add
2545 new field outer_non_cleanup.
2546 (cleanup_is_dead_in): Pass leh_state instead
2547 of eh_region. Add a checking that state->outer_non_cleanup
2548 points to outer non-clean up region.
2549 (lower_try_finally): Record outer_non_cleanup
2551 (lower_catch): Likewise.
2552 (lower_eh_filter): Likewise.
2553 (lower_eh_must_not_throw): Likewise.
2554 (lower_cleanup): Likewise.
2556 2020-01-20 Richard Biener <rguenther@suse.de>
2558 PR tree-optimization/93094
2559 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2560 (vect_transform_loop): Likewise.
2561 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2562 loop_vectorized_call to vect_transform_loop.
2563 * tree-vect-loop.c (vect_transform_loop): Pass down
2564 loop_vectorized_call to vect_loop_versioning.
2565 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2566 the earlier discovered loop_vectorized_call.
2568 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2570 * doc/contribute.texi: Update for SVN -> Git transition.
2571 * doc/install.texi: Likewise.
2573 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2576 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2577 speculative targets.
2578 (verify_speculative_call): New function
2579 (cgraph_node::verify_node): Use it.
2580 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2583 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2586 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2587 (cgraph_edge::make_direct): Remove all indirect targets.
2588 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2589 (cgraph_node::verify_node): Verify that only one call_stmt or
2590 lto_stmt_uid is set.
2591 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2593 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2594 (lto_output_ref): Simplify streaming of stmt.
2595 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2597 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2599 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2600 Mark parameter unused.
2602 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2604 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2606 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2608 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2610 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2612 * Makefile.in: Add coroutine-passes.o.
2613 * builtin-types.def (BT_CONST_SIZE): New.
2614 (BT_FN_BOOL_PTR): New.
2615 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2616 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2617 * coroutine-builtins.def: New file.
2618 * coroutine-passes.cc: New file.
2619 * function.h (struct GTY function): Add a bit to indicate that the
2620 function is a coroutine component.
2621 * internal-fn.c (expand_CO_FRAME): New.
2622 (expand_CO_YIELD): New.
2623 (expand_CO_SUSPN): New.
2624 (expand_CO_ACTOR): New.
2625 * internal-fn.def (CO_ACTOR): New.
2629 * passes.def: Add pass_coroutine_lower_builtins,
2630 pass_coroutine_early_expand_ifns.
2631 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2632 (make_pass_coroutine_early_expand_ifns): New.
2633 * doc/invoke.texi: Document the fcoroutines command line
2636 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2638 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2641 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2642 after checking the argument is a REG. Don't use REGNO (reg)
2643 again to set last_regno, reuse regno variable instead.
2645 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2647 * doc/analyzer.texi (Limitations): Add note about NaN.
2649 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2650 Sudakshina Das <sudi.das@arm.com>
2652 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2653 and valid immediate.
2654 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2655 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2656 * config/arm/constraints.md (Pg): New.
2657 * config/arm/predicates.md (long_shift_imm): New.
2658 (arm_reg_or_long_shift_imm): Likewise.
2659 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2660 (thumb2_lsll): Likewise.
2663 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2664 Sudakshina Das <sudi.das@arm.com>
2666 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2667 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2668 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2669 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2670 * config/arm/thumb2.md (thumb2_asrl): New.
2671 (thumb2_lsll): Likewise.
2673 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2675 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2678 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2680 * gdbinit.in (help-gcc-hooks): New command.
2681 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2682 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2685 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2687 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2688 correct target macro.
2690 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2692 * config/aarch64/aarch64-protos.h
2693 (aarch64_sve_ld1ro_operand_p): New.
2694 * config/aarch64/aarch64-sve-builtins-base.cc
2695 (class load_replicate): New.
2696 (class svld1ro_impl): New.
2697 (class svld1rq_impl): Change to inherit from load_replicate.
2698 (svld1ro): New sve intrinsic function base.
2699 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2700 New DEF_SVE_FUNCTION.
2701 * config/aarch64/aarch64-sve-builtins-base.h
2702 (svld1ro): New decl.
2703 * config/aarch64/aarch64-sve-builtins.cc
2704 (function_expander::add_mem_operand): Modify assert to allow
2706 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2708 * config/aarch64/aarch64.c
2709 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2710 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2711 (aarch64_sve_ld1ro_operand_p): New.
2712 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2713 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2714 * config/aarch64/predicates.md
2715 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2717 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2719 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2720 Introduce this ACLE specified predefined macro.
2721 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2722 (fp): Disabling this disables f64mm.
2723 (simd): Disabling this disables f64mm.
2724 (fp16): Disabling this disables f64mm.
2725 (sve): Disabling this disables f64mm.
2726 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2727 (AARCH64_ISA_F64MM): New.
2728 (TARGET_F64MM): New.
2729 * doc/invoke.texi (f64mm): Document new option.
2731 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2733 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2734 (neoversen1_tunings): Likewise.
2736 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2739 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2740 Add assert to ensure prolog has been emitted.
2741 (aarch64_split_atomic_op): Likewise.
2742 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2743 Use epilogue_completed rather than reload_completed.
2744 (aarch64_atomic_exchange<mode>): Likewise.
2745 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2746 (atomic_nand<mode>): Likewise.
2747 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2748 (atomic_fetch_nand<mode>): Likewise.
2749 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2750 (atomic_nand_fetch<mode>): Likewise.
2752 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2755 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2757 (REVERSE_CONDITION): Delete.
2758 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2759 (CCFP_CCFPE): Likewise.
2760 (e): New mode attribute.
2761 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2762 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2763 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2764 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2765 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2766 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2767 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2768 name of generator from gen_ccmpdi to gen_ccmpccdi.
2769 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2770 the previous comparison but aren't able to, use the new ccmp_rev
2773 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2775 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2776 than testing directly for INTEGER_CST.
2777 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2779 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2781 PR tree-optimization/93292
2782 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2783 get_vectype_for_scalar_type returns NULL.
2785 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2787 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2788 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2790 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2792 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2794 * params.opt: (max-predicted-iterations): Set bounds.
2795 * predict.c (real_almost_one, real_br_prob_base,
2796 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2797 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2798 probabilities; do not truncate to reg_br_prob_bases.
2799 (estimate_loops_at_level): Pass max_cyclic_prob.
2800 (estimate_loops): Compute max_cyclic_prob.
2801 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2803 * profile-count.c (profile_probability::to_sreal): New.
2804 * profile-count.h (class sreal): Move up in file.
2805 (profile_probability::to_sreal): Declare.
2807 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2810 (arm_invalid_conversion): New function for target hook.
2811 (arm_invalid_unary_op): New function for target hook.
2812 (arm_invalid_binary_op): New function for target hook.
2814 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2816 * config.gcc: Add arm_bf16.h.
2817 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2818 (arm_simd_builtin_std_type): Add BFmode.
2819 (arm_init_simd_builtin_types): Define element types for vector types.
2820 (arm_init_bf16_types): New function.
2821 (arm_init_builtins): Add arm_init_bf16_types function call.
2822 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2823 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2824 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2825 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2826 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2827 (arm_mangle_type): Add __bf16.
2828 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2829 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2830 arm_bf16_ptr_type_node.
2831 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2832 define_split between ARM registers.
2833 * config/arm/arm_bf16.h: New file.
2834 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2835 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2837 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2838 * config/arm/vfp.md: Add BFmode to movhf patterns.
2840 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2841 Andre Vieira <andre.simoesdiasvieira@arm.com>
2843 * config/arm/arm-cpus.in (mve, mve_float): New features.
2844 (dsp, mve, mve.fp): New options.
2845 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2846 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2847 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2849 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2850 Thomas Preud'homme <thomas.preudhomme@arm.com>
2852 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2854 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2855 error for using -mcmse when targeting Armv8.1-M Mainline.
2857 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2858 Thomas Preud'homme <thomas.preudhomme@arm.com>
2860 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2861 address in r4 when targeting Armv8.1-M Mainline.
2862 (nonsecure_call_value_internal): Likewise.
2863 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2864 a register match_operand again. Emit BLXNS when targeting
2866 (nonsecure_call_value_reg_thumb2): Likewise.
2868 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2869 Thomas Preud'homme <thomas.preudhomme@arm.com>
2871 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2872 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2873 variable as true when floating-point ABI is not hard. Replace
2874 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2875 Generate VLSTM and VLLDM instruction respectively before and
2876 after a function call to cmse_nonsecure_call function.
2877 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2878 (VUNSPEC_VLLDM): Likewise.
2879 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2880 (lazy_load_multiple_insn): Likewise.
2882 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2883 Thomas Preud'homme <thomas.preudhomme@arm.com>
2885 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2886 (arm_emit_vfp_multi_reg_pop): Likewise.
2887 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2888 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2889 restore callee-saved VFP registers.
2891 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2892 Thomas Preud'homme <thomas.preudhomme@arm.com>
2894 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2895 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2896 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2897 callee-saved GPRs as well as clear ip register before doing a nonsecure
2898 call then restore callee-saved GPRs after it when targeting
2900 (arm_reorg): Adapt to function rename.
2902 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2903 Thomas Preud'homme <thomas.preudhomme@arm.com>
2905 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2906 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2907 clear_vfp_multiple pattern based on a new vfp parameter.
2908 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2909 targeting Armv8.1-M Mainline.
2910 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2911 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2912 whether VFP registers are available before looking call_used_regs for a
2914 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2915 of prototype of clear_operation_p.
2916 (clear_vfp_multiple_operation): New predicate.
2917 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2918 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2920 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2921 Thomas Preud'homme <thomas.preudhomme@arm.com>
2923 * config/arm/arm-protos.h (clear_operation_p): Declare.
2924 * config/arm/arm.c (clear_operation_p): New function.
2925 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2926 targeting Armv8.1-M Mainline or successor.
2927 (output_return_instruction): Only output APSR register clearing if
2928 Armv8.1-M Mainline instructions not available.
2929 (thumb_exit): Likewise.
2930 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2931 * config/arm/thumb2.md (clear_apsr): New define_insn.
2932 (clear_multiple): Likewise.
2933 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2935 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2936 Thomas Preud'homme <thomas.preudhomme@arm.com>
2938 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2939 (use_return_insn): Also return false for Armv8.1-M Mainline.
2940 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2941 Mainline instructions are available.
2942 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2943 when targeting Armv8.1-M Mainline Security Extensions.
2944 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2945 Mainline entry function.
2946 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2947 targeting Armv8.1-M Mainline or successor.
2948 (arm_expand_epilogue): Fix indentation of caller-saved register
2949 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2951 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2952 (FP_SYSREGS): Likewise.
2953 (enum vfp_sysregs_encoding): Define enum.
2954 (fp_sysreg_names): Declare.
2955 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2956 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2957 (pop_fpsysreg_insn): Likewise.
2959 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2960 Thomas Preud'homme <thomas.preudhomme@arm.com>
2962 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2963 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2964 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2965 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2966 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2967 (ARMv8_1m_main): New feature group.
2968 (armv8.1-m.main): New architecture.
2969 * config/arm/arm-tables.opt: Regenerate.
2970 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2971 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2972 (arm_options_perform_arch_sanity_checks): Error out when targeting
2973 Armv8.1-M Mainline Security Extensions.
2974 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2976 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2978 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2979 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2980 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2981 aarch64_bfdot_laneq): New.
2982 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2983 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2984 vbfdotq_laneq_f32): New.
2985 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2986 VBFMLA_W, VBF): New.
2987 (isquadop): Add V4BF, V8BF.
2989 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2991 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2992 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2993 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2994 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2995 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2996 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2997 usdot_laneq, sudot_lane,sudot_laneq): New.
2998 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2999 (aarch64_<sur>dot_lane): New.
3000 * config/aarch64/arm_neon.h (vusdot_s32): New.
3002 (vusdot_lane_s32): New.
3003 (vsudot_lane_s32): New.
3004 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3005 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3007 2020-01-16 Martin Liska <mliska@suse.cz>
3009 * value-prof.c (dump_histogram_value): Fix
3010 obvious spacing issue.
3012 2020-01-16 Andrew Pinski <apinski@marvell.com>
3014 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3015 !storage_order_barrier_p.
3017 2020-01-16 Andrew Pinski <apinski@marvell.com>
3019 * sched-int.h (_dep): Add unused bit-field field for the padding.
3020 * sched-deps.c (init_dep_1): Init unused field.
3022 2020-01-16 Andrew Pinski <apinski@marvell.com>
3024 * optabs.h (create_expand_operand): Initialize target field also.
3026 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3028 PR tree-optimization/92429
3029 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3030 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3032 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3035 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3037 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3038 aarch64_sve_int_mode to each mode.
3040 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3042 * doc/analyzer.texi (Overview): Add note about
3043 -fdump-ipa-analyzer.
3045 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3047 PR tree-optimization/93231
3048 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3049 input_type is unsigned. Use tree_to_shwi for shift constant.
3050 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3051 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3052 inputs more efficiently.
3054 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3056 * config/i386/i386.md (*movsf_internal): Do not require
3057 SSE2 ISA for alternatives 14 and 15.
3059 2020-01-15 Richard Biener <rguenther@suse.de>
3062 * tree-eh.c (sink_clobbers): If we already visited the destination
3063 block do not defer insertion.
3064 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3065 the purpose of defered insertion.
3067 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3069 * BASE-VER: Bump to 10.0.1.
3071 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3073 PR tree-optimization/93247
3074 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3075 type of the stmt that we're going to vectorize.
3077 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3079 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3080 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3083 2020-01-15 Martin Liska <mliska@suse.cz>
3085 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3086 2 calls of streamer_read_hwi in a function call.
3088 2020-01-15 Richard Biener <rguenther@suse.de>
3090 * alias.c (record_alias_subset): Avoid redundant work when
3091 subset is already recorded.
3093 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3095 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3096 the analyzer options provide CWE identifiers.
3098 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3100 * tree-diagnostic-path.cc (path_summary::event_range::print):
3101 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3102 using get_pure_location.
3104 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3106 PR tree-optimization/93262
3107 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3108 perform head trimming only if the last argument is constant,
3109 either all ones, or larger or equal to head trim, in the latter
3110 case decrease the last argument by head_trim.
3112 PR tree-optimization/93249
3113 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3114 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3115 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3116 perform head trim unless we can prove there are no '\0' chars
3117 from the source among the first head_trim chars.
3119 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3121 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3123 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3126 * config/i386/sse.md
3127 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3128 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3129 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3130 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3131 just a single alternative instead of two, make operands 1 and 2
3134 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3137 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3140 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3142 * Makefile.in (lang_opt_files): Add analyzer.opt.
3143 (ANALYZER_OBJS): New.
3144 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3145 tristate.o and ANALYZER_OBJS.
3146 (TEXI_GCCINT_FILES): Add analyzer.texi.
3147 * common.opt (-fanalyzer): New driver option.
3148 * config.in: Regenerate.
3149 * configure: Regenerate.
3150 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3151 (gccdepdir): Also create depdir for "analyzer" subdir.
3152 * digraph.cc: New file.
3153 * digraph.h: New file.
3154 * doc/analyzer.texi: New file.
3155 * doc/gccint.texi ("Static Analyzer") New menu item.
3156 (analyzer.texi): Include it.
3157 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3158 ("Warning Options"): Add static analysis warnings to the list.
3159 (-Wno-analyzer-double-fclose): New option.
3160 (-Wno-analyzer-double-free): New option.
3161 (-Wno-analyzer-exposure-through-output-file): New option.
3162 (-Wno-analyzer-file-leak): New option.
3163 (-Wno-analyzer-free-of-non-heap): New option.
3164 (-Wno-analyzer-malloc-leak): New option.
3165 (-Wno-analyzer-possible-null-argument): New option.
3166 (-Wno-analyzer-possible-null-dereference): New option.
3167 (-Wno-analyzer-null-argument): New option.
3168 (-Wno-analyzer-null-dereference): New option.
3169 (-Wno-analyzer-stale-setjmp-buffer): New option.
3170 (-Wno-analyzer-tainted-array-index): New option.
3171 (-Wno-analyzer-use-after-free): New option.
3172 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3173 (-Wno-analyzer-use-of-uninitialized-value): New option.
3174 (-Wanalyzer-too-complex): New option.
3175 (-fanalyzer-call-summaries): New warning.
3176 (-fanalyzer-checker=): New warning.
3177 (-fanalyzer-fine-grained): New warning.
3178 (-fno-analyzer-state-merge): New warning.
3179 (-fno-analyzer-state-purge): New warning.
3180 (-fanalyzer-transitivity): New warning.
3181 (-fanalyzer-verbose-edges): New warning.
3182 (-fanalyzer-verbose-state-changes): New warning.
3183 (-fanalyzer-verbosity=): New warning.
3184 (-fdump-analyzer): New warning.
3185 (-fdump-analyzer-callgraph): New warning.
3186 (-fdump-analyzer-exploded-graph): New warning.
3187 (-fdump-analyzer-exploded-nodes): New warning.
3188 (-fdump-analyzer-exploded-nodes-2): New warning.
3189 (-fdump-analyzer-exploded-nodes-3): New warning.
3190 (-fdump-analyzer-supergraph): New warning.
3191 * doc/sourcebuild.texi (dg-require-dot): New.
3192 (dg-check-dot): New.
3193 * gdbinit.in (break-on-saved-diagnostic): New command.
3194 * graphviz.cc: New file.
3195 * graphviz.h: New file.
3196 * ordered-hash-map-tests.cc: New file.
3197 * ordered-hash-map.h: New file.
3198 * passes.def (pass_analyzer): Add before
3199 pass_ipa_whole_program_visibility.
3200 * selftest-run-tests.c (selftest::run_tests): Call
3201 selftest::ordered_hash_map_tests_cc_tests.
3202 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3204 * shortest-paths.h: New file.
3205 * timevar.def (TV_ANALYZER): New timevar.
3206 (TV_ANALYZER_SUPERGRAPH): Likewise.
3207 (TV_ANALYZER_STATE_PURGE): Likewise.
3208 (TV_ANALYZER_PLAN): Likewise.
3209 (TV_ANALYZER_SCC): Likewise.
3210 (TV_ANALYZER_WORKLIST): Likewise.
3211 (TV_ANALYZER_DUMP): Likewise.
3212 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3213 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3214 * tree-pass.h (make_pass_analyzer): New decl.
3215 * tristate.cc: New file.
3216 * tristate.h: New file.
3218 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3221 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3222 alternatives 9 and 10.
3224 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3226 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3227 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3228 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3229 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3230 (selftest::hash_map_tests_c_tests): Call it.
3231 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3232 New static constant, using the value of = H::empty_zero_p.
3233 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3234 from default_hash_traits <Value>.
3235 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3237 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3238 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3239 calls to mark_empty with !Descriptor::empty_zero_p.
3240 (hash_table::empty_slow): Conditionalize the memset call with a
3241 check that Descriptor::empty_zero_p; otherwise, loop through the
3242 entries calling mark_empty on them.
3243 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3244 (pointer_hash::empty_zero_p): Likewise.
3245 (pair_hash::empty_zero_p): Likewise.
3246 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3248 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3249 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3250 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3251 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3252 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3253 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3254 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3255 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3257 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3260 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3262 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3263 fix typo on return value.
3265 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3268 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3270 (cgraph_edge::make_speculative): Add param for setting speculative_id
3272 (cgraph_edge::speculative_call_info): Update comments and find reference
3273 by speculative_id for multiple indirect targets.
3274 (cgraph_edge::resolve_speculation): Decrease the speculations
3275 for indirect edge, drop it's speculative if not direct target
3276 left. Update comments.
3277 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3278 (cgraph_node::dump): Print num_speculative_call_targets.
3279 (cgraph_node::verify_node): Don't report error if speculative
3280 edge not include statement.
3281 (cgraph_edge::num_speculative_call_targets_p): New function.
3282 * cgraph.h (int common_target_id): Remove.
3283 (int common_target_probability): Remove.
3284 (num_speculative_call_targets): New variable.
3285 (make_speculative): Add param for setting speculative_id.
3286 (cgraph_edge::num_speculative_call_targets_p): New declare.
3287 (target_prob): New variable.
3288 (speculative_id): New variable.
3289 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3290 call summaries for multiple speculative call targets.
3291 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3292 * ipa-profile.c (struct speculative_call_target): New struct.
3293 (class speculative_call_summary): New class.
3294 (class speculative_call_summaries): New class.
3295 (call_sums): New variable.
3296 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3297 (ipa_profile_write_edge_summary): New function.
3298 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3299 (ipa_profile_dump_all_summaries): New function.
3300 (ipa_profile_read_edge_summary): New function.
3301 (ipa_profile_read_summary_section): New function.
3302 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3303 (ipa_profile): Generate num_speculative_call_targets from
3305 * ipa-ref.h (speculative_id): New variable.
3306 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3307 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3308 common_target_probability. Stream out speculative_id and
3309 num_speculative_call_targets.
3310 (input_edge): Likewise.
3311 * predict.c (dump_prediction): Remove edges count assert to be
3313 * symtab.c (symtab_node::create_reference): Init speculative_id.
3314 (symtab_node::clone_references): Clone speculative_id.
3315 (symtab_node::clone_referring): Clone speculative_id.
3316 (symtab_node::clone_reference): Clone speculative_id.
3317 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3318 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3319 if indirect call contains multiple speculative targets.
3320 * value-prof.h (check_ic_target): Remove.
3321 * value-prof.c (gimple_value_profile_transformations):
3322 Use void function gimple_ic_transform.
3323 * value-prof.c (gimple_ic_transform): Handle topn case.
3324 Fix comment typos. Change it to a void function.
3326 2020-01-13 Andrew Pinski <apinski@marvell.com>
3328 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3329 (octeontx2t98): New define.
3330 (octeontx2t96): New define.
3331 (octeontx2t93): New define.
3332 (octeontx2f95): New define.
3333 (octeontx2f95n): New define.
3334 (octeontx2f95mm): New define.
3335 * config/aarch64/aarch64-tune.md: Regenerate.
3336 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3338 2020-01-13 Jason Merrill <jason@redhat.com>
3340 PR c++/33799 - destroy return value if local cleanup throws.
3341 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3343 2020-01-13 Martin Liska <mliska@suse.cz>
3345 * ipa-cp.c (get_max_overall_size): Use newly
3346 renamed param param_ipa_cp_unit_growth.
3347 * params.opt: Remove legacy param name.
3349 2020-01-13 Martin Sebor <msebor@redhat.com>
3351 PR tree-optimization/93213
3352 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3353 stores to be eliminated.
3355 2020-01-13 Martin Liska <mliska@suse.cz>
3357 * opts.c (print_help): Do not print CL_PARAM
3358 and CL_WARNING for CL_OPTIMIZATION.
3360 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3363 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3364 depending on optimization settings.
3366 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3368 PR tree-optimization/90838
3369 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3370 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3371 argument rather than to initialize temporary for targets that
3372 don't use the mode argument at all. Initialize ctzval to avoid
3375 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3377 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3378 * tree-core.h: Document it.
3379 * gimplify.c (gimplify_omp_workshare): Set it.
3380 * omp-low.c (lower_omp_target): Use it.
3381 * tree-pretty-print.c (dump_omp_clause): Print it.
3383 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3384 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3386 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3388 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3389 * common.opt (fdiagnostics-path-format=): New option.
3390 (diagnostic_path_format): New enum.
3391 (fdiagnostics-show-path-depths): New option.
3392 * coretypes.h (diagnostic_event_id_t): New forward decl.
3393 * diagnostic-color.c (color_dict): Add "path".
3394 * diagnostic-event-id.h: New file.
3395 * diagnostic-format-json.cc (json_from_expanded_location): Make
3397 (json_end_diagnostic): Call context->make_json_for_path if it
3398 exists and the diagnostic has a path.
3399 (diagnostic_output_format_init): Clear context->print_path.
3400 * diagnostic-path.h: New file.
3401 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3402 when printing a run of events in a diagnostic_path so that they
3403 all get the same color.
3404 (layout::m_diagnostic_path_p): New field.
3405 (layout::layout): Initialize it.
3406 (layout::print_any_labels): Don't colorize the label text for an
3407 event in a diagnostic_path.
3408 (gcc_rich_location::add_location_if_nearby): Add
3409 "restrict_to_current_line_spans" and "label" params. Pass the
3410 former to layout.maybe_add_location_range; pass the latter
3411 when calling add_range.
3412 * diagnostic.c: Include "diagnostic-path.h".
3413 (diagnostic_initialize): Initialize context->path_format and
3414 context->show_path_depths.
3415 (diagnostic_show_any_path): New function.
3416 (diagnostic_path::interprocedural_p): New function.
3417 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3418 (simple_diagnostic_path::num_events): New function.
3419 (simple_diagnostic_path::get_event): New function.
3420 (simple_diagnostic_path::add_event): New function.
3421 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3422 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3423 (debug): New overload taking a diagnostic_path *.
3424 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3425 * diagnostic.h (enum diagnostic_path_format): New enum.
3426 (json::value): New forward decl.
3427 (diagnostic_context::path_format): New field.
3428 (diagnostic_context::show_path_depths): New field.
3429 (diagnostic_context::print_path): New callback field.
3430 (diagnostic_context::make_json_for_path): New callback field.
3431 (diagnostic_show_any_path): New decl.
3432 (json_from_expanded_location): New decl.
3433 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3434 (-fdiagnostics-show-path-depths): New option.
3435 (-fdiagnostics-color): Add "path" to description of default
3436 GCC_COLORS; describe it.
3437 (-fdiagnostics-format=json): Document how diagnostic paths are
3438 represented in the JSON output format.
3439 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3440 Add optional params "restrict_to_current_line_spans" and "label".
3441 * opts.c (common_handle_option): Handle
3442 OPT_fdiagnostics_path_format_ and
3443 OPT_fdiagnostics_show_path_depths.
3444 * pretty-print.c: Include "diagnostic-event-id.h".
3445 (pp_format): Implement "%@" format code for printing
3446 diagnostic_event_id_t *.
3447 (selftest::test_pp_format): Add tests for "%@".
3448 * selftest-run-tests.c (selftest::run_tests): Call
3449 selftest::tree_diagnostic_path_cc_tests.
3450 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3451 * toplev.c (general_init): Initialize global_dc->path_format and
3452 global_dc->show_path_depths.
3453 * tree-diagnostic-path.cc: New file.
3454 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3455 non-static. Drop "diagnostic" param in favor of storing the
3456 original value of "where" and re-using it.
3457 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3458 maybe_unwind_expanded_macro_loc.
3459 (tree_diagnostics_defaults): Initialize context->print_path and
3460 context->make_json_for_path.
3461 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3463 (default_tree_make_json_for_path): New decl.
3464 (maybe_unwind_expanded_macro_loc): New decl.
3466 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3468 PR tree-optimization/93210
3469 * fold-const.h (native_encode_initializer,
3470 can_native_interpret_type_p): Declare.
3471 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3473 (native_encode_initializer): New function, moved from dwarf2out.c.
3474 Adjust to native_encode_expr compatible arguments, including dry-run
3475 and partial extraction modes. Don't handle STRING_CST.
3476 (can_native_interpret_type_p): No longer static.
3477 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3478 offset / BITS_PER_UNIT fits into int and don't call it if
3479 can_native_interpret_type_p fails. If suboff is NULL and for
3480 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3481 native_encode_initializer.
3482 (fold_const_aggregate_ref_1): Formatting fix.
3483 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3484 (tree_add_const_value_attribute): Adjust caller.
3486 PR tree-optimization/90838
3487 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3488 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3489 CTZ_DEFINED_VALUE_AT_ZERO.
3491 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3494 * lra-constraints.c (match_reload): Permit input operands have the
3495 same mode as output while other input operands have a different
3498 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3500 PR tree-optimization/90838
3501 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3502 (check_ctz_string): Likewise.
3503 (optimize_count_trailing_zeroes): Likewise.
3504 (simplify_count_trailing_zeroes): Likewise.
3505 (pass_forwprop::execute): Try ctz simplification.
3506 * match.pd: Add matching for ctz idioms.
3508 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3510 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3512 (aarch64_invalid_unary_op): New function for target hook.
3513 (aarch64_invalid_binary_op): New function for target hook.
3515 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3517 * config.gcc: Add arm_bf16.h.
3518 * config/aarch64/aarch64-builtins.c
3519 (aarch64_simd_builtin_std_type): Add BFmode.
3520 (aarch64_init_simd_builtin_types): Define element types for vector
3522 (aarch64_init_bf16_types): New function.
3523 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3524 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3526 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3527 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3529 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3530 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3531 * config/aarch64/aarch64.c
3532 (aarch64_classify_vector_mode): Add support for BF types.
3533 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3534 (aarch64_vq_mode): Add support for BF types.
3535 (aarch64_simd_container_mode): Add support for BF types.
3536 (aarch64_mangle_type): Add support for BF scalar type.
3537 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3538 * config/aarch64/arm_bf16.h: New file.
3539 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3540 * config/aarch64/iterators.md: Add BF types to mode attributes.
3541 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3543 2020-01-10 Jason Merrill <jason@redhat.com>
3545 PR c++/93173 - incorrect tree sharing.
3546 * gimplify.c (copy_if_shared): No longer static.
3547 * gimplify.h: Declare it.
3549 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3551 * doc/invoke.texi (-msve-vector-bits=): Document that
3552 -msve-vector-bits=128 now generates VL-specific code for
3553 little-endian targets.
3554 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3555 build_vector_type_for_mode to construct the data vector types.
3556 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3557 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3558 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3559 for 128-bit vectors.
3561 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3563 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3566 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3568 * config/aarch64/aarch64-builtins.c
3569 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3570 rather than checking the number of elements and the element mode.
3572 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3574 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3575 get_related_vectype_for_scalar_type rather than build_vector_type
3576 to create the index type for a conditional reduction.
3578 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3580 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3581 for any type of gather or scatter, including strided accesses.
3583 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3585 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3588 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3590 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3592 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3593 parameter and its use to reset DR_OFFSET's.
3594 (vect_transform_loop): Remove orig_drs_init argument.
3595 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3596 member of dr_vec_info rather than the offset of the associated
3597 data_reference's innermost_loop_behavior.
3598 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3599 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3600 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3601 get_dr_vinfo_offset.
3602 (vectorizable_store): Likewise.
3603 (vectorizable_load): Likewise.
3605 2020-01-10 Richard Biener <rguenther@suse.de>
3607 * gimple-ssa-store-merging
3608 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3610 2020-01-10 Martin Liska <mliska@suse.cz>
3613 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3614 encapsulation that was there before r280040.
3616 2020-01-10 Richard Biener <rguenther@suse.de>
3619 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3620 sequences to avoid walking them again for secondary opportunities.
3621 (pass_lower_eh_dispatch::execute): Instead actually insert
3624 2020-01-10 Richard Biener <rguenther@suse.de>
3627 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3628 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3629 avoid quadraticness.
3631 2020-01-10 Martin Jambor <mjambor@suse.cz>
3633 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3634 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3635 to get param_ipa_sra_max_replacements.
3636 (param_splitting_across_edge): Pass the caller to
3637 pull_accesses_from_callee.
3639 2020-01-10 Martin Jambor <mjambor@suse.cz>
3641 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3642 * ipa-cp.c (max_new_size): Removed.
3643 (orig_overall_size): New variable.
3644 (get_max_overall_size): New function.
3645 (estimate_local_effects): Use it. Adjust dump.
3646 (decide_about_value): Likewise.
3647 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3648 orig_overall_size. Adjust dump.
3649 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3651 2020-01-10 Martin Jambor <mjambor@suse.cz>
3653 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3654 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3655 instead of param_ipa_max_agg_items.
3656 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3657 optimization info for the callee.
3659 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3661 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3662 markers if debug_inline_points is false.
3664 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3666 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3668 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3669 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3670 aarch64-sve-builtins-sve2.h.
3671 (aarch64-sve-builtins-sve2.o): New rule.
3672 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3673 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3674 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3675 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3676 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3677 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3679 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3680 instructions that are handled here.
3681 (@cond_asrd<mode>): Generalize to...
3682 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3683 (*cond_asrd<mode>_2): Generalize to...
3684 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3685 (*cond_asrd<mode>_z): Generalize to...
3686 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3687 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3688 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3689 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3690 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3692 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3693 (@aarch64_scatter_stnt<mode>): Likewise.
3694 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3695 (@aarch64_mul_lane_<mode>): Likewise.
3696 (@aarch64_sve_suqadd<mode>_const): Likewise.
3697 (*<sur>h<addsub><mode>): Generalize to...
3698 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3700 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3701 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3702 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3703 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3704 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3705 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3706 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3707 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3708 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3709 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3710 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3711 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3712 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3713 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3714 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3715 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3716 (@aarch64_sve2_xar<mode>): Likewise.
3717 (@aarch64_sve2_bcax<mode>): Likewise.
3718 (*aarch64_sve2_eor3<mode>): Rename to...
3719 (@aarch64_sve2_eor3<mode>): ...this.
3720 (@aarch64_sve2_bsl<mode>): New expander.
3721 (@aarch64_sve2_nbsl<mode>): Likewise.
3722 (@aarch64_sve2_bsl1n<mode>): Likewise.
3723 (@aarch64_sve2_bsl2n<mode>): Likewise.
3724 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3725 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3726 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3727 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3728 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3729 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3730 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3731 (<su>mull<bt><Vwide>): Generalize to...
3732 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3734 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3735 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3736 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3737 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3738 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3739 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3740 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3741 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3742 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3743 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3744 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3745 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3746 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3747 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3748 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3749 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3750 (<SHRNB:r>shrnb<mode>): Generalize to...
3751 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3753 (<SHRNT:r>shrnt<mode>): Generalize to...
3754 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3756 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3757 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3758 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3759 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3760 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3761 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3762 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3763 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3764 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3765 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3766 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3767 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3768 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3769 (@aarch64_sve2_cvtnt<mode>): Likewise.
3770 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3771 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3772 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3773 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3774 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3775 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3776 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3777 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3778 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3779 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3780 (@aarch64_sve2_pmul<mode>): Likewise.
3781 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3782 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3783 (@aarch64_sve2_tbl2<mode>): Likewise.
3784 (@aarch64_sve2_tbx<mode>): Likewise.
3785 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3786 (@aarch64_sve2_histcnt<mode>): Likewise.
3787 (@aarch64_sve2_histseg<mode>): Likewise.
3788 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3789 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3790 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3791 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3792 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3793 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3794 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3795 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3796 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3797 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3798 (SVE2_PMULL_PAIR_I): New mode iterators.
3799 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3800 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3801 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3802 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3803 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3804 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3805 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3806 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3807 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3808 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3809 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3810 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3811 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3812 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3813 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3814 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3815 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3816 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3817 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3818 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3819 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3820 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3821 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3822 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3823 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3824 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3825 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3826 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3827 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3828 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3829 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3831 (VNARROW, Ventype): New mode attributes.
3832 (Vewtype): Handle VNx2DI. Fix typo in comment.
3833 (VDOUBLE): New mode attribute.
3834 (sve_lane_con): Handle VNx8HI.
3835 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3836 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3837 (sve_int_op, sve_int_op_rev): Handle the above codes.
3838 (sve_pred_int_rhs2_operand): Likewise.
3839 (MULLBT, SHRNB, SHRNT): Delete.
3840 (SVE_INT_SHIFT_IMM): New int iterator.
3841 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3842 and UNSPEC_WHILEHS for TARGET_SVE2.
3843 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3844 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3845 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3846 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3847 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3848 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3849 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3850 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3851 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3852 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3853 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3854 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3855 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3856 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3857 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3858 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3859 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3860 (optab): Handle the new unspecs.
3861 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3863 (lr): Handle the new unspecs.
3865 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3866 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3867 (sve_int_qsub_op): New int attributes.
3868 (sve_fp_op, rot): Handle the new unspecs.
3869 * config/aarch64/aarch64-sve-builtins.h
3870 (function_resolver::require_matching_pointer_type): Declare.
3871 (function_resolver::resolve_unary): Add an optional boolean argument.
3872 (function_resolver::finish_opt_n_resolution): Add an optional
3873 type_suffix_index argument.
3874 (gimple_folder::redirect_call): Declare.
3875 (gimple_expander::prepare_gather_address_operands): Add an optional
3877 * config/aarch64/aarch64-sve-builtins.cc: Include
3878 aarch64-sve-builtins-sve2.h.
3879 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3880 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3881 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3882 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3883 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3884 (TYPES_s_unsigned): Likewise.
3885 (TYPES_s_integer): Use TYPES_s_unsigned.
3886 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3887 (TYPES_sd_integer): Use them.
3888 (TYPES_d_unsigned): New macro.
3889 (TYPES_d_integer): Use it.
3890 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3891 (TYPES_cvt_narrow): Likewise.
3892 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3893 (preds_mx): New variable.
3894 (function_builder::add_overloaded_function): Allow the new feature
3895 set to be more restrictive than the original one.
3896 (function_resolver::infer_pointer_type): Remove qualifiers from
3897 the pointer type before printing it.
3898 (function_resolver::require_matching_pointer_type): New function.
3899 (function_resolver::resolve_sv_displacement): Handle functions
3900 that don't support 32-bit vector indices or svint32_t vector offsets.
3901 (function_resolver::finish_opt_n_resolution): Take the inferred type
3902 as a separate argument.
3903 (function_resolver::resolve_unary): Optionally treat all forms in
3904 the same way as normal merging functions.
3905 (gimple_folder::redirect_call): New function.
3906 (function_expander::prepare_gather_address_operands): Add an argument
3907 that says whether scaled forms are available. If they aren't,
3908 handle scaling of vector indices and don't add the extension and
3910 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3911 fall back to using cond_* instead.
3912 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3913 Split out the member variables into...
3914 (rtx_code_function_base): ...this new base class.
3915 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3916 (unspec_based_function): Split out the member variables into...
3917 (unspec_based_function_base): ...this new base class.
3918 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3919 (unspec_based_function_exact_insn): New class.
3920 (unspec_based_add_function, unspec_based_add_lane_function)
3921 (unspec_based_lane_function, unspec_based_pred_function)
3922 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3923 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3924 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3926 (unspec_based_fused_function): New class.
3927 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3928 (unspec_based_fused_lane_function): New class.
3929 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3931 (CODE_FOR_MODE1): New macro.
3932 (fixed_insn_function): New class.
3933 (while_comparison): Likewise.
3934 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3935 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3936 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3937 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3938 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3939 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3940 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3941 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3942 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3943 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3944 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3945 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3946 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3947 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3948 Also add an initial argument for unary_convert_narrowt, regardless
3949 of the predication type.
3950 (build_32_64): Allow loads and stores to specify MODE_none.
3951 (build_sv_index64, build_sv_uint_offset): New functions.
3952 (long_type_suffix): New function.
3953 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3954 (binary_imm_long_base, load_gather_sv_base): Likewise.
3955 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3956 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3957 (unary_narrowb_base, unary_narrowt_base): Likewise.
3958 (binary_long_lane_def, binary_long_lane): New shape.
3959 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3960 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3961 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3962 (binary_to_uint_def, binary_to_uint): Likewise.
3963 (binary_wide_def, binary_wide): Likewise.
3964 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3965 (compare_def, compare): Likewise.
3966 (compare_ptr_def, compare_ptr): Likewise.
3967 (load_ext_gather_index_restricted_def,
3968 load_ext_gather_index_restricted): Likewise.
3969 (load_ext_gather_offset_restricted_def,
3970 load_ext_gather_offset_restricted): Likewise.
3971 (load_gather_sv_def): Inherit from load_gather_sv_base.
3972 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3973 (shift_left_imm_def, shift_left_imm): Likewise.
3974 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3975 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3976 (store_scatter_index_restricted_def,
3977 store_scatter_index_restricted): Likewise.
3978 (store_scatter_offset_restricted_def,
3979 store_scatter_offset_restricted): Likewise.
3980 (tbl_tuple_def, tbl_tuple): Likewise.
3981 (ternary_long_lane_def, ternary_long_lane): Likewise.
3982 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3983 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3984 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3985 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3986 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3987 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3988 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3989 (ternary_uint_def, ternary_uint): Likewise.
3990 (unary_convert): Fix typo in comment.
3991 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3992 (unary_long_def, unary_long): Likewise.
3993 (unary_narrowb_def, unary_narrowb): Likewise.
3994 (unary_narrowt_def, unary_narrowt): Likewise.
3995 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3996 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3997 (unary_to_int_def, unary_to_int): Likewise.
3998 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3999 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4000 (svasrd_impl): Delete.
4001 (svcadd_impl::expand): Handle integer operations too.
4002 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4003 new functions to derive the unspec numbers.
4004 (svmla_svmls_lane_impl): Replace with...
4005 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
4006 integer operations too.
4007 (svwhile_impl): Rename to...
4008 (svwhilelx_impl): ...this and inherit from while_comparison.
4009 (svasrd): Use unspec_based_function.
4010 (svmla_lane): Use svmla_lane_impl.
4011 (svmls_lane): Use svmls_lane_impl.
4012 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4013 (svwhilele, svwhilelt): Use svwhilelx_impl.
4014 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4015 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4016 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4017 * config/aarch64/aarch64-sve-builtins.def: Include
4018 aarch64-sve-builtins-sve2.def.
4020 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4022 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4023 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4024 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4025 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4026 immediates as well as vector ones.
4027 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4028 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4029 (aarch64_sve_qsub_immediate): Update calls accordingly.
4031 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4033 * config/aarch64/aarch64-sve2.md: Add banner comments.
4034 (<su>mulh<r>s<mode>3): Move further up file.
4035 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4036 (*aarch64_sve2_sra<mode>): Move further down file.
4037 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4039 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4041 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4043 (while_optab_cmp): Handle them.
4044 * config/aarch64/aarch64-sve.md
4045 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4046 and add a "@" marker.
4047 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4048 instead of gen_aarch64_sve2_while_ptest.
4049 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4051 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4053 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4054 (UNSPEC_WHILELE): ...this.
4055 (UNSPEC_WHILE_LO): Rename to...
4056 (UNSPEC_WHILELO): ...this.
4057 (UNSPEC_WHILE_LS): Rename to...
4058 (UNSPEC_WHILELS): ...this.
4059 (UNSPEC_WHILE_LT): Rename to...
4060 (UNSPEC_WHILELT): ...this.
4061 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4062 (cmp_op, while_optab_cmp): Likewise.
4063 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4064 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4065 (svwhilelt): Likewise.
4067 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4069 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4070 (unary_to_uint): Define.
4071 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4072 (unary_count): Rename to...
4073 (unary_to_uint_def, unary_to_uint): ...this.
4074 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4076 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4078 * config/aarch64/aarch64-sve-builtins-functions.h
4079 (code_for_mode_function): New class.
4080 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4081 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4082 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4083 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4084 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4086 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4088 * config/aarch64/iterators.md (addsub): New code attribute.
4089 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4091 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4092 in the asm string and attributes. Fix indentation.
4093 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4095 (@aarch64_sve_<optab><mode>): ...this.
4096 * config/aarch64/aarch64-sve-builtins.h
4097 (function_expander::expand_signed_unpred_op): Delete.
4098 * config/aarch64/aarch64-sve-builtins.cc
4099 (function_expander::expand_signed_unpred_op): Likewise.
4100 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4101 try using code_for_aarch64_sve instead.
4102 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4103 (svqsub_impl): Likewise.
4104 (svqadd, svqsub): Use rtx_code_function instead.
4106 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4108 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4109 (HADDSUB, sur, addsub): Remove them.
4111 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4113 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4114 functions the same way, rather than singling out those that
4115 aren't mapped directly to optabs.
4117 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4119 * target.def (compatible_vector_types_p): New target hook.
4120 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4121 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4122 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4123 * doc/tm.texi: Regenerate.
4124 * gimple-expr.c: Include target.h.
4125 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4126 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4128 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4129 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4130 Use the original predicate if it already has a suitable type.
4132 2020-01-09 Martin Jambor <mjambor@suse.cz>
4134 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4135 resolve_speculation and redirect_call_stmt_to_callee static. Change
4136 return type of set_call_stmt to cgraph_edge *.
4137 * auto-profile.c (afdo_indirect_call): Adjust call to
4138 redirect_call_stmt_to_callee.
4139 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4140 make the this pointer explicit, adjust self-recursive calls and the
4141 call top make_direct. Return the resulting edge.
4142 (cgraph_edge::remove): Make this pointer explicit.
4143 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4144 (cgraph_edge::make_direct): Likewise, adjust call to
4145 resolve_speculation.
4146 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4147 call to set_call_stmt.
4148 (cgraph_update_edges_for_call_stmt_node): Update call to
4149 set_call_stmt and remove.
4150 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4151 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4152 (cgraph_node::create_edge_including_clones): Moved "first" definition
4153 of edge to the block where it was used. Adjusted calls to
4155 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4156 cgraph_edge::remove.
4157 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4158 make_direct and redirect_call_stmt_to_callee.
4159 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4160 resolve_speculation and make_direct.
4161 * ipa-inline-transform.c (inline_transform): Adjust call to
4162 redirect_call_stmt_to_callee.
4163 (check_speculations_1):: Adjust call to resolve_speculation.
4164 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4165 resolve-speculation.
4166 (inline_small_functions): Adjust call to resolve_speculation.
4167 (ipa_inline): Likewise.
4168 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4170 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4171 safe with regards to edge removal, adjust calls to
4172 redirect_call_stmt_to_callee.
4173 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4174 and redirect_call_stmt_to_callee.
4175 * multiple_target.c (create_dispatcher_calls): Adjust call to
4176 redirect_call_stmt_to_callee
4177 (redirect_to_specific_clone): Likewise.
4178 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4179 Adjust calls to cgraph_edge::remove.
4180 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4181 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4182 (expand_call_inline): Adjust call to cgraph_edge::remove.
4184 2020-01-09 Martin Liska <mliska@suse.cz>
4186 * params.opt: Set Optimization for
4187 param_max_speculative_devirt_maydefs.
4189 2020-01-09 Martin Sebor <msebor@redhat.com>
4193 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4195 2020-01-09 Martin Liska <mliska@suse.cz>
4197 * auto-profile.c (auto_profile): Use opt_for_fn
4199 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4200 (propagate_vals_across_arith_jfunc): Likewise.
4201 (hint_time_bonus): Likewise.
4202 (incorporate_penalties): Likewise.
4203 (good_cloning_opportunity_p): Likewise.
4204 (perform_estimation_of_a_value): Likewise.
4205 (estimate_local_effects): Likewise.
4206 (ipcp_propagate_stage): Likewise.
4207 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4208 (set_switch_stmt_execution_predicate): Likewise.
4209 (analyze_function_body): Likewise.
4210 * ipa-inline-analysis.c (offline_size): Likewise.
4211 * ipa-inline.c (early_inliner): Likewise.
4212 * ipa-prop.c (ipa_analyze_node): Likewise.
4213 (ipcp_transform_function): Likewise.
4214 * ipa-sra.c (process_scan_results): Likewise.
4215 (ipa_sra_summarize_function): Likewise.
4216 * params.opt: Rename ipcp-unit-growth to
4217 ipa-cp-unit-growth. Add Optimization for various
4218 IPA-related parameters.
4220 2020-01-09 Richard Biener <rguenther@suse.de>
4223 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4225 2020-01-09 Richard Biener <rguenther@suse.de>
4227 PR tree-optimization/93040
4228 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4230 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4232 * common/config/avr/avr-common.c (avr_option_optimization_table)
4233 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4235 2020-01-09 Martin Liska <mliska@suse.cz>
4237 * cgraphclones.c (symbol_table::materialize_all_clones):
4238 Use cgraph_node::dump_name.
4240 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4243 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4244 output_operand_lossage instead of gcc_unreachable.
4245 * doc/md.texi (riscv f constraint): Fix typo.
4248 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4249 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4250 CONST_SCALAR_INT_P instead of CONST_INT_P.
4251 (*subv<mode>4_1): Rename to ...
4252 (subv<mode>4_1): ... this.
4253 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4254 define_insn_and_split patterns.
4255 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4258 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4260 * vec.c (class selftest::count_dtor): New class.
4261 (selftest::test_auto_delete_vec): New test.
4262 (selftest::vec_c_tests): Call it.
4263 * vec.h (class auto_delete_vec): New class template.
4264 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4266 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4268 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4270 2020-01-08 Jim Wilson <jimw@sifive.com>
4272 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4273 use of TLS_MODEL_LOCAL_EXEC when not pic.
4275 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4277 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4280 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4283 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4284 *stack_protect_set_3 peephole2): Also check that the second
4285 insns source is general_operand.
4288 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4289 predicate for output operand instead of register_operand.
4290 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4291 memory destination and non-memory operands[2].
4293 2020-01-08 Martin Liska <mliska@suse.cz>
4295 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4296 ::dump_asm_name instead of (::name or ::asm_name).
4297 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4298 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4299 (analyze_functions): Likewise.
4300 (expand_all_functions): Likewise.
4301 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4302 (propagate_bits_across_jump_function): Likewise.
4303 (dump_profile_updates): Likewise.
4304 (ipcp_store_bits_results): Likewise.
4305 (ipcp_store_vr_results): Likewise.
4306 * ipa-devirt.c (dump_targets): Likewise.
4307 * ipa-fnsummary.c (analyze_function_body): Likewise.
4308 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4309 (process_hsa_functions): Likewise.
4310 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4311 (set_alias_uids): Likewise.
4312 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4313 * ipa-inline.c (recursive_inlining): Likewise.
4314 (inline_to_all_callers_1): Likewise.
4315 (ipa_inline): Likewise.
4316 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4317 (ipa_propagate_frequency): Likewise.
4318 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4319 (remove_described_reference): Likewise.
4320 * ipa-pure-const.c (worse_state): Likewise.
4321 (check_retval_uses): Likewise.
4322 (analyze_function): Likewise.
4323 (propagate_pure_const): Likewise.
4324 (propagate_nothrow): Likewise.
4325 (dump_malloc_lattice): Likewise.
4326 (propagate_malloc): Likewise.
4327 (pass_local_pure_const::execute): Likewise.
4328 * ipa-visibility.c (optimize_weakref): Likewise.
4329 (function_and_variable_visibility): Likewise.
4330 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4331 (ipa_discover_variable_flags): Likewise.
4332 * lto-streamer-out.c (output_function): Likewise.
4333 (output_constructor): Likewise.
4334 * tree-inline.c (copy_bb): Likewise.
4335 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4336 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4338 2020-01-08 Richard Biener <rguenther@suse.de>
4341 * tree-eh.c (sink_clobbers): Update virtual operands for
4342 the first and last stmt only. Add a dry-run capability.
4343 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4344 after CFG manipulations and in RPO order to catch all
4345 secondary opportunities reliably.
4347 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4350 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4352 2019-01-08 Richard Biener <rguenther@suse.de>
4355 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4356 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4357 virtual operand, also updating SSA use.
4358 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4359 Update stmt after resetting virtual operand.
4360 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4361 * gimple-iterator.c (gsi_remove): When not removing the stmt
4362 permanently do not delink immediate uses or mark the stmt modified.
4364 2020-01-08 Martin Liska <mliska@suse.cz>
4366 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4367 (ipa_call_context::estimate_size_and_time): Likewise.
4368 (inline_analyze_function): Likewise.
4370 2020-01-08 Martin Liska <mliska@suse.cz>
4372 * cgraph.c (cgraph_node::dump): Use systematically
4375 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4377 Add -nodevicespecs option for avr.
4380 * config/avr/avr.opt (-nodevicespecs): New driver option.
4381 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4382 "-specs=device-specs/..." if that option is not set.
4383 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4385 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4387 Implement 64-bit double functions for avr.
4390 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4391 --with-double-comparison.
4392 * doc/install.texi: Document them.
4393 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4394 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4395 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4396 * doc/invoke.texi (AVR Built-in Macros): Document them.
4397 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4398 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4399 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4401 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4404 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4405 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4406 when only building rm-profile multilibs.
4408 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4411 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4412 lattice for a value to check.
4413 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4414 finite propagation in self-recursive scc.
4416 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4418 * ipa-inline.c (caller_growth_limits): Restore the AND.
4420 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4422 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4423 (VEC_ALLREG_ALT): New iterator.
4424 (VEC_ALLREG_INT_MODE): New iterator.
4425 (VCMP_MODE): New iterator.
4426 (VCMP_MODE_INT): New iterator.
4427 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4428 (vec_cmp<u>v64qidi): New define_expand.
4429 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4430 (vec_cmpu<mode>di_exec): New define_expand.
4431 (vec_cmp<u>v64qidi_exec): New define_expand.
4432 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4433 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4434 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4435 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4436 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4437 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4438 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4439 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4440 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4441 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4443 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4444 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4446 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4448 * config/gcn/constraints.md (DA): Update description and match.
4450 (Db): New constraint.
4451 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4453 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4454 Implement 'Db' mixed immediate type.
4455 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4456 (addcv64si3_dup<exec_vcc>): Delete.
4457 (subcv64si3<exec_vcc>): Rework constraints.
4458 (addv64di3): Rework constraints.
4459 (addv64di3_exec): Rework constraints.
4460 (subv64di3): Rework constraints.
4461 (addv64di3_dup): Delete.
4462 (addv64di3_dup_exec): Delete.
4463 (addv64di3_zext): Rework constraints.
4464 (addv64di3_zext_exec): Rework constraints.
4465 (addv64di3_zext_dup): Rework constraints.
4466 (addv64di3_zext_dup_exec): Rework constraints.
4467 (addv64di3_zext_dup2): Rework constraints.
4468 (addv64di3_zext_dup2_exec): Rework constraints.
4469 (addv64di3_sext_dup2): Rework constraints.
4470 (addv64di3_sext_dup2_exec): Rework constraints.
4472 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4474 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4475 existing target checks.
4477 2020-01-07 Richard Biener <rguenther@suse.de>
4479 * doc/install.texi: Bump minimal supported MPC version.
4481 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4483 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4484 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4485 * langhooks.c: Include stor-layout.h.
4486 (lhd_simulate_enum_decl): New function.
4487 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4488 handle_arm_sve_h for the LTO frontend.
4489 (register_vector_type): Cope with null returns from pushdecl.
4491 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4493 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4494 (aarch64_sve::nvectors_if_data_type): Replace with...
4495 (aarch64_sve::builtin_type_p): ...this.
4496 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4497 (find_vector_type): Delete.
4498 (add_sve_type_attribute): New function.
4499 (lookup_sve_type_attribute): Likewise.
4500 (register_builtin_types): Add an "SVE type" attribute to each type.
4501 (register_tuple_type): Likewise.
4502 (svbool_type_p, nvectors_if_data_type): Delete.
4503 (mangle_builtin_type): Use lookup_sve_type_attribute.
4504 (builtin_type_p): Likewise. Add an overload that returns the
4505 number of constituent vector and predicate registers.
4506 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4507 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4508 instead of aarch64_sve_argument_p.
4509 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4510 (aarch64_pass_by_reference): Likewise.
4511 (aarch64_function_value_1): Likewise.
4512 (aarch64_return_in_memory): Likewise.
4513 (aarch64_layout_arg): Likewise.
4515 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4517 PR tree-optimization/93156
4518 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4519 least significant bit is always clear.
4521 PR tree-optimization/93118
4522 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4523 simplifier with two intermediate conversions.
4525 2020-01-07 Martin Liska <mliska@suse.cz>
4527 * params.opt: Add Optimization for various parameters.
4529 2020-01-07 Martin Liska <mliska@suse.cz>
4532 * doc/extend.texi: Explain cloning for target_clone
4535 2020-01-07 Martin Liska <mliska@suse.cz>
4537 PR tree-optimization/92860
4538 * common.opt: Make in Optimization option
4539 as it is affected by -O0, which is an Optimization
4541 * tree-inline.c (tree_inlinable_function_p):
4542 Use opt_for_fn for warn_inline.
4543 (expand_call_inline): Likewise.
4545 2020-01-07 Martin Liska <mliska@suse.cz>
4547 PR tree-optimization/92860
4548 * common.opt: Make flag_ree as optimization
4551 2020-01-07 Martin Liska <mliska@suse.cz>
4553 PR optimization/92860
4554 * params.opt: Mark param_min_crossjump_insns with Optimization
4557 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4559 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4560 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4562 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4564 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4565 helper function to return the valid addressing formats for a given
4566 hard register and mode.
4567 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4569 * config/rs6000/constraints.md (Q constraint): Update
4571 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4574 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4575 Use 'Q' for doing vector extract from memory.
4576 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4578 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4579 doing vector extract from memory.
4580 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4581 extract from memory.
4583 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4584 for the offset being 34-bits when -mcpu=future is used.
4586 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4588 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4589 instead of cmpib_comparison_operator in cmpib patterns.
4590 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4591 of cmpib_comparison_operator. Revise comment.
4593 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4595 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4596 in an IFN_DIV_POW2 node to be equal.
4598 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4600 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4601 (vect_check_scalar_mask): ...this.
4602 (vectorizable_store, vectorizable_load): Update call accordingly.
4603 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4604 argument in calls to conditional internal functions.
4606 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4608 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4609 '0' matching inputs.
4610 (subv64di3_exec): Likewise.
4612 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4614 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4615 * doc/md.texi (movstr): Likewise.
4617 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4619 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4622 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4624 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4626 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4627 to a temporary file and use move-if-change to update the real
4628 file where necessary.
4630 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4632 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4633 rather than Upa for CPY /M.
4635 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4637 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4640 2020-01-06 Martin Liska <mliska@suse.cz>
4642 PR tree-optimization/92860
4643 * params.opt: Mark param_max_combine_insns with Optimization
4646 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4649 * config/i386/i386.md (SWIDWI): New mode iterator.
4650 (DWI, dwi): Add TImode variants.
4651 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4652 <general_hilo_operand> instead of <general_operand>. Use
4653 CONST_SCALAR_INT_P instead of CONST_INT_P.
4654 (*addv<mode>4_1): Rename to ...
4655 (addv<mode>4_1): ... this.
4656 (QWI): New mode attribute.
4657 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4658 define_insn_and_split patterns.
4659 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4661 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4662 <general_hilo_operand> instead of <general_operand>.
4663 (*addcarry<mode>_1): New define_insn.
4664 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4666 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4668 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4669 Use "call" instead of "set".
4671 2020-01-03 Martin Jambor <mjambor@suse.cz>
4674 * ipa-cp.c (print_all_lattices): Skip functions without info.
4676 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4679 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4680 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4681 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4682 for 'e' simd clones.
4685 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4687 (mprefer-vector-width=): Add Save.
4688 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4689 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4690 (ix86_debug_options, ix86_function_specific_print): Adjust
4691 ix86_target_string callers.
4692 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4693 (ix86_valid_target_attribute_tree): Likewise.
4694 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4695 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4696 ix86_target_string caller.
4699 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4700 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4701 instead of gen_int_shift_amount + convert_modes.
4703 PR rtl-optimization/93088
4704 * loop-iv.c (find_single_def_src): Punt after looking through
4705 128 reg copies for regs with single definitions. Move definitions
4708 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4710 * config/arm/arm-c.c (arm_cpu_builtins): Define
4711 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4712 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4713 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4714 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4715 * config/arm/arm-tables.opt: Regenerated.
4716 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4717 arm_arch_i8mm and arm_arch_bf16 when enabled.
4718 * config/arm/arm.h (TARGET_I8MM): New macro.
4719 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4720 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4721 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4722 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4723 (v8_6_a_simd_variants): New.
4724 (v8_*_a_simd_variants): Add i8mm and bf16.
4725 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4727 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4730 * predict.c (compute_function_frequency): Don't call
4731 warn_function_cold on functions that already have cold attribute.
4733 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4736 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4737 COMDAT group function labels in .data.rel.ro.local section.
4738 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4741 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4742 comparison_operator in B and S integer comparisons. Likewise, use
4743 ordered_comparison_operator instead of cmpib_comparison_operator in
4745 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4747 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4749 Update copyright years.
4751 * gcc.c (process_command): Update copyright notice dates.
4752 * gcov-dump.c (print_version): Ditto.
4753 * gcov.c (print_version): Ditto.
4754 * gcov-tool.c (print_version): Ditto.
4755 * gengtype.c (create_file): Ditto.
4756 * doc/cpp.texi: Bump @copying's copyright year.
4757 * doc/cppinternals.texi: Ditto.
4758 * doc/gcc.texi: Ditto.
4759 * doc/gccint.texi: Ditto.
4760 * doc/gcov.texi: Ditto.
4761 * doc/install.texi: Ditto.
4762 * doc/invoke.texi: Ditto.
4764 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4766 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4769 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4771 PR tree-optimization/93098
4772 * match.pd (popcount): For shift amounts, use integer_onep
4773 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4774 tests. Make sure that precision is power of two larger than or equal
4775 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4776 instead of ULL suffixed constants. Formatting fixes.
4778 Copyright (C) 2020 Free Software Foundation, Inc.
4780 Copying and distribution of this file, with or without modification,
4781 are permitted in any medium without royalty provided the copyright
4782 notice and this notice are preserved.