1 2020-02-27 Jakub Jelinek <jakub@redhat.com>
4 * gimplify.c (gimplify_init_constructor): Don't promote readonly
5 DECL_REGISTER variables to TREE_STATIC.
7 PR tree-optimization/93582
8 PR tree-optimization/93945
9 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
10 non-zero INTEGER_CST second argument and ref->offset or ref->size
11 not a multiple of BITS_PER_UNIT.
13 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
15 * doc/install.texi (Binaries): Update description of BullFreeware.
17 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
21 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
22 C++ Language Options, Warning Options, and Static Analyzer
23 Options lists. Document negative form of options enabled by
24 default. Move some things around to more accurately sort
26 (C++ Dialect Options, Warning Options, Static Analyzer
27 Options): Document negative form of options when enabled by
28 default. Move some things around to more accurately sort
29 warnings by category. Add some missing index entries.
32 2020-02-26 Carl Love <cel@us.ibm.com>
35 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
36 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
37 for the vector unsigned short arguments. It is also listed as the
38 name of the built-in for arguments vector unsigned short,
39 vector unsigned int and vector unsigned long long built-ins. The
40 name of the builtins for these arguments should be:
41 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
42 __builtin_crypto_vpmsumd respectively.
44 2020-02-26 Richard Biener <rguenther@suse.de>
46 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
49 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
52 * optabs-tree.c (supportable_convert_operation): Reject types with
55 2020-02-26 David Malcolm <dmalcolm@redhat.com>
57 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
59 2020-02-26 Jakub Jelinek <jakub@redhat.com>
61 PR tree-optimization/93820
62 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
63 argument to ALL_INTEGER_CST_P boolean.
64 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
65 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
66 adjacent INTEGER_CST store into merged_store->only_constants like
69 2020-02-25 Jakub Jelinek <jakub@redhat.com>
72 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
74 * cfghooks.c (verify_flow_info): Likewise.
75 * predict.c (combine_predictions_for_bb): Likewise.
76 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
77 sucessor -> successor.
78 (find_traces_1_round): Fix comment typo, destinarion -> destination.
79 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
81 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
82 message typo, sucessors -> successors.
84 2020-02-25 Martin Sebor <msebor@redhat.com>
86 * doc/extend.texi (attribute access): Correct an example.
88 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
90 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
92 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
94 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
95 (VD): Enable for V4BF.
97 (VQ): Enable for V8BF.
100 (VDBL, Vdbl): Add V4BF.
101 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
102 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
103 (bfloat16x8x2_t): Likewise.
104 (bfloat16x4x3_t): Likewise.
105 (bfloat16x8x3_t): Likewise.
106 (bfloat16x4x4_t): Likewise.
107 (bfloat16x8x4_t): Likewise.
108 (vcombine_bf16): New.
109 (vld1_bf16, vld1_bf16_x2): New.
110 (vld1_bf16_x3, vld1_bf16_x4): New.
111 (vld1q_bf16, vld1q_bf16_x2): New.
112 (vld1q_bf16_x3, vld1q_bf16_x4): New.
113 (vld1_lane_bf16): New.
114 (vld1q_lane_bf16): New.
115 (vld1_dup_bf16): New.
116 (vld1q_dup_bf16): New.
119 (vld2_dup_bf16): New.
120 (vld2q_dup_bf16): New.
123 (vld3_dup_bf16): New.
124 (vld3q_dup_bf16): New.
127 (vld4_dup_bf16): New.
128 (vld4q_dup_bf16): New.
129 (vst1_bf16, vst1_bf16_x2): New.
130 (vst1_bf16_x3, vst1_bf16_x4): New.
131 (vst1q_bf16, vst1q_bf16_x2): New.
132 (vst1q_bf16_x3, vst1q_bf16_x4): New.
133 (vst1_lane_bf16): New.
134 (vst1q_lane_bf16): New.
142 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
144 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
145 (VALL_F16): Likewise.
146 (VALLDI_F16): Likewise.
149 (vswap_width_name): Likewise.
150 (VSWAP_WIDTH): Likewise.
154 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
155 (vget_lane_bf16, vgetq_lane_bf16): New.
157 (vdup_n_bf16, vdupq_n_bf16): New.
158 (vdup_lane_bf16, vdup_laneq_bf16): New.
159 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
160 (vduph_lane_bf16, vduph_laneq_bf16): New.
161 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
162 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
163 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
164 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
165 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
166 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
167 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
168 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
169 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
170 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
171 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
172 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
173 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
174 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
175 (vreinterpretq_bf16_p128): New.
176 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
177 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
178 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
179 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
180 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
181 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
182 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
183 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
184 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
185 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
186 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
187 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
188 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
189 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
190 (vreinterpretq_p128_bf16): New.
192 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
194 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
195 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
196 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
197 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
198 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
199 * config/arm/iterators.md (VSF2BF): New attribute.
200 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
201 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
202 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
204 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
206 * config/arm/arm.md (required_for_purecode): New attribute.
207 (enabled): Handle required_for_purecode.
208 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
209 work with -mpure-code.
211 2020-02-25 Jakub Jelinek <jakub@redhat.com>
213 PR rtl-optimization/93908
214 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
217 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
219 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
221 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
223 * doc/install.texi (--enable-checking): Adjust wording.
225 2020-02-25 Richard Biener <rguenther@suse.de>
227 PR tree-optimization/93868
228 * tree-vect-slp.c (slp_copy_subtree): New function.
229 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
230 re-arranging stmts in it.
232 2020-02-25 Jakub Jelinek <jakub@redhat.com>
235 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
236 dummy function and remove it at the end.
239 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
240 paramter -> parameter.
241 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
242 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
244 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
246 * doc/install.texi (--enable-checking): Properly document current
248 (--enable-stage1-checking): Minor clarification about bootstrap.
250 2020-02-24 David Malcolm <dmalcolm@redhat.com>
253 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
254 -fanalyzer-checker=taint is also required.
255 (-fanalyzer-checker=): Note that providing this option enables the
256 given checker, and doing so may be required for checkers that are
259 2020-02-24 David Malcolm <dmalcolm@redhat.com>
261 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
262 significant control flow events; add a "3" which shows all
263 control flow events; the old "3" becomes "4".
265 2020-02-24 Jakub Jelinek <jakub@redhat.com>
267 PR tree-optimization/93582
268 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
269 pd.offset and pd.size to be counted in bits rather than bytes, add
270 support for maxsizei that is not a multiple of BITS_PER_UNIT and
271 handle bitfield stores and loads.
272 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
273 uncomparable quantities - bytes vs. bits. Allow push_partial_def
274 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
275 pd.offset/pd.size to be counted in bits rather than bytes.
276 Formatting fix. Rename shadowed len variable to buflen.
278 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
279 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
282 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
283 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
284 * opts-common.c (parse_options_from_collect_gcc_options): New function.
285 (prepend_xassembler_to_collect_as_options): Likewise.
286 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
287 (prepend_xassembler_to_collect_as_options): Likewise.
288 * lto-opts.c (lto_write_options): Stream assembler options
289 in COLLECT_AS_OPTIONS.
290 * lto-wrapper.c (xassembler_options_error): New static variable.
291 (get_options_from_collect_gcc_options): Move parsing options code to
292 parse_options_from_collect_gcc_options and call it.
293 (merge_and_complain): Validate -Xassembler options.
294 (append_compiler_options): Handle OPT_Xassembler.
295 (run_gcc): Append command line -Xassembler options to
297 * doc/invoke.texi: Add documentation about using Xassembler
300 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
302 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
304 (riscv_rtx_costs): Update cost model for LTGT.
306 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
308 PR rtl-optimization/93564
309 * ira-color.c (struct update_cost_queue_elem): New member start.
310 (queue_update_cost, get_next_update_cost): Add new arg start.
311 (allocnos_conflict_p): New function.
312 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
313 Add checking conflicts with allocnos_conflict_p.
314 (update_costs_from_prefs, restore_costs_from_copies): Adjust
315 update_costs_from_allocno calls.
316 (update_conflict_hard_regno_costs): Add checking conflicts with
317 allocnos_conflict_p. Adjust calls of queue_update_cost and
318 get_next_update_cost.
319 (assign_hard_reg): Adjust calls of queue_update_cost. Add
321 (bucket_allocno_compare_func): Restore previous version.
323 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
325 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
326 double-word size when handling aggregate return values.
327 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
328 that homogeneous SFmode and DFmode aggregates are passed and returned
329 in general registers.
331 2020-02-21 Jakub Jelinek <jakub@redhat.com>
334 * opts.c (print_filtered_help): Translate help before appending
335 messages to it rather than after that.
337 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
339 PR rtl-optimization/PR92989
340 * lra-lives.c (process_bb_lives): Restore the original order
341 of the bb liveness update. Call make_hard_regno_dead for each
342 register clobbered at the start of an EH receiver.
344 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
347 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
348 self-recursively generated.
350 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
353 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
356 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
358 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
359 Document new target supports option.
361 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
363 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
364 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
365 * config/arm/iterators.md (MATMUL): New iterator.
366 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
367 (mmla_sfx): New attribute.
368 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
369 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
370 (UNSPEC_MATMUL_US): New.
372 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
374 * config/arm/arm.md: Prevent scalar shifts from being used when big
377 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
378 Richard Biener <rguenther@suse.de>
380 PR tree-optimization/93586
381 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
382 after mismatched array refs; do not sure type size information to
383 recover from unmatched referneces with !flag_strict_aliasing_p.
385 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
387 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
388 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
389 (scatter_store<mode>): Rename to ...
390 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
391 (scatter<mode>_exec): Delete. Move contents ...
392 (mask_scatter_store<mode>): ... here, and rename that to ...
393 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
394 Remove mode conversion.
395 (mask_gather_load<mode>): Rename to ...
396 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
397 Remove mode conversion.
398 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
400 2020-02-21 Martin Jambor <mjambor@suse.cz>
402 PR tree-optimization/93845
403 * tree-sra.c (verify_sra_access_forest): Only test access size of
406 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
408 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
409 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
410 (addv64di3_exec): Likewise.
411 (subv64di3): Likewise.
412 (subv64di3_exec): Likewise.
413 (addv64di3_zext): Likewise.
414 (addv64di3_zext_exec): Likewise.
415 (addv64di3_zext_dup): Likewise.
416 (addv64di3_zext_dup_exec): Likewise.
417 (addv64di3_zext_dup2): Likewise.
418 (addv64di3_zext_dup2_exec): Likewise.
419 (addv64di3_sext_dup2): Likewise.
420 (addv64di3_sext_dup2_exec): Likewise.
421 (<expander>v64di3): Likewise.
422 (<expander>v64di3_exec): Likewise.
423 (*<reduc_op>_dpp_shr_v64di): Likewise.
424 (*plus_carry_dpp_shr_v64di): Likewise.
425 * config/gcn/gcn.md (adddi3): Likewise.
426 (addptrdi3): Likewise.
427 (<expander>di3): Likewise.
429 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
431 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
433 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
435 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
436 support. Use aarch64_emit_mult instead of emitting multiplication
437 instructions directly.
438 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
439 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
441 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
443 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
444 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
445 instead of emitting multiplication instructions directly.
446 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
447 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
448 (@aarch64_frecps<mode>): New expanders.
450 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
452 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
453 on and produce uint64_ts rather than ints.
454 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
455 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
457 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
459 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
460 an unused xmsk register when handling approximate rsqrt.
462 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
464 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
465 flag_finite_math_only condition.
467 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
470 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
471 to destination operand for shufps alternative.
472 (*vec_extractv2si_1): Ditto.
474 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
477 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
480 2020-02-20 Martin Liska <mliska@suse.cz>
483 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
485 2020-02-20 Martin Liska <mliska@suse.cz>
488 * common/config/avr/avr-common.c: Remote trailing "|".
490 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
492 * collect2.c (maybe_run_lto_and_relink): Fix typo in
495 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
497 PR tree-optimization/93767
498 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
499 access-size bias from the offset calculations for negative strides.
501 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
503 * collect2.c (c_file, o_file): Make const again.
504 (ldout,lderrout, dump_ld_file): Remove.
505 (tool_cleanup): Avoid calling not signal-safe functions.
506 (maybe_run_lto_and_relink): Avoid possible signal handler
507 access to unintialzed memory (lto_o_files).
508 (main): Avoid leaking temp files in $TMPDIR.
509 Initialize c_file/o_file with concat, which avoids exposing
510 uninitialized memory to signal handler, which calls unlink(!).
511 Avoid calling maybe_unlink when the main function returns,
512 since the atexit handler is already doing this.
513 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
515 2020-02-19 Martin Jambor <mjambor@suse.cz>
517 PR tree-optimization/93776
518 * tree-sra.c (create_access): Do not create zero size accesses.
519 (get_access_for_expr): Do not search for zero sized accesses.
521 2020-02-19 Martin Jambor <mjambor@suse.cz>
523 PR tree-optimization/93667
524 * tree-sra.c (scalarizable_type_p): Return false if record fields
525 do not follow wach other.
527 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
529 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
530 rather than fmv.x.s/fmv.s.x.
532 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
534 * config/aarch64/aarch64-simd-builtins.def
535 (intrinsic_vec_smult_lo_): New.
536 (intrinsic_vec_umult_lo_): Likewise.
537 (vec_widen_smult_hi_): Likewise.
538 (vec_widen_umult_hi_): Likewise.
539 * config/aarch64/aarch64-simd.md
540 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
541 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
542 (vmull_high_s16): Likewise.
543 (vmull_high_s32): Likewise.
544 (vmull_high_u8): Likewise.
545 (vmull_high_u16): Likewise.
546 (vmull_high_u32): Likewise.
547 (vmull_s8): Likewise.
548 (vmull_s16): Likewise.
549 (vmull_s32): Likewise.
550 (vmull_u8): Likewise.
551 (vmull_u16): Likewise.
552 (vmull_u32): Likewise.
554 2020-02-18 Martin Liska <mliska@suse.cz>
556 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
557 bootstrap by missing removal of invalid sanity check.
559 2020-02-18 Martin Liska <mliska@suse.cz>
562 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
563 Always compare LHS of gimple_assign.
565 2020-02-18 Martin Liska <mliska@suse.cz>
568 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
569 and return type of functions.
570 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
571 Drop MALLOC attribute for void functions.
572 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
573 malloc_state for a new VOID clone.
575 2020-02-18 Martin Liska <mliska@suse.cz>
578 * common.opt: Add -fprofile-reproducibility.
579 * doc/invoke.texi: Document it.
580 * value-prof.c (dump_histogram_value):
581 Document and support behavior for counters[0]
582 being a negative value.
583 (get_nth_most_common_value): Handle negative
584 counters[0] in respect to flag_profile_reproducible.
586 2020-02-18 Jakub Jelinek <jakub@redhat.com>
589 * cgraph.c (verify_speculative_call): Use speculative_id instead of
590 speculative_uid in messages. Remove trailing whitespace from error
591 message. Use num_speculative_call_targets instead of
592 num_speculative_targets in a message.
593 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
594 edge messages and stmt instead of cal_stmt in reference message.
596 PR tree-optimization/93780
597 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
598 before calling build_vector_type.
599 (execute_update_addresses_taken): Likewise.
602 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
603 typo, functoin -> function.
604 * tree.c (free_lang_data_in_decl): Fix comment typo,
605 functoin -> function.
606 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
608 2020-02-17 David Malcolm <dmalcolm@redhat.com>
610 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
612 (print_option_information): Don't call get_option_url if URLs
615 2020-02-17 Alexandre Oliva <oliva@adacore.com>
617 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
618 handling of register_common-less targets.
620 2020-02-17 Martin Liska <mliska@suse.cz>
623 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
625 2020-02-17 Martin Liska <mliska@suse.cz>
628 * config/rs6000/rs6000.c (rs6000_option_override_internal):
631 2020-02-17 Martin Liska <mliska@suse.cz>
634 * config/rx/elf.opt: Fix typo.
636 2020-02-17 Richard Biener <rguenther@suse.de>
639 * opts-global.c (print_ignored_options): Use inform and
642 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
645 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
647 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
650 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
651 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
653 2020-02-15 Jason Merrill <jason@redhat.com>
655 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
657 2020-02-15 Jakub Jelinek <jakub@redhat.com>
659 PR tree-optimization/93744
660 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
661 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
662 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
663 sure @2 in the first and @1 in the other patterns has no side-effects.
665 2020-02-15 David Malcolm <dmalcolm@redhat.com>
666 Bernd Edlinger <bernd.edlinger@hotmail.de>
670 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
671 * configure.ac (--with-diagnostics-urls): New configuration
672 option, based on --with-diagnostics-color.
673 (DIAGNOSTICS_URLS_DEFAULT): New define.
674 * config.h: Regenerate.
675 * configure: Regenerate.
676 * diagnostic.c (diagnostic_urls_init): Handle -1 for
677 DIAGNOSTICS_URLS_DEFAULT from configure-time
678 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
679 and TERM_URLS environment variable.
680 * diagnostic-url.h (diagnostic_url_format): New enum type.
681 (diagnostic_urls_enabled_p): rename to...
682 (determine_url_format): ... this, and change return type.
683 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
684 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
685 the linux console, and mingw.
686 (diagnostic_urls_enabled_p): rename to...
687 (determine_url_format): ... this, and adjust.
688 * pretty-print.h (pretty_printer::show_urls): rename to...
689 (pretty_printer::url_format): ... this, and change to enum.
690 * pretty-print.c (pretty_printer::pretty_printer,
691 pp_begin_url, pp_end_url, test_urls): Adjust.
692 * doc/install.texi (--with-diagnostics-urls): Document the new
693 configuration option.
694 (--with-diagnostics-color): Document the existing interaction
695 with GCC_COLORS better.
696 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
697 vindex reference. Update description of defaults based on the above.
698 (-fdiagnostics-color): Update description of how -fdiagnostics-color
699 interacts with GCC_COLORS.
701 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
704 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
705 conjunction with TARGET_GNU_TLS in early return.
707 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
709 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
710 the mode is not wider than UNITS_PER_WORD.
712 2020-02-14 Martin Jambor <mjambor@suse.cz>
714 PR tree-optimization/93516
715 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
716 access of the same type as the parent.
717 (propagate_subaccesses_from_lhs): Likewise.
719 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
722 * config/i386/avx512vbmi2intrin.h
723 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
724 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
725 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
726 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
727 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
728 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
729 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
730 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
731 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
732 of lacking a closing parenthesis.
733 * config/i386/avx512vbmi2vlintrin.h
734 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
735 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
736 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
737 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
738 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
739 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
740 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
741 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
742 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
743 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
744 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
745 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
746 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
747 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
748 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
749 _mm_shldi_epi32, _mm_mask_shldi_epi32,
750 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
751 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
753 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
756 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
757 the target function entry.
759 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
761 * common/config/arc/arc-common.c (arc_option_optimization_table):
762 Disable if-conversion step when optimized for size.
764 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
766 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
767 R12-R15 are always in ARCOMPACT16_REGS register class.
768 * config/arc/arc.opt (mq-class): Deprecate.
769 * config/arc/constraint.md ("q"): Remove dependency on mq-class
771 * doc/invoke.texi (mq-class): Update text.
772 * common/config/arc/arc-common.c (arc_option_optimization_table):
775 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
777 * config/arc/arc.c (arc_insn_cost): New function.
778 (TARGET_INSN_COST): Define.
779 * config/arc/arc.md (cost): New attribute.
780 (add_n): Use arc_nonmemory_operand.
781 (ashlsi3_insn): Likewise, also update constraints.
782 (ashrsi3_insn): Likewise.
784 (add_shift): Likewise.
785 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
787 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
789 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
791 (umulsidi_600): Likewise.
793 2020-02-13 Jakub Jelinek <jakub@redhat.com>
796 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
797 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
798 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
799 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
800 pass __A to the builtin followed by __W instead of __A followed by
802 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
803 _mm512_mask_popcnt_epi64): Likewise.
804 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
805 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
806 _mm256_mask_popcnt_epi64): Likewise.
808 PR tree-optimization/93582
809 * fold-const.h (shift_bytes_in_array_left,
810 shift_bytes_in_array_right): Declare.
811 * fold-const.c (shift_bytes_in_array_left,
812 shift_bytes_in_array_right): New function, moved from
813 gimple-ssa-store-merging.c, no longer static.
814 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
815 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
816 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
817 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
818 shift_bytes_in_array.
819 (verify_shift_bytes_in_array): Rename to ...
820 (verify_shift_bytes_in_array_left): ... this. Use
821 shift_bytes_in_array_left instead of shift_bytes_in_array.
822 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
823 instead of verify_shift_bytes_in_array.
824 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
825 / native_interpret_expr where the store covers all needed bits,
826 punt on PDP-endian, otherwise allow all involved offsets and sizes
827 not to be byte-aligned.
830 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
831 use const_0_to_255_operand predicate instead of immediate_operand.
832 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
833 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
834 vgf2p8affineinvqb_<mode><mask_name>,
835 vgf2p8affineqb_<mode><mask_name>): Drop mode from
836 const_0_to_255_operand predicated operands.
838 2020-02-12 Jeff Law <law@redhat.com>
840 * config/h8300/h8300.md (comparison shortening peepholes): Use
841 a mode iterator to merge the HImode and SImode peepholes.
843 2020-02-12 Jakub Jelinek <jakub@redhat.com>
846 * real.c (is_even): Make static. Function comment fix.
847 (is_halfway_below): Make static, don't assert R is not inf/nan,
848 instead return false for those. Small formatting fixes.
850 2020-02-12 Martin Sebor <msebor@redhat.com>
853 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
854 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
855 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
856 (strlen_check_and_optimize_call): Adjust callee name.
858 2020-02-12 Jeff Law <law@redhat.com>
860 * config/h8300/h8300.md (comparison shortening peepholes): Drop
861 (and (xor)) variant. Combine other two into single peephole.
863 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
865 PR rtl-optimization/93565
866 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
868 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
870 * config/aarch64/aarch64-simd.md
871 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
872 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
873 generating separate ADDV and zero_extend patterns.
874 * config/aarch64/iterators.md (VDQV_E): New iterator.
876 2020-02-12 Jeff Law <law@redhat.com>
878 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
879 expanders, splits, etc.
880 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
881 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
882 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
883 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
884 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
885 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
887 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
889 2020-02-12 Jakub Jelinek <jakub@redhat.com>
892 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
893 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
894 TARGET_AVX512DQ from condition.
895 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
896 instead of <mask_mode512bit_condition> in condition. If
897 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
899 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
902 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
905 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
907 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
909 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
910 where strlen is more legible.
911 (rs6000_builtin_vectorized_libmass): Ditto.
912 (rs6000_print_options_internal): Ditto.
914 2020-02-11 Martin Sebor <msebor@redhat.com>
916 PR tree-optimization/93683
917 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
919 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
921 * config/rs6000/predicates.md (cint34_operand): Rename the
922 -mprefixed-addr option to be -mprefixed.
923 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
924 the -mprefixed-addr option to be -mprefixed.
925 (OTHER_FUTURE_MASKS): Likewise.
926 (POWERPC_MASKS): Likewise.
927 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
928 the -mprefixed-addr option to be -mprefixed. Change error
929 messages to refer to -mprefixed.
930 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
932 (rs6000_legitimate_offset_address_p): Likewise.
933 (rs6000_mode_dependent_address): Likewise.
934 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
935 "-mprefixed" for target attributes and pragmas.
936 (address_to_insn_form): Rename the -mprefixed-addr option to be
938 (rs6000_adjust_insn_length): Likewise.
939 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
940 -mprefixed-addr option to be -mprefixed.
941 (ASM_OUTPUT_OPCODE): Likewise.
942 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
943 -mprefixed-addr option to be -mprefixed.
944 * config/rs6000/rs6000.opt (-mprefixed): Rename the
945 -mprefixed-addr option to be prefixed. Change the option from
946 being undocumented to being documented.
947 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
948 -mprefixed option. Update the -mpcrel documentation to mention
951 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
953 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
954 including FIRST_PSEUDO_REGISTER - 1.
955 * ira-color.c (print_hard_reg_set): Ditto.
957 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
959 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
960 (USTERNOP_QUALIFIERS): New define.
961 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
962 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
963 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
964 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
965 * config/arm/arm_neon.h (vusdot_s32): New.
966 (vusdot_lane_s32): New.
967 (vusdotq_lane_s32): New.
968 (vsudot_lane_s32): New.
969 (vsudotq_lane_s32): New.
970 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
971 * config/arm/iterators.md (DOTPROD_I8MM): New.
972 (sup, opsuffix): Add <us/su>.
973 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
974 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
976 2020-02-11 Richard Biener <rguenther@suse.de>
978 PR tree-optimization/93661
979 PR tree-optimization/93662
980 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
982 * tree-sra.c (get_access_for_expr): Likewise.
984 2020-02-10 Jakub Jelinek <jakub@redhat.com>
987 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
988 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
989 Change condition from TARGET_AVX2 to TARGET_AVX.
991 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
994 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
997 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
999 Try to generate zero-based comparisons.
1000 * config/cris/cris.c (cris_reduce_compare): New function.
1001 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1002 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1003 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1005 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1008 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1009 in Thumb state and also as a destination in Arm state. Add T16
1012 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1014 * md.texi (Define Subst): Match closing paren in example.
1016 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1020 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1021 arguments of strncmp.
1023 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1026 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1027 but different source value.
1028 (adjust_callers_for_value_intersection): New function.
1029 (gather_edges_for_value): Adjust order of callers to let a
1030 non-self-recursive caller be the first element.
1031 (self_recursive_pass_through_p): Add a new parameter "simple", and
1032 check generalized self-recursive pass-through jump function.
1033 (self_recursive_agg_pass_through_p): Likewise.
1034 (find_more_scalar_values_for_callers_subset): Compute value from
1035 pass-through jump function for self-recursive.
1036 (intersect_with_plats): Cleanup previous implementation code for value
1037 itersection with self-recursive call edge.
1038 (intersect_with_agg_replacements): Likewise.
1039 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1040 function for self-recursive call edge. Cleanup previous implementation
1041 code for value intersection with self-recursive call edge.
1042 (decide_whether_version_node): Remove dead callers and adjust order
1043 to let a non-self-recursive caller be the first element.
1045 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1047 * recog.c: Move pass_split_before_sched2 code in front of
1048 pass_split_before_regstack.
1049 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1050 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1051 (rest_of_handle_split_before_sched2): Remove.
1052 (pass_split_before_sched2::execute): Unconditionally call
1054 (enable_split_before_sched2): New function.
1055 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1056 (pass_split_before_regstack::gate): Ditto.
1057 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1058 Update name check for renamed split4 pass.
1059 * config/sh/sh.c (register_sh_passes): Update pass insertion
1060 point for renamed split4 pass.
1062 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1064 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1065 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1066 copying them around between host and target.
1068 2020-02-08 Andrew Pinski <apinski@marvell.com>
1071 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1072 STRICT_ALIGNMENT also.
1074 2020-02-08 Jim Wilson <jimw@sifive.com>
1077 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1079 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1080 Jakub Jelinek <jakub@redhat.com>
1083 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1084 xmm16-xmm31 call-used even in 64-bit ms-abi.
1086 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1088 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1089 (simd_ummla, simd_usmmla): Likewise.
1090 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1091 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1092 (vusmmlaq_s32): New.
1094 2020-02-07 Richard Biener <rguenther@suse.de>
1097 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1098 skipping unreachable regions.
1099 (optimize_inline_calls): Skip folding stmts when we didn't
1102 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1105 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1106 Don't return aggregates with only SFmode and DFmode in SSE
1108 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1110 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1113 * config/rs6000/rs6000-logue.c
1114 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1115 if it fails, move rs into end_addr and retry. Add
1116 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1117 the insn pattern doesn't describe well what exactly happens to
1121 * config/i386/predicates.md (avx_identity_operand): Remove.
1122 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1123 (avx_<castmode><avxsizesuffix>_<castmode>,
1124 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1125 a VEC_CONCAT of the operand and UNSPEC_CAST.
1126 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1127 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1131 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1132 recog_data.insn if distance_non_agu_define changed it.
1134 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1137 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1138 we only had X-FORM (reg+reg) addressing for vectors. Also before
1139 ISA 3.0, we only had X-FORM addressing for scalars in the
1140 traditional Altivec registers.
1142 2020-02-06 <zhongyunde@huawei.com>
1143 Vladimir Makarov <vmakarov@redhat.com>
1145 PR rtl-optimization/93561
1146 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1147 hard register range.
1149 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1151 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1154 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1156 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1157 where the low and the high 32 bits are equal to each other specially,
1158 with an rldimi instruction.
1160 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1162 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1164 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1166 * config/arm/arm-tables.opt: Regenerate.
1168 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1171 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1172 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1173 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1175 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1177 PR rtl-optimization/87763
1178 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1180 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1182 * config/aarch64/aarch64-simd-builtins.def
1183 (bfmlaq): New built-in function.
1184 (bfmlalb): New built-in function.
1185 (bfmlalt): New built-in function.
1186 (bfmlalb_lane): New built-in function.
1187 (bfmlalt_lane): New built-in function.
1188 * config/aarch64/aarch64-simd.md
1189 (aarch64_bfmmlaqv4sf): New pattern.
1190 (aarch64_bfmlal<bt>v4sf): New pattern.
1191 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1192 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1193 (vbfmlalbq_f32): New intrinsic.
1194 (vbfmlaltq_f32): New intrinsic.
1195 (vbfmlalbq_lane_f32): New intrinsic.
1196 (vbfmlaltq_lane_f32): New intrinsic.
1197 (vbfmlalbq_laneq_f32): New intrinsic.
1198 (vbfmlaltq_laneq_f32): New intrinsic.
1199 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1200 (bt): New int attribute.
1202 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1204 * config/i386/i386.md (*pushtf): Emit "#" instead of
1205 calling gcc_unreachable in insn output.
1208 (*pushsf_rex64): Ditto for alternatives other than 1.
1209 (*pushsf): Ditto for alternatives other than 1.
1211 2020-02-06 Martin Liska <mliska@suse.cz>
1213 PR gcov-profile/91971
1214 PR gcov-profile/93466
1215 * coverage.c (coverage_init): Revert mangling of
1216 path into filename. It can lead to huge filename length.
1217 Creation of subfolders seem more natural.
1219 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1222 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1223 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1224 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1226 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1229 * config/i386/predicates.md (avx_identity_operand): New predicate.
1230 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1231 define_insn_and_split.
1234 * omp-low.c (use_pointer_for_field): For nested constructs, also
1235 look for map clauses on target construct.
1236 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1237 taskreg_nesting_level.
1240 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1241 shared clause, call omp_notice_variable on outer context if any.
1243 2020-02-05 Jason Merrill <jason@redhat.com>
1246 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1247 non-zero address even if weak and not yet defined.
1249 2020-02-05 Martin Sebor <msebor@redhat.com>
1251 PR tree-optimization/92765
1252 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1253 * tree-ssa-strlen.c (compute_string_length): Remove.
1254 (determine_min_objsize): Remove.
1255 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1256 Avoid using type size as the upper bound on string length.
1257 (handle_builtin_string_cmp): Add an argument. Adjust.
1258 (strlen_check_and_optimize_call): Pass additional argument to
1259 handle_builtin_string_cmp.
1261 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1263 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1264 (*pushdi2_rex64 peephole2): Unconditionally split after
1266 (*ashl<mode>3_doubleword): Ditto.
1267 (*<shift_insn><mode>3_doubleword): Ditto.
1269 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1272 * config/rs6000/rs6000.c (get_vector_offset): Fix
1274 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1276 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1278 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1281 (Special Functions for Debugging the Analyzer): Update description
1282 of __analyzer_dump_exploded_nodes.
1284 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1287 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1288 include sets and not clobbers in the vzeroupper pattern.
1289 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1290 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1291 (*avx_vzeroupper_1): New define_insn_and_split.
1294 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1295 don't run when !optimize.
1296 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1299 2020-02-05 Richard Biener <rguenther@suse.de>
1302 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1303 checks before matching calls.
1305 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1307 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1308 function comment typo.
1311 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1312 simd_clone_create failed when i == 0, adjust clone->nargs by
1315 2020-02-05 Martin Liska <mliska@suse.cz>
1318 * doc/invoke.texi: Document that one should
1319 not combine ASLR and -fpch.
1321 2020-02-04 Richard Biener <rguenther@suse.de>
1323 PR tree-optimization/93538
1324 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1326 2020-02-04 Richard Biener <rguenther@suse.de>
1328 PR tree-optimization/91123
1329 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1330 (vn_walk_cb_data::last_vuse): New member.
1331 (vn_walk_cb_data::saved_operands): Likewsie.
1332 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1333 (vn_walk_cb_data::push_partial_def): Use finish.
1334 (vn_reference_lookup_2): Update last_vuse and use finish if
1335 we've saved operands.
1336 (vn_reference_lookup_3): Use finish and update calls to
1337 push_partial_defs everywhere. When translating through
1338 memcpy or aggregate copies save off operands and alias-set.
1339 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1340 operation for redundant store removal.
1342 2020-02-04 Richard Biener <rguenther@suse.de>
1344 PR tree-optimization/92819
1345 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1346 generating more stmts than before.
1348 2020-02-04 Martin Liska <mliska@suse.cz>
1350 * config/arm/arm.c (arm_gen_far_branch): Move the function
1351 outside of selftests.
1353 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1355 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1356 function to adjust PC-relative vector addresses.
1357 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1358 handle vectors with PC-relative addresses.
1360 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1362 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1364 (hard_reg_and_mode_to_addr_mask): Delete.
1365 (rs6000_adjust_vec_address): If the original vector address
1366 was REG+REG or REG+OFFSET and the element is not zero, do the add
1367 of the elements in the original address before adding the offset
1368 for the vector element. Use address_to_insn_form to validate the
1369 address using the register being loaded, rather than guessing
1370 whether the address is a DS-FORM or DQ-FORM address.
1372 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1374 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1375 to calculate the offset in memory from the start of a vector of a
1376 particular element. Add code to keep the element number in
1377 bounds if the element number is variable.
1378 (rs6000_adjust_vec_address): Move calculation of offset of the
1379 vector element to get_vector_offset.
1380 (rs6000_split_vec_extract_var): Do not do the initial AND of
1381 element here, move the code to get_vector_offset.
1383 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1385 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1388 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1390 * config/rs6000/constraints.md: Improve documentation.
1392 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1395 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1396 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1398 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1400 * config.gcc: Remove "carrizo" support.
1401 * config/gcn/gcn-opts.h (processor_type): Likewise.
1402 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1403 * config/gcn/gcn.opt (gpu_type): Likewise.
1404 * config/gcn/t-omp-device: Likewise.
1406 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1409 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1410 * config/arm/arm.c (arm_gen_far_branch): New function
1412 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1414 2020-02-03 Julian Brown <julian@codesourcery.com>
1415 Tobias Burnus <tobias@codesourcery.com>
1417 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1419 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1422 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1423 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1426 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1428 PR rtl-optimization/91333
1429 * ira-color.c (struct allocno_color_data): Add member
1431 (init_allocno_threads): Set the member up.
1432 (bucket_allocno_compare_func): Add compare hard reg
1435 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1437 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1439 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1440 * config.in: Regenerated.
1441 * configure: Regenerated.
1442 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1443 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1444 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1446 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1448 * configure: Regenerate.
1450 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1452 PR rtl-optimization/91333
1453 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1454 reg preferences comparison up.
1456 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1458 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1459 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1460 aarch64-sve-builtins-base.h.
1461 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1462 aarch64-sve-builtins-base.cc.
1463 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1464 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1466 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1467 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1468 (svcvtnt): New functions.
1469 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1470 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1471 (svcvtnt): New functions.
1472 (svcvt): Add a form that converts f32 to bf16.
1473 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1474 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1476 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1477 Treat B as bfloat16_t.
1478 (ternary_bfloat_lane_base): New class.
1479 (ternary_bfloat_def): Likewise.
1480 (ternary_bfloat): New shape.
1481 (ternary_bfloat_lane_def): New class.
1482 (ternary_bfloat_lane): New shape.
1483 (ternary_bfloat_lanex2_def): New class.
1484 (ternary_bfloat_lanex2): New shape.
1485 (ternary_bfloat_opt_n_def): New class.
1486 (ternary_bfloat_opt_n): New shape.
1487 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1488 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1489 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1490 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1491 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1492 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1493 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1494 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1495 the pattern off the narrow mode instead of the wider one.
1496 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1497 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1498 (sve_fp_op): Handle them.
1499 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1500 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1502 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1504 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1505 * config/aarch64/aarch64-modes.def (BF): Move definition before
1506 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1507 (SVE_MODES): Handle BF modes.
1508 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1510 (aarch64_full_sve_mode): Likewise.
1511 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1513 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1514 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1515 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1516 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1518 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1520 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1521 (TYPES_all_data): Add bf16.
1522 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1523 (register_tuple_type): Increase buffer size.
1524 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1525 (bf16): New type suffix.
1526 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1527 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1528 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1529 Change type from all_data to all_arith.
1530 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1533 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1534 Matthew Malcomson <matthew.malcomson@arm.com>
1535 Richard Sandiford <richard.sandiford@arm.com>
1537 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1538 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1539 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1540 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1541 __ARM_FEATURE_MATMUL_FP64.
1542 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1543 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1544 be disabled at the same time.
1545 (f32mm): New extension.
1546 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1547 (AARCH64_FL_F64MM): Bump to the next bit up.
1548 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1549 (TARGET_SVE_F64MM): New macros.
1550 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1551 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1552 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1553 (UNSPEC_ZIP2Q): New unspeccs.
1554 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1555 (optab, sur, perm_insn): Handle the new unspecs.
1556 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1557 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1558 TARGET_SVE_F64MM instead of separate tests.
1559 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1560 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1561 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1562 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1563 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1564 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1565 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1566 (TYPES_s_signed): New macro.
1567 (TYPES_s_integer): Use it.
1568 (TYPES_d_float): New macro.
1569 (TYPES_d_data): Use it.
1570 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1571 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1572 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1573 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1574 (svmmla): New shape.
1575 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1576 template parameters.
1577 (ternary_resize2_lane_base): Likewise.
1578 (ternary_resize2_base): New class.
1579 (ternary_qq_lane_base): Likewise.
1580 (ternary_intq_uintq_lane_def): Likewise.
1581 (ternary_intq_uintq_lane): New shape.
1582 (ternary_intq_uintq_opt_n_def): New class
1583 (ternary_intq_uintq_opt_n): New shape.
1584 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1585 (ternary_uintq_intq_def): New class.
1586 (ternary_uintq_intq): New shape.
1587 (ternary_uintq_intq_lane_def): New class.
1588 (ternary_uintq_intq_lane): New shape.
1589 (ternary_uintq_intq_opt_n_def): New class.
1590 (ternary_uintq_intq_opt_n): New shape.
1591 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1592 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1593 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1594 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1596 (svdotprod_lane_impl): ...this new class.
1597 (svmmla_impl, svusdot_impl): New classes.
1598 (svdot_lane): Update to use svdotprod_lane_impl.
1599 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1600 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1602 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1603 function, with no types defined.
1604 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1605 AARCH64_FL_I8MM functions.
1606 (svmmla): New AARCH64_FL_F32MM function.
1607 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1608 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1609 AARCH64_FL_F64MM function.
1610 (REQUIRED_EXTENSIONS):
1612 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1614 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1617 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1619 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1620 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1621 (*movti_internal): Do not check for
1622 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1623 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1624 just after check for TARGET_AVX.
1625 (*movdf_internal): Ditto.
1626 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1627 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1628 * config/i386/sse.md (mov<mode>_internal): Only check
1629 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1630 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1631 (<sse>_andnot<mode>3<mask_name>): Move check for
1632 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1633 (<code><mode>3<mask_name>): Ditto.
1634 (*andnot<mode>3): Ditto.
1635 (*andnottf3): Ditto.
1636 (*<code><mode>3): Ditto.
1637 (*<code>tf3): Ditto.
1638 (*andnot<VI:mode>3): Remove
1639 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1640 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1641 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1642 (sse4_1_blendv<ssemodesuffix>): Ditto.
1643 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1644 Explain that tune applies to 128bit instructions only.
1646 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1648 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1649 to definition of hsa_kernel_description. Parse assembly to find SGPR
1650 and VGPR count of kernel and store in hsa_kernel_description.
1652 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1654 PR rtl-optimization/91838
1655 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1656 to truncate if allowed or reject combination.
1658 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1660 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1661 (find_inv_vars_cb): Likewise.
1663 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1665 * calls.c (special_function_p): Split out the check for DECL_NAME
1666 being non-NULL and fndecl being extern at file scope into a
1667 new maybe_special_function_p and call it. Drop check for fndecl
1668 being non-NULL that was after a usage of DECL_NAME (fndecl).
1669 * tree.h (maybe_special_function_p): New inline function.
1671 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1673 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1674 (mask_gather_load<mode>): ... here, and zero-initialize the
1676 (maskload<mode>di): Zero-initialize the destination.
1679 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1682 * doc/analyzer.texi (Limitations): Note that constraints on
1683 floating-point values are currently ignored.
1685 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1688 * symtab.c (symtab_node::noninterposable_alias): If localalias
1689 already exists, but is not usable, append numbers after it until
1690 a unique name is found. Formatting fix.
1693 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1696 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1698 * config/gcn/gcn.c (print_operand): Handle LTGT.
1699 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1701 2020-01-30 Richard Biener <rguenther@suse.de>
1703 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1704 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1706 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1708 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1709 without a DECL in .data.rel.ro.local.
1711 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1714 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1718 * config/i386/sse.md
1719 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1720 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1721 any_extend code iterator instead of always zero_extend.
1722 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1723 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1724 Use any_extend code iterator instead of always zero_extend.
1725 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1726 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1727 Use any_extend code iterator instead of always zero_extend.
1728 (*sse2_pmovmskb_ext): New define_insn.
1729 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1732 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1733 (*popcountsi2_zext_falsedep): New define_insn.
1735 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1737 * config.in: Regenerated.
1738 * configure: Regenerated.
1740 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1743 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1744 LLVM's assembler changed the default in version 9.
1746 2020-01-24 Jeff Law <law@redhat.com>
1748 PR tree-optimization/89689
1749 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1751 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1755 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1757 PR rtl-optimization/87763
1758 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1759 simplification to handle subregs as well as bare regs.
1760 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1762 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1765 * ira.c (ira): Revert use of simplified LRA algorithm.
1767 2020-01-29 Martin Jambor <mjambor@suse.cz>
1769 PR tree-optimization/92706
1770 * tree-sra.c (struct access): Fields first_link, last_link,
1771 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1772 next_rhs_queued and grp_rhs_queued respectively, new fields
1773 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1774 (struct assign_link): Field next renamed to next_rhs, new field
1775 next_lhs. Updated comment.
1776 (work_queue_head): Renamed to rhs_work_queue_head.
1777 (lhs_work_queue_head): New variable.
1778 (add_link_to_lhs): New function.
1779 (relink_to_new_repr): Also relink LHS lists.
1780 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1781 (add_access_to_lhs_work_queue): New function.
1782 (pop_access_from_work_queue): Renamed to
1783 pop_access_from_rhs_work_queue.
1784 (pop_access_from_lhs_work_queue): New function.
1785 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1787 (child_would_conflict_in_lacc): Renamed to
1788 child_would_conflict_in_acc. Adjusted parameter names.
1789 (create_artificial_child_access): New parameter set_grp_read, use it.
1790 (subtree_mark_written_and_enqueue): Renamed to
1791 subtree_mark_written_and_rhs_enqueue.
1792 (propagate_subaccesses_across_link): Renamed to
1793 propagate_subaccesses_from_rhs.
1794 (propagate_subaccesses_from_lhs): New function.
1795 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1798 2020-01-29 Martin Jambor <mjambor@suse.cz>
1800 PR tree-optimization/92706
1801 * tree-sra.c (struct access): Adjust comment of
1802 grp_total_scalarization.
1803 (find_access_in_subtree): Look for single children spanning an entire
1805 (scalarizable_type_p): Allow register accesses, adjust callers.
1806 (completely_scalarize): Remove function.
1807 (scalarize_elem): Likewise.
1808 (create_total_scalarization_access): Likewise.
1809 (sort_and_splice_var_accesses): Do not track total scalarization
1811 (analyze_access_subtree): New parameter totally, adjust to new meaning
1812 of grp_total_scalarization.
1813 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1814 (can_totally_scalarize_forest_p): New function.
1815 (create_total_scalarization_access): Likewise.
1816 (create_total_access_and_reshape): Likewise.
1817 (total_should_skip_creating_access): Likewise.
1818 (totally_scalarize_subtree): Likewise.
1819 (analyze_all_variable_accesses): Perform total scalarization after
1820 subaccess propagation using the new functions above.
1821 (initialize_constant_pool_replacements): Output initializers by
1822 traversing the access tree.
1824 2020-01-29 Martin Jambor <mjambor@suse.cz>
1826 * tree-sra.c (verify_sra_access_forest): New function.
1827 (verify_all_sra_access_forests): Likewise.
1828 (create_artificial_child_access): Set parent.
1829 (analyze_all_variable_accesses): Call the verifier.
1831 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1833 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1834 if called on indirect edge.
1835 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1836 speculative call if needed.
1838 2020-01-29 Richard Biener <rguenther@suse.de>
1840 PR tree-optimization/93428
1841 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1842 permutation when the load node is created.
1843 (vect_analyze_slp_instance): Re-use it here.
1845 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1847 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1849 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1851 PR rtl-optimization/93272
1852 * ira-lives.c (process_out_of_region_eh_regs): New function.
1853 (process_bb_node_lives): Call it.
1855 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1857 * coverage.c (read_counts_file): Make error message lowercase.
1859 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1861 * profile-count.c (profile_quality_display_names): Fix ordering.
1863 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1866 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1867 hash only when edge is first within the sequence.
1868 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1869 (symbol_table::create_edge): Do not set target_prob.
1870 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1872 (cgraph_edge::make_speculative): Drop target_prob parameter.
1873 (cgraph_edge::speculative_call_info): Remove.
1874 (cgraph_edge::first_speculative_call_target): New member function.
1875 (update_call_stmt_hash_for_removing_direct_edge): New function.
1876 (cgraph_edge::resolve_speculation): Rewrite to new API.
1877 (cgraph_edge::speculative_call_for_target): New member function.
1878 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1879 multiple speculation targets.
1880 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1882 (verify_speculative_call): Verify that targets form an interval.
1883 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1884 (cgraph_edge::first_speculative_call_target): New member function.
1885 (cgraph_edge::next_speculative_call_target): New member function.
1886 (cgraph_edge::speculative_call_target_ref): New member function.
1887 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1888 (cgraph_edge): Remove target_prob.
1889 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1890 Fix handling of speculative calls.
1891 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1892 * ipa-fnsummary.c (analyze_function_body): Likewise.
1893 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1894 * ipa-profile.c (dump_histogram): Fix formating.
1895 (ipa_profile_generate_summary): Watch for overflows.
1896 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1897 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1898 (update_indirect_edges_after_inlining): Update to new API.
1899 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1901 * profile-count.h: (profile_probability::adjusted): New.
1902 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1903 updating of profile.
1904 * value-prof.c (gimple_ic_transform): Rename to ...
1905 (dump_ic_profile): ... this one; update dumping.
1906 (stream_in_histogram_value): Fix formating.
1907 (gimple_value_profile_transformations): Update.
1909 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1912 * config/i386/i386.md (*movoi_internal_avx): Remove
1913 TARGET_SSE_TYPELESS_STORES check.
1914 (*movti_internal): Prefer TARGET_AVX over
1915 TARGET_SSE_TYPELESS_STORES.
1916 (*movtf_internal): Likewise.
1917 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1918 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1919 from TARGET_SSE_TYPELESS_STORES.
1921 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1923 * diagnostic-core.h (warning_at): Rename overload to...
1924 (warning_meta): ...this.
1925 (emit_diagnostic_valist): Delete decl of overload taking
1926 diagnostic_metadata.
1927 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1928 (warning_at): Rename overload taking diagnostic_metadata to...
1929 (warning_meta): ...this.
1931 2020-01-28 Richard Biener <rguenther@suse.de>
1933 PR tree-optimization/93439
1934 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1935 * tree-cfg.c (move_sese_region_to_fn): ... here.
1936 (verify_types_in_gimple_reference): Verify used cliques are
1939 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1942 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1943 argument of a pointer to struct gcc_options and pass it to
1944 parse_mtune_ctrl_str.
1945 (ix86_function_specific_restore): Pass opts to
1946 set_ix86_tune_features.
1947 (ix86_option_override_internal): Likewise.
1948 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1949 gcc_options and use it for x_ix86_tune_ctrl_string.
1951 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1953 PR rtl-optimization/87763
1954 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1955 simplification to handle subregs as well as bare regs.
1956 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1958 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1960 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1961 for reduction chains that (now) include a call.
1963 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1965 PR tree-optimization/92822
1966 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1967 out the don't-care elements of a vector whose significant elements
1968 are duplicates, make the don't-care elements duplicates too.
1970 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1972 PR tree-optimization/93434
1973 * tree-predcom.c (split_data_refs_to_components): Record which
1974 components have had aliasing loads removed. Prevent store-store
1975 commoning for all such components.
1977 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1980 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1981 -1 or is_vshift is true, use new_vector with number of elts npatterns
1982 rather than new_unary_operation.
1984 PR tree-optimization/93454
1985 * gimple-fold.c (fold_array_ctor_reference): Perform
1986 elt_size.to_uhwi () just once, instead of calling it in every
1987 iteration. Punt if that value is above size of the temporary
1988 buffer. Decrease third native_encode_expr argument when
1989 bufoff + elt_sz is above size of buf.
1991 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1993 * config/mips/mips.c (mips_declare_object_name)
1994 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1996 2020-01-27 Martin Liska <mliska@suse.cz>
1998 PR gcov-profile/93403
1999 * tree-profile.c (gimple_init_gcov_profiler): Generate
2000 both __gcov_indirect_call_profiler_v4 and
2001 __gcov_indirect_call_profiler_v4_atomic.
2003 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2006 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2008 (@aarch64_split_simd_mov<mode>): Use it.
2009 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2010 Leave the vec_extract patterns to handle 2-element vectors.
2011 (aarch64_simd_mov_from_<mode>high): Likewise.
2012 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2013 (vec_extractv2dfv1df): Likewise.
2015 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2017 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2018 jump conditions for *compare_condjump<GPI:mode>.
2020 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2023 * digraph.cc (test_edge::test_edge): Specify template for base
2026 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2028 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2030 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2032 * config/arc/arc-protos.h (gen_mlo): Remove.
2033 (gen_mhi): Likewise.
2034 * config/arc/arc.c (AUX_MULHI): Define.
2035 (arc_must_save_reister): Special handling for r58/59.
2036 (arc_compute_frame_size): Consider mlo/mhi registers.
2037 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2039 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2040 mlo/mhi name selection.
2041 (arc_restore_callee_saves): Don't early restore blink when ISR.
2042 (arc_expand_prologue): Add mlo/mhi saving.
2043 (arc_expand_epilogue): Add mlo/mhi restoring.
2046 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2047 numbering when MUL64 option is used.
2048 (DWARF2_FRAME_REG_OUT): Define.
2049 * config/arc/arc.md (arc600_stall): New pattern.
2050 (VUNSPEC_ARC_ARC600_STALL): Define.
2051 (mulsi64): Use correct mlo/mhi registers.
2052 (mulsi_600): Clean it up.
2053 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2055 (mhi_operand): Likewise.
2057 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2058 Petro Karashchenko <petro.karashchenko@ring.com>
2060 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2061 attributes if needed.
2062 (prepare_move_operands): Generate special unspec instruction for
2064 (arc_isuncached_mem_p): Propagate uncached attribute to each
2066 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2067 (VUNSPEC_ARC_STDI): Likewise.
2068 (ALLI): New mode iterator.
2069 (mALLI): New mode attribute.
2070 (lddi): New instruction pattern.
2072 (stdidi_split): Split instruction for architectures which are not
2073 supporting ll64 option.
2074 (lddidi_split): Likewise.
2076 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2078 PR rtl-optimization/92989
2079 * lra-lives.c (process_bb_lives): Update the live-in set before
2080 processing additional clobbers.
2082 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2084 PR rtl-optimization/93170
2085 * cselib.c (cselib_invalidate_regno_val): New function, split out
2087 (cselib_invalidate_regno): ...here.
2088 (cselib_invalidated_by_call_p): New function.
2089 (cselib_process_insn): Iterate over all the hard-register entries in
2090 REG_VALUES and invalidate any that cross call-clobbered registers.
2092 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2094 * dojump.c (split_comparison): Use HONOR_NANS rather than
2095 HONOR_SNANS when splitting LTGT.
2097 2020-01-27 Martin Liska <mliska@suse.cz>
2100 * opts.c (print_filtered_help): Exclude language-specific
2101 options from --help=common unless enabled in all FEs.
2103 2020-01-27 Martin Liska <mliska@suse.cz>
2105 * opts.c (print_help): Exclude params from
2106 all except --help=param.
2108 2020-01-27 Martin Liska <mliska@suse.cz>
2111 * config/i386/i386-features.c (make_resolver_func):
2112 Align the code with ppc64 target implementation.
2113 Do not generate a unique name for resolver function.
2115 2020-01-27 Richard Biener <rguenther@suse.de>
2117 PR tree-optimization/93397
2118 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2119 converted reduction chain SLP graph adjustment.
2121 2020-01-26 Marek Polacek <polacek@redhat.com>
2124 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2127 2020-01-26 Jason Merrill <jason@redhat.com>
2130 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2133 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2135 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2136 (rx_setmem): Likewise.
2138 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2141 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2142 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2143 drop <di> from constraint of last operand.
2146 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2147 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2148 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2150 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2153 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2156 2020-01-24 Jeff Law <law@redhat.com>
2158 PR tree-optimization/92788
2159 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2162 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2165 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2166 *avx_vperm_broadcast_<mode>,
2167 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2168 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2169 Move before avx2_perm<mode>/avx512f_perm<mode>.
2172 * simplify-rtx.c (simplify_const_unary_operation,
2173 simplify_const_binary_operation): Punt for mode precision above
2174 MAX_BITSIZE_MODE_ANY_INT.
2176 2020-01-24 Andrew Pinski <apinski@marvell.com>
2178 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2181 2020-01-24 Jeff Law <law@redhat.com>
2184 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2185 for REGs. Call output_operand_lossage to get more reasonable
2188 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2190 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2191 gcn_fp_compare_operator.
2192 (vec_cmpu<mode>di): Use gcn_compare_operator.
2193 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2194 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2195 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2196 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2197 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2198 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2199 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2200 gcn_fp_compare_operator.
2201 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2202 gcn_fp_compare_operator.
2203 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2204 gcn_fp_compare_operator.
2205 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2206 gcn_fp_compare_operator.
2208 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2210 * doc/install.texi (Cross-Compiler-Specific Options): Document
2211 `--with-toolexeclibdir' option.
2213 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2215 * target.def (flags_regnum): Also mention effect on delay slot filling.
2216 * doc/tm.texi: Regenerate.
2218 2020-01-23 Jeff Law <law@redhat.com>
2220 PR translation/90162
2221 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2223 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2226 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2229 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2231 PR rtl-optimization/93402
2232 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2235 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2237 * config.in: Regenerated.
2238 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2239 for TARGET_LIBC_GNUSTACK.
2240 * configure: Regenerated.
2241 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2242 found to be 2.31 or greater.
2244 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2246 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2248 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2249 (mips_asm_file_end): New function. Delegate to
2250 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2251 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2253 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2256 * config/i386/i386-modes.def (POImode): New mode.
2257 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2258 * config/i386/i386.md (DPWI): New mode attribute.
2259 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2261 (QPWI): ... this. Use POI instead of OI for TImode.
2262 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2263 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2266 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2269 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2271 (speculation_tracker_rev): New pattern.
2272 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2273 Use speculation_tracker_rev to track the inverse condition.
2275 2020-01-23 Richard Biener <rguenther@suse.de>
2277 PR tree-optimization/93381
2278 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2279 alias-set of the def as argument and record the first one.
2280 (vn_walk_cb_data::first_set): New member.
2281 (vn_reference_lookup_3): Pass the alias-set of the current def
2282 to push_partial_def. Fix alias-set used in the aggregate copy
2284 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2285 * real.c (clear_significand_below): Fix out-of-bound access.
2287 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2290 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2291 New define_insn patterns.
2293 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2295 * doc/sourcebuild.texi (check-function-bodies): Add an
2296 optional target/xfail selector.
2298 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2300 PR rtl-optimization/93124
2301 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2302 bare USE and CLOBBER insns.
2304 2020-01-22 Andrew Pinski <apinski@marvell.com>
2306 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2308 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2311 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2312 diagnostic_manager into "ana" namespace.
2313 * selftest-run-tests.c (selftest::run_tests): Update for move of
2314 selftest::run_analyzer_selftests to
2315 ana::selftest::run_analyzer_selftests.
2317 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2319 * cfgexpand.c (union_stack_vars): Update the size.
2321 2020-01-22 Richard Biener <rguenther@suse.de>
2323 PR tree-optimization/93381
2324 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2325 throughout, handle all conversions the same.
2327 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2330 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2331 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2332 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2333 Call force_reg on high_in2 unconditionally.
2335 2020-01-22 Martin Liska <mliska@suse.cz>
2337 PR tree-optimization/92924
2338 * profile.c (compute_value_histograms): Divide
2341 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2344 * output.h (assemble_name_resolve): Declare.
2345 * varasm.c (assemble_name_resolve): New function.
2346 (assemble_name): Use it.
2347 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2349 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2351 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2352 update_web_docs_git instead of update_web_docs_svn.
2354 2020-01-21 Andrew Pinski <apinski@marvell.com>
2357 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2358 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2359 (*tlsgd_small_<mode>): Likewise.
2360 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2361 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2362 register. Convert that register back to dest using convert_mode.
2364 2020-01-21 Jim Wilson <jimw@sifive.com>
2366 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2369 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2370 Uros Bizjak <ubizjak@gmail.com>
2373 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2375 (legitimize_tls_address): Do GNU2 TLS address computation in
2376 ptr_mode and zero-extend result to Pmode.
2377 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2378 :P with :PTR and Pmode with ptr_mode.
2379 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2380 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2381 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2383 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2386 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2387 the last two operands are CONST_INT_P before using them as such.
2389 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2391 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2392 to get the integer element types.
2394 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2396 * config/aarch64/aarch64-sve-builtins.h
2397 (function_expander::convert_to_pmode): Declare.
2398 * config/aarch64/aarch64-sve-builtins.cc
2399 (function_expander::convert_to_pmode): New function.
2400 (function_expander::get_contiguous_base): Use it.
2401 (function_expander::prepare_gather_address_operands): Likewise.
2402 * config/aarch64/aarch64-sve-builtins-sve2.cc
2403 (svwhilerw_svwhilewr_impl::expand): Likewise.
2405 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2408 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2409 cfun->machine->label_is_assembled.
2410 (aarch64_print_patchable_function_entry): New.
2411 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2412 * config/aarch64/aarch64.h (struct machine_function): New field,
2415 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2418 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2421 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2424 * cgraph.c (cgraph_edge::resolve_speculation,
2425 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2426 call_stmt_site_hash.
2428 2020-01-21 Martin Liska <mliska@suse.cz>
2430 * config/rs6000/rs6000.c (common_mode_defined): Remove
2433 2020-01-21 Richard Biener <rguenther@suse.de>
2435 PR tree-optimization/92328
2436 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2437 type when value-numbering same-sized store by inserting a
2439 (eliminate_dom_walker::eliminate_stmt): When eliminating
2440 a redundant store handle bit-reinterpretation of the same value.
2442 2020-01-21 Andrew Pinski <apinski@marvel.com>
2445 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2447 (prepare_block_for_update): This. Use a worklist instead of
2450 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2452 * gcc/config/arm/arm.c (clear_operation_p):
2453 Initialise last_regno, skip first iteration
2454 based on the first_set value and use ints instead
2455 of the unnecessary HOST_WIDE_INTs.
2457 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2460 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2461 compare_mode other than SFmode or DFmode.
2463 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2466 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2467 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2468 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2470 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2472 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2474 2020-01-20 Andrew Pinski <apinski@marvell.com>
2477 * targhooks.c (default_print_patchable_function_entry): Use
2478 output_asm_insn to emit the nop instruction.
2480 2020-01-20 Fangrui Song <maskray@google.com>
2483 * targhooks.c (default_print_patchable_function_entry): Align to
2486 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2489 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2490 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2491 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2492 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2493 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2494 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2495 Remove the {q} suffix from lea.
2496 (*tls_dynamic_gnu2_call_64): Renamed to ...
2497 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2498 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2499 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2500 Pass Pmode to gen_tls_dynamic_gnu2_64.
2502 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2504 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2506 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2508 * config/aarch64/aarch64-sve-builtins-base.cc
2509 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2511 2020-01-20 Richard Biener <rguenther@suse.de>
2514 * dwarf2out.c (prune_unused_types): Unconditionally mark
2515 called function DIEs.
2517 2020-01-20 Martin Liska <mliska@suse.cz>
2519 PR tree-optimization/93199
2520 * tree-eh.c (struct leh_state): Add
2521 new field outer_non_cleanup.
2522 (cleanup_is_dead_in): Pass leh_state instead
2523 of eh_region. Add a checking that state->outer_non_cleanup
2524 points to outer non-clean up region.
2525 (lower_try_finally): Record outer_non_cleanup
2527 (lower_catch): Likewise.
2528 (lower_eh_filter): Likewise.
2529 (lower_eh_must_not_throw): Likewise.
2530 (lower_cleanup): Likewise.
2532 2020-01-20 Richard Biener <rguenther@suse.de>
2534 PR tree-optimization/93094
2535 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2536 (vect_transform_loop): Likewise.
2537 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2538 loop_vectorized_call to vect_transform_loop.
2539 * tree-vect-loop.c (vect_transform_loop): Pass down
2540 loop_vectorized_call to vect_loop_versioning.
2541 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2542 the earlier discovered loop_vectorized_call.
2544 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2546 * doc/contribute.texi: Update for SVN -> Git transition.
2547 * doc/install.texi: Likewise.
2549 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2552 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2553 speculative targets.
2554 (verify_speculative_call): New function
2555 (cgraph_node::verify_node): Use it.
2556 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2559 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2562 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2563 (cgraph_edge::make_direct): Remove all indirect targets.
2564 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2565 (cgraph_node::verify_node): Verify that only one call_stmt or
2566 lto_stmt_uid is set.
2567 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2569 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2570 (lto_output_ref): Simplify streaming of stmt.
2571 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2573 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2575 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2576 Mark parameter unused.
2578 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2580 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2582 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2584 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2586 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2588 * Makefile.in: Add coroutine-passes.o.
2589 * builtin-types.def (BT_CONST_SIZE): New.
2590 (BT_FN_BOOL_PTR): New.
2591 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2592 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2593 * coroutine-builtins.def: New file.
2594 * coroutine-passes.cc: New file.
2595 * function.h (struct GTY function): Add a bit to indicate that the
2596 function is a coroutine component.
2597 * internal-fn.c (expand_CO_FRAME): New.
2598 (expand_CO_YIELD): New.
2599 (expand_CO_SUSPN): New.
2600 (expand_CO_ACTOR): New.
2601 * internal-fn.def (CO_ACTOR): New.
2605 * passes.def: Add pass_coroutine_lower_builtins,
2606 pass_coroutine_early_expand_ifns.
2607 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2608 (make_pass_coroutine_early_expand_ifns): New.
2609 * doc/invoke.texi: Document the fcoroutines command line
2612 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2614 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2617 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2618 after checking the argument is a REG. Don't use REGNO (reg)
2619 again to set last_regno, reuse regno variable instead.
2621 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2623 * doc/analyzer.texi (Limitations): Add note about NaN.
2625 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2626 Sudakshina Das <sudi.das@arm.com>
2628 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2629 and valid immediate.
2630 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2631 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2632 * config/arm/constraints.md (Pg): New.
2633 * config/arm/predicates.md (long_shift_imm): New.
2634 (arm_reg_or_long_shift_imm): Likewise.
2635 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2636 (thumb2_lsll): Likewise.
2639 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2640 Sudakshina Das <sudi.das@arm.com>
2642 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2643 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2644 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2645 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2646 * config/arm/thumb2.md (thumb2_asrl): New.
2647 (thumb2_lsll): Likewise.
2649 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2651 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2654 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2656 * gdbinit.in (help-gcc-hooks): New command.
2657 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2658 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2661 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2663 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2664 correct target macro.
2666 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2668 * config/aarch64/aarch64-protos.h
2669 (aarch64_sve_ld1ro_operand_p): New.
2670 * config/aarch64/aarch64-sve-builtins-base.cc
2671 (class load_replicate): New.
2672 (class svld1ro_impl): New.
2673 (class svld1rq_impl): Change to inherit from load_replicate.
2674 (svld1ro): New sve intrinsic function base.
2675 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2676 New DEF_SVE_FUNCTION.
2677 * config/aarch64/aarch64-sve-builtins-base.h
2678 (svld1ro): New decl.
2679 * config/aarch64/aarch64-sve-builtins.cc
2680 (function_expander::add_mem_operand): Modify assert to allow
2682 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2684 * config/aarch64/aarch64.c
2685 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2686 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2687 (aarch64_sve_ld1ro_operand_p): New.
2688 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2689 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2690 * config/aarch64/predicates.md
2691 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2693 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2695 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2696 Introduce this ACLE specified predefined macro.
2697 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2698 (fp): Disabling this disables f64mm.
2699 (simd): Disabling this disables f64mm.
2700 (fp16): Disabling this disables f64mm.
2701 (sve): Disabling this disables f64mm.
2702 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2703 (AARCH64_ISA_F64MM): New.
2704 (TARGET_F64MM): New.
2705 * doc/invoke.texi (f64mm): Document new option.
2707 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2709 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2710 (neoversen1_tunings): Likewise.
2712 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2715 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2716 Add assert to ensure prolog has been emitted.
2717 (aarch64_split_atomic_op): Likewise.
2718 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2719 Use epilogue_completed rather than reload_completed.
2720 (aarch64_atomic_exchange<mode>): Likewise.
2721 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2722 (atomic_nand<mode>): Likewise.
2723 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2724 (atomic_fetch_nand<mode>): Likewise.
2725 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2726 (atomic_nand_fetch<mode>): Likewise.
2728 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2731 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2733 (REVERSE_CONDITION): Delete.
2734 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2735 (CCFP_CCFPE): Likewise.
2736 (e): New mode attribute.
2737 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2738 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2739 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2740 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2741 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2742 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2743 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2744 name of generator from gen_ccmpdi to gen_ccmpccdi.
2745 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2746 the previous comparison but aren't able to, use the new ccmp_rev
2749 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2751 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2752 than testing directly for INTEGER_CST.
2753 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2755 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2757 PR tree-optimization/93292
2758 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2759 get_vectype_for_scalar_type returns NULL.
2761 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2763 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2764 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2766 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2768 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2770 * params.opt: (max-predicted-iterations): Set bounds.
2771 * predict.c (real_almost_one, real_br_prob_base,
2772 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2773 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2774 probabilities; do not truncate to reg_br_prob_bases.
2775 (estimate_loops_at_level): Pass max_cyclic_prob.
2776 (estimate_loops): Compute max_cyclic_prob.
2777 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2779 * profile-count.c (profile_probability::to_sreal): New.
2780 * profile-count.h (class sreal): Move up in file.
2781 (profile_probability::to_sreal): Declare.
2783 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2786 (arm_invalid_conversion): New function for target hook.
2787 (arm_invalid_unary_op): New function for target hook.
2788 (arm_invalid_binary_op): New function for target hook.
2790 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2792 * config.gcc: Add arm_bf16.h.
2793 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2794 (arm_simd_builtin_std_type): Add BFmode.
2795 (arm_init_simd_builtin_types): Define element types for vector types.
2796 (arm_init_bf16_types): New function.
2797 (arm_init_builtins): Add arm_init_bf16_types function call.
2798 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2799 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2800 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2801 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2802 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2803 (arm_mangle_type): Add __bf16.
2804 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2805 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2806 arm_bf16_ptr_type_node.
2807 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2808 define_split between ARM registers.
2809 * config/arm/arm_bf16.h: New file.
2810 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2811 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2813 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2814 * config/arm/vfp.md: Add BFmode to movhf patterns.
2816 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2817 Andre Vieira <andre.simoesdiasvieira@arm.com>
2819 * config/arm/arm-cpus.in (mve, mve_float): New features.
2820 (dsp, mve, mve.fp): New options.
2821 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2822 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2823 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2825 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2826 Thomas Preud'homme <thomas.preudhomme@arm.com>
2828 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2830 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2831 error for using -mcmse when targeting Armv8.1-M Mainline.
2833 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2834 Thomas Preud'homme <thomas.preudhomme@arm.com>
2836 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2837 address in r4 when targeting Armv8.1-M Mainline.
2838 (nonsecure_call_value_internal): Likewise.
2839 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2840 a register match_operand again. Emit BLXNS when targeting
2842 (nonsecure_call_value_reg_thumb2): Likewise.
2844 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2845 Thomas Preud'homme <thomas.preudhomme@arm.com>
2847 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2848 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2849 variable as true when floating-point ABI is not hard. Replace
2850 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2851 Generate VLSTM and VLLDM instruction respectively before and
2852 after a function call to cmse_nonsecure_call function.
2853 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2854 (VUNSPEC_VLLDM): Likewise.
2855 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2856 (lazy_load_multiple_insn): Likewise.
2858 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2859 Thomas Preud'homme <thomas.preudhomme@arm.com>
2861 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2862 (arm_emit_vfp_multi_reg_pop): Likewise.
2863 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2864 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2865 restore callee-saved VFP registers.
2867 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2868 Thomas Preud'homme <thomas.preudhomme@arm.com>
2870 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2871 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2872 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2873 callee-saved GPRs as well as clear ip register before doing a nonsecure
2874 call then restore callee-saved GPRs after it when targeting
2876 (arm_reorg): Adapt to function rename.
2878 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2879 Thomas Preud'homme <thomas.preudhomme@arm.com>
2881 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2882 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2883 clear_vfp_multiple pattern based on a new vfp parameter.
2884 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2885 targeting Armv8.1-M Mainline.
2886 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2887 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2888 whether VFP registers are available before looking call_used_regs for a
2890 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2891 of prototype of clear_operation_p.
2892 (clear_vfp_multiple_operation): New predicate.
2893 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2894 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2896 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2897 Thomas Preud'homme <thomas.preudhomme@arm.com>
2899 * config/arm/arm-protos.h (clear_operation_p): Declare.
2900 * config/arm/arm.c (clear_operation_p): New function.
2901 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2902 targeting Armv8.1-M Mainline or successor.
2903 (output_return_instruction): Only output APSR register clearing if
2904 Armv8.1-M Mainline instructions not available.
2905 (thumb_exit): Likewise.
2906 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2907 * config/arm/thumb2.md (clear_apsr): New define_insn.
2908 (clear_multiple): Likewise.
2909 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2911 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2912 Thomas Preud'homme <thomas.preudhomme@arm.com>
2914 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2915 (use_return_insn): Also return false for Armv8.1-M Mainline.
2916 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2917 Mainline instructions are available.
2918 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2919 when targeting Armv8.1-M Mainline Security Extensions.
2920 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2921 Mainline entry function.
2922 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2923 targeting Armv8.1-M Mainline or successor.
2924 (arm_expand_epilogue): Fix indentation of caller-saved register
2925 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2927 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2928 (FP_SYSREGS): Likewise.
2929 (enum vfp_sysregs_encoding): Define enum.
2930 (fp_sysreg_names): Declare.
2931 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2932 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2933 (pop_fpsysreg_insn): Likewise.
2935 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2936 Thomas Preud'homme <thomas.preudhomme@arm.com>
2938 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2939 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2940 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2941 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2942 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2943 (ARMv8_1m_main): New feature group.
2944 (armv8.1-m.main): New architecture.
2945 * config/arm/arm-tables.opt: Regenerate.
2946 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2947 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2948 (arm_options_perform_arch_sanity_checks): Error out when targeting
2949 Armv8.1-M Mainline Security Extensions.
2950 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2952 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2954 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2955 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2956 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2957 aarch64_bfdot_laneq): New.
2958 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2959 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2960 vbfdotq_laneq_f32): New.
2961 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2962 VBFMLA_W, VBF): New.
2963 (isquadop): Add V4BF, V8BF.
2965 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2967 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2968 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2969 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2970 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2971 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2972 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2973 usdot_laneq, sudot_lane,sudot_laneq): New.
2974 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2975 (aarch64_<sur>dot_lane): New.
2976 * config/aarch64/arm_neon.h (vusdot_s32): New.
2978 (vusdot_lane_s32): New.
2979 (vsudot_lane_s32): New.
2980 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2981 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2983 2020-01-16 Martin Liska <mliska@suse.cz>
2985 * value-prof.c (dump_histogram_value): Fix
2986 obvious spacing issue.
2988 2020-01-16 Andrew Pinski <apinski@marvell.com>
2990 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2991 !storage_order_barrier_p.
2993 2020-01-16 Andrew Pinski <apinski@marvell.com>
2995 * sched-int.h (_dep): Add unused bit-field field for the padding.
2996 * sched-deps.c (init_dep_1): Init unused field.
2998 2020-01-16 Andrew Pinski <apinski@marvell.com>
3000 * optabs.h (create_expand_operand): Initialize target field also.
3002 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3004 PR tree-optimization/92429
3005 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3006 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3008 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3011 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3013 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3014 aarch64_sve_int_mode to each mode.
3016 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3018 * doc/analyzer.texi (Overview): Add note about
3019 -fdump-ipa-analyzer.
3021 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3023 PR tree-optimization/93231
3024 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3025 input_type is unsigned. Use tree_to_shwi for shift constant.
3026 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3027 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3028 inputs more efficiently.
3030 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3032 * config/i386/i386.md (*movsf_internal): Do not require
3033 SSE2 ISA for alternatives 14 and 15.
3035 2020-01-15 Richard Biener <rguenther@suse.de>
3038 * tree-eh.c (sink_clobbers): If we already visited the destination
3039 block do not defer insertion.
3040 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3041 the purpose of defered insertion.
3043 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3045 * BASE-VER: Bump to 10.0.1.
3047 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3049 PR tree-optimization/93247
3050 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3051 type of the stmt that we're going to vectorize.
3053 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3055 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3056 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3059 2020-01-15 Martin Liska <mliska@suse.cz>
3061 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3062 2 calls of streamer_read_hwi in a function call.
3064 2020-01-15 Richard Biener <rguenther@suse.de>
3066 * alias.c (record_alias_subset): Avoid redundant work when
3067 subset is already recorded.
3069 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3071 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3072 the analyzer options provide CWE identifiers.
3074 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3076 * tree-diagnostic-path.cc (path_summary::event_range::print):
3077 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3078 using get_pure_location.
3080 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3082 PR tree-optimization/93262
3083 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3084 perform head trimming only if the last argument is constant,
3085 either all ones, or larger or equal to head trim, in the latter
3086 case decrease the last argument by head_trim.
3088 PR tree-optimization/93249
3089 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3090 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3091 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3092 perform head trim unless we can prove there are no '\0' chars
3093 from the source among the first head_trim chars.
3095 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3097 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3099 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3102 * config/i386/sse.md
3103 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3104 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3105 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3106 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3107 just a single alternative instead of two, make operands 1 and 2
3110 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3113 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3116 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3118 * Makefile.in (lang_opt_files): Add analyzer.opt.
3119 (ANALYZER_OBJS): New.
3120 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3121 tristate.o and ANALYZER_OBJS.
3122 (TEXI_GCCINT_FILES): Add analyzer.texi.
3123 * common.opt (-fanalyzer): New driver option.
3124 * config.in: Regenerate.
3125 * configure: Regenerate.
3126 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3127 (gccdepdir): Also create depdir for "analyzer" subdir.
3128 * digraph.cc: New file.
3129 * digraph.h: New file.
3130 * doc/analyzer.texi: New file.
3131 * doc/gccint.texi ("Static Analyzer") New menu item.
3132 (analyzer.texi): Include it.
3133 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3134 ("Warning Options"): Add static analysis warnings to the list.
3135 (-Wno-analyzer-double-fclose): New option.
3136 (-Wno-analyzer-double-free): New option.
3137 (-Wno-analyzer-exposure-through-output-file): New option.
3138 (-Wno-analyzer-file-leak): New option.
3139 (-Wno-analyzer-free-of-non-heap): New option.
3140 (-Wno-analyzer-malloc-leak): New option.
3141 (-Wno-analyzer-possible-null-argument): New option.
3142 (-Wno-analyzer-possible-null-dereference): New option.
3143 (-Wno-analyzer-null-argument): New option.
3144 (-Wno-analyzer-null-dereference): New option.
3145 (-Wno-analyzer-stale-setjmp-buffer): New option.
3146 (-Wno-analyzer-tainted-array-index): New option.
3147 (-Wno-analyzer-use-after-free): New option.
3148 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3149 (-Wno-analyzer-use-of-uninitialized-value): New option.
3150 (-Wanalyzer-too-complex): New option.
3151 (-fanalyzer-call-summaries): New warning.
3152 (-fanalyzer-checker=): New warning.
3153 (-fanalyzer-fine-grained): New warning.
3154 (-fno-analyzer-state-merge): New warning.
3155 (-fno-analyzer-state-purge): New warning.
3156 (-fanalyzer-transitivity): New warning.
3157 (-fanalyzer-verbose-edges): New warning.
3158 (-fanalyzer-verbose-state-changes): New warning.
3159 (-fanalyzer-verbosity=): New warning.
3160 (-fdump-analyzer): New warning.
3161 (-fdump-analyzer-callgraph): New warning.
3162 (-fdump-analyzer-exploded-graph): New warning.
3163 (-fdump-analyzer-exploded-nodes): New warning.
3164 (-fdump-analyzer-exploded-nodes-2): New warning.
3165 (-fdump-analyzer-exploded-nodes-3): New warning.
3166 (-fdump-analyzer-supergraph): New warning.
3167 * doc/sourcebuild.texi (dg-require-dot): New.
3168 (dg-check-dot): New.
3169 * gdbinit.in (break-on-saved-diagnostic): New command.
3170 * graphviz.cc: New file.
3171 * graphviz.h: New file.
3172 * ordered-hash-map-tests.cc: New file.
3173 * ordered-hash-map.h: New file.
3174 * passes.def (pass_analyzer): Add before
3175 pass_ipa_whole_program_visibility.
3176 * selftest-run-tests.c (selftest::run_tests): Call
3177 selftest::ordered_hash_map_tests_cc_tests.
3178 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3180 * shortest-paths.h: New file.
3181 * timevar.def (TV_ANALYZER): New timevar.
3182 (TV_ANALYZER_SUPERGRAPH): Likewise.
3183 (TV_ANALYZER_STATE_PURGE): Likewise.
3184 (TV_ANALYZER_PLAN): Likewise.
3185 (TV_ANALYZER_SCC): Likewise.
3186 (TV_ANALYZER_WORKLIST): Likewise.
3187 (TV_ANALYZER_DUMP): Likewise.
3188 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3189 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3190 * tree-pass.h (make_pass_analyzer): New decl.
3191 * tristate.cc: New file.
3192 * tristate.h: New file.
3194 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3197 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3198 alternatives 9 and 10.
3200 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3202 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3203 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3204 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3205 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3206 (selftest::hash_map_tests_c_tests): Call it.
3207 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3208 New static constant, using the value of = H::empty_zero_p.
3209 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3210 from default_hash_traits <Value>.
3211 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3213 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3214 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3215 calls to mark_empty with !Descriptor::empty_zero_p.
3216 (hash_table::empty_slow): Conditionalize the memset call with a
3217 check that Descriptor::empty_zero_p; otherwise, loop through the
3218 entries calling mark_empty on them.
3219 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3220 (pointer_hash::empty_zero_p): Likewise.
3221 (pair_hash::empty_zero_p): Likewise.
3222 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3224 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3225 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3226 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3227 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3228 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3229 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3230 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3231 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3233 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3236 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3238 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3239 fix typo on return value.
3241 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3244 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3246 (cgraph_edge::make_speculative): Add param for setting speculative_id
3248 (cgraph_edge::speculative_call_info): Update comments and find reference
3249 by speculative_id for multiple indirect targets.
3250 (cgraph_edge::resolve_speculation): Decrease the speculations
3251 for indirect edge, drop it's speculative if not direct target
3252 left. Update comments.
3253 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3254 (cgraph_node::dump): Print num_speculative_call_targets.
3255 (cgraph_node::verify_node): Don't report error if speculative
3256 edge not include statement.
3257 (cgraph_edge::num_speculative_call_targets_p): New function.
3258 * cgraph.h (int common_target_id): Remove.
3259 (int common_target_probability): Remove.
3260 (num_speculative_call_targets): New variable.
3261 (make_speculative): Add param for setting speculative_id.
3262 (cgraph_edge::num_speculative_call_targets_p): New declare.
3263 (target_prob): New variable.
3264 (speculative_id): New variable.
3265 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3266 call summaries for multiple speculative call targets.
3267 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3268 * ipa-profile.c (struct speculative_call_target): New struct.
3269 (class speculative_call_summary): New class.
3270 (class speculative_call_summaries): New class.
3271 (call_sums): New variable.
3272 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3273 (ipa_profile_write_edge_summary): New function.
3274 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3275 (ipa_profile_dump_all_summaries): New function.
3276 (ipa_profile_read_edge_summary): New function.
3277 (ipa_profile_read_summary_section): New function.
3278 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3279 (ipa_profile): Generate num_speculative_call_targets from
3281 * ipa-ref.h (speculative_id): New variable.
3282 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3283 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3284 common_target_probability. Stream out speculative_id and
3285 num_speculative_call_targets.
3286 (input_edge): Likewise.
3287 * predict.c (dump_prediction): Remove edges count assert to be
3289 * symtab.c (symtab_node::create_reference): Init speculative_id.
3290 (symtab_node::clone_references): Clone speculative_id.
3291 (symtab_node::clone_referring): Clone speculative_id.
3292 (symtab_node::clone_reference): Clone speculative_id.
3293 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3294 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3295 if indirect call contains multiple speculative targets.
3296 * value-prof.h (check_ic_target): Remove.
3297 * value-prof.c (gimple_value_profile_transformations):
3298 Use void function gimple_ic_transform.
3299 * value-prof.c (gimple_ic_transform): Handle topn case.
3300 Fix comment typos. Change it to a void function.
3302 2020-01-13 Andrew Pinski <apinski@marvell.com>
3304 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3305 (octeontx2t98): New define.
3306 (octeontx2t96): New define.
3307 (octeontx2t93): New define.
3308 (octeontx2f95): New define.
3309 (octeontx2f95n): New define.
3310 (octeontx2f95mm): New define.
3311 * config/aarch64/aarch64-tune.md: Regenerate.
3312 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3314 2020-01-13 Jason Merrill <jason@redhat.com>
3316 PR c++/33799 - destroy return value if local cleanup throws.
3317 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3319 2020-01-13 Martin Liska <mliska@suse.cz>
3321 * ipa-cp.c (get_max_overall_size): Use newly
3322 renamed param param_ipa_cp_unit_growth.
3323 * params.opt: Remove legacy param name.
3325 2020-01-13 Martin Sebor <msebor@redhat.com>
3327 PR tree-optimization/93213
3328 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3329 stores to be eliminated.
3331 2020-01-13 Martin Liska <mliska@suse.cz>
3333 * opts.c (print_help): Do not print CL_PARAM
3334 and CL_WARNING for CL_OPTIMIZATION.
3336 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3339 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3340 depending on optimization settings.
3342 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3344 PR tree-optimization/90838
3345 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3346 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3347 argument rather than to initialize temporary for targets that
3348 don't use the mode argument at all. Initialize ctzval to avoid
3351 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3353 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3354 * tree-core.h: Document it.
3355 * gimplify.c (gimplify_omp_workshare): Set it.
3356 * omp-low.c (lower_omp_target): Use it.
3357 * tree-pretty-print.c (dump_omp_clause): Print it.
3359 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3360 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3362 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3364 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3365 * common.opt (fdiagnostics-path-format=): New option.
3366 (diagnostic_path_format): New enum.
3367 (fdiagnostics-show-path-depths): New option.
3368 * coretypes.h (diagnostic_event_id_t): New forward decl.
3369 * diagnostic-color.c (color_dict): Add "path".
3370 * diagnostic-event-id.h: New file.
3371 * diagnostic-format-json.cc (json_from_expanded_location): Make
3373 (json_end_diagnostic): Call context->make_json_for_path if it
3374 exists and the diagnostic has a path.
3375 (diagnostic_output_format_init): Clear context->print_path.
3376 * diagnostic-path.h: New file.
3377 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3378 when printing a run of events in a diagnostic_path so that they
3379 all get the same color.
3380 (layout::m_diagnostic_path_p): New field.
3381 (layout::layout): Initialize it.
3382 (layout::print_any_labels): Don't colorize the label text for an
3383 event in a diagnostic_path.
3384 (gcc_rich_location::add_location_if_nearby): Add
3385 "restrict_to_current_line_spans" and "label" params. Pass the
3386 former to layout.maybe_add_location_range; pass the latter
3387 when calling add_range.
3388 * diagnostic.c: Include "diagnostic-path.h".
3389 (diagnostic_initialize): Initialize context->path_format and
3390 context->show_path_depths.
3391 (diagnostic_show_any_path): New function.
3392 (diagnostic_path::interprocedural_p): New function.
3393 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3394 (simple_diagnostic_path::num_events): New function.
3395 (simple_diagnostic_path::get_event): New function.
3396 (simple_diagnostic_path::add_event): New function.
3397 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3398 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3399 (debug): New overload taking a diagnostic_path *.
3400 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3401 * diagnostic.h (enum diagnostic_path_format): New enum.
3402 (json::value): New forward decl.
3403 (diagnostic_context::path_format): New field.
3404 (diagnostic_context::show_path_depths): New field.
3405 (diagnostic_context::print_path): New callback field.
3406 (diagnostic_context::make_json_for_path): New callback field.
3407 (diagnostic_show_any_path): New decl.
3408 (json_from_expanded_location): New decl.
3409 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3410 (-fdiagnostics-show-path-depths): New option.
3411 (-fdiagnostics-color): Add "path" to description of default
3412 GCC_COLORS; describe it.
3413 (-fdiagnostics-format=json): Document how diagnostic paths are
3414 represented in the JSON output format.
3415 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3416 Add optional params "restrict_to_current_line_spans" and "label".
3417 * opts.c (common_handle_option): Handle
3418 OPT_fdiagnostics_path_format_ and
3419 OPT_fdiagnostics_show_path_depths.
3420 * pretty-print.c: Include "diagnostic-event-id.h".
3421 (pp_format): Implement "%@" format code for printing
3422 diagnostic_event_id_t *.
3423 (selftest::test_pp_format): Add tests for "%@".
3424 * selftest-run-tests.c (selftest::run_tests): Call
3425 selftest::tree_diagnostic_path_cc_tests.
3426 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3427 * toplev.c (general_init): Initialize global_dc->path_format and
3428 global_dc->show_path_depths.
3429 * tree-diagnostic-path.cc: New file.
3430 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3431 non-static. Drop "diagnostic" param in favor of storing the
3432 original value of "where" and re-using it.
3433 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3434 maybe_unwind_expanded_macro_loc.
3435 (tree_diagnostics_defaults): Initialize context->print_path and
3436 context->make_json_for_path.
3437 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3439 (default_tree_make_json_for_path): New decl.
3440 (maybe_unwind_expanded_macro_loc): New decl.
3442 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3444 PR tree-optimization/93210
3445 * fold-const.h (native_encode_initializer,
3446 can_native_interpret_type_p): Declare.
3447 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3449 (native_encode_initializer): New function, moved from dwarf2out.c.
3450 Adjust to native_encode_expr compatible arguments, including dry-run
3451 and partial extraction modes. Don't handle STRING_CST.
3452 (can_native_interpret_type_p): No longer static.
3453 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3454 offset / BITS_PER_UNIT fits into int and don't call it if
3455 can_native_interpret_type_p fails. If suboff is NULL and for
3456 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3457 native_encode_initializer.
3458 (fold_const_aggregate_ref_1): Formatting fix.
3459 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3460 (tree_add_const_value_attribute): Adjust caller.
3462 PR tree-optimization/90838
3463 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3464 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3465 CTZ_DEFINED_VALUE_AT_ZERO.
3467 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3470 * lra-constraints.c (match_reload): Permit input operands have the
3471 same mode as output while other input operands have a different
3474 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3476 PR tree-optimization/90838
3477 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3478 (check_ctz_string): Likewise.
3479 (optimize_count_trailing_zeroes): Likewise.
3480 (simplify_count_trailing_zeroes): Likewise.
3481 (pass_forwprop::execute): Try ctz simplification.
3482 * match.pd: Add matching for ctz idioms.
3484 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3486 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3488 (aarch64_invalid_unary_op): New function for target hook.
3489 (aarch64_invalid_binary_op): New function for target hook.
3491 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3493 * config.gcc: Add arm_bf16.h.
3494 * config/aarch64/aarch64-builtins.c
3495 (aarch64_simd_builtin_std_type): Add BFmode.
3496 (aarch64_init_simd_builtin_types): Define element types for vector
3498 (aarch64_init_bf16_types): New function.
3499 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3500 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3502 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3503 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3505 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3506 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3507 * config/aarch64/aarch64.c
3508 (aarch64_classify_vector_mode): Add support for BF types.
3509 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3510 (aarch64_vq_mode): Add support for BF types.
3511 (aarch64_simd_container_mode): Add support for BF types.
3512 (aarch64_mangle_type): Add support for BF scalar type.
3513 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3514 * config/aarch64/arm_bf16.h: New file.
3515 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3516 * config/aarch64/iterators.md: Add BF types to mode attributes.
3517 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3519 2020-01-10 Jason Merrill <jason@redhat.com>
3521 PR c++/93173 - incorrect tree sharing.
3522 * gimplify.c (copy_if_shared): No longer static.
3523 * gimplify.h: Declare it.
3525 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3527 * doc/invoke.texi (-msve-vector-bits=): Document that
3528 -msve-vector-bits=128 now generates VL-specific code for
3529 little-endian targets.
3530 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3531 build_vector_type_for_mode to construct the data vector types.
3532 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3533 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3534 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3535 for 128-bit vectors.
3537 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3539 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3542 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3544 * config/aarch64/aarch64-builtins.c
3545 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3546 rather than checking the number of elements and the element mode.
3548 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3550 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3551 get_related_vectype_for_scalar_type rather than build_vector_type
3552 to create the index type for a conditional reduction.
3554 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3556 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3557 for any type of gather or scatter, including strided accesses.
3559 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3561 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3564 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3566 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3568 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3569 parameter and its use to reset DR_OFFSET's.
3570 (vect_transform_loop): Remove orig_drs_init argument.
3571 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3572 member of dr_vec_info rather than the offset of the associated
3573 data_reference's innermost_loop_behavior.
3574 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3575 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3576 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3577 get_dr_vinfo_offset.
3578 (vectorizable_store): Likewise.
3579 (vectorizable_load): Likewise.
3581 2020-01-10 Richard Biener <rguenther@suse.de>
3583 * gimple-ssa-store-merging
3584 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3586 2020-01-10 Martin Liska <mliska@suse.cz>
3589 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3590 encapsulation that was there before r280040.
3592 2020-01-10 Richard Biener <rguenther@suse.de>
3595 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3596 sequences to avoid walking them again for secondary opportunities.
3597 (pass_lower_eh_dispatch::execute): Instead actually insert
3600 2020-01-10 Richard Biener <rguenther@suse.de>
3603 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3604 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3605 avoid quadraticness.
3607 2020-01-10 Martin Jambor <mjambor@suse.cz>
3609 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3610 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3611 to get param_ipa_sra_max_replacements.
3612 (param_splitting_across_edge): Pass the caller to
3613 pull_accesses_from_callee.
3615 2020-01-10 Martin Jambor <mjambor@suse.cz>
3617 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3618 * ipa-cp.c (max_new_size): Removed.
3619 (orig_overall_size): New variable.
3620 (get_max_overall_size): New function.
3621 (estimate_local_effects): Use it. Adjust dump.
3622 (decide_about_value): Likewise.
3623 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3624 orig_overall_size. Adjust dump.
3625 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3627 2020-01-10 Martin Jambor <mjambor@suse.cz>
3629 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3630 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3631 instead of param_ipa_max_agg_items.
3632 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3633 optimization info for the callee.
3635 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3637 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3638 markers if debug_inline_points is false.
3640 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3642 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3644 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3645 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3646 aarch64-sve-builtins-sve2.h.
3647 (aarch64-sve-builtins-sve2.o): New rule.
3648 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3649 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3650 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3651 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3652 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3653 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3655 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3656 instructions that are handled here.
3657 (@cond_asrd<mode>): Generalize to...
3658 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3659 (*cond_asrd<mode>_2): Generalize to...
3660 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3661 (*cond_asrd<mode>_z): Generalize to...
3662 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3663 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3664 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3665 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3666 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3668 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3669 (@aarch64_scatter_stnt<mode>): Likewise.
3670 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3671 (@aarch64_mul_lane_<mode>): Likewise.
3672 (@aarch64_sve_suqadd<mode>_const): Likewise.
3673 (*<sur>h<addsub><mode>): Generalize to...
3674 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3676 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3677 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3678 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3679 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3680 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3681 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3682 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3683 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3684 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3685 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3686 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3687 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3688 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3689 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3690 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3691 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3692 (@aarch64_sve2_xar<mode>): Likewise.
3693 (@aarch64_sve2_bcax<mode>): Likewise.
3694 (*aarch64_sve2_eor3<mode>): Rename to...
3695 (@aarch64_sve2_eor3<mode>): ...this.
3696 (@aarch64_sve2_bsl<mode>): New expander.
3697 (@aarch64_sve2_nbsl<mode>): Likewise.
3698 (@aarch64_sve2_bsl1n<mode>): Likewise.
3699 (@aarch64_sve2_bsl2n<mode>): Likewise.
3700 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3701 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3702 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3703 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3704 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3705 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3706 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3707 (<su>mull<bt><Vwide>): Generalize to...
3708 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3710 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3711 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3712 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3713 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3714 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3715 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3716 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3717 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3718 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3719 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3720 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3721 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3722 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3723 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3724 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3725 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3726 (<SHRNB:r>shrnb<mode>): Generalize to...
3727 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3729 (<SHRNT:r>shrnt<mode>): Generalize to...
3730 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3732 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3733 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3734 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3735 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3736 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3737 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3738 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3739 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3740 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3741 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3742 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3743 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3744 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3745 (@aarch64_sve2_cvtnt<mode>): Likewise.
3746 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3747 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3748 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3749 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3750 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3751 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3752 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3753 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3754 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3755 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3756 (@aarch64_sve2_pmul<mode>): Likewise.
3757 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3758 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3759 (@aarch64_sve2_tbl2<mode>): Likewise.
3760 (@aarch64_sve2_tbx<mode>): Likewise.
3761 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3762 (@aarch64_sve2_histcnt<mode>): Likewise.
3763 (@aarch64_sve2_histseg<mode>): Likewise.
3764 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3765 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3766 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3767 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3768 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3769 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3770 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3771 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3772 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3773 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3774 (SVE2_PMULL_PAIR_I): New mode iterators.
3775 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3776 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3777 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3778 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3779 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3780 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3781 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3782 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3783 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3784 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3785 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3786 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3787 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3788 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3789 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3790 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3791 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3792 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3793 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3794 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3795 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3796 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3797 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3798 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3799 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3800 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3801 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3802 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3803 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3804 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3805 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3807 (VNARROW, Ventype): New mode attributes.
3808 (Vewtype): Handle VNx2DI. Fix typo in comment.
3809 (VDOUBLE): New mode attribute.
3810 (sve_lane_con): Handle VNx8HI.
3811 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3812 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3813 (sve_int_op, sve_int_op_rev): Handle the above codes.
3814 (sve_pred_int_rhs2_operand): Likewise.
3815 (MULLBT, SHRNB, SHRNT): Delete.
3816 (SVE_INT_SHIFT_IMM): New int iterator.
3817 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3818 and UNSPEC_WHILEHS for TARGET_SVE2.
3819 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3820 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3821 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3822 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3823 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3824 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3825 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3826 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3827 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3828 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3829 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3830 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3831 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3832 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3833 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3834 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3835 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3836 (optab): Handle the new unspecs.
3837 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3839 (lr): Handle the new unspecs.
3841 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3842 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3843 (sve_int_qsub_op): New int attributes.
3844 (sve_fp_op, rot): Handle the new unspecs.
3845 * config/aarch64/aarch64-sve-builtins.h
3846 (function_resolver::require_matching_pointer_type): Declare.
3847 (function_resolver::resolve_unary): Add an optional boolean argument.
3848 (function_resolver::finish_opt_n_resolution): Add an optional
3849 type_suffix_index argument.
3850 (gimple_folder::redirect_call): Declare.
3851 (gimple_expander::prepare_gather_address_operands): Add an optional
3853 * config/aarch64/aarch64-sve-builtins.cc: Include
3854 aarch64-sve-builtins-sve2.h.
3855 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3856 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3857 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3858 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3859 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3860 (TYPES_s_unsigned): Likewise.
3861 (TYPES_s_integer): Use TYPES_s_unsigned.
3862 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3863 (TYPES_sd_integer): Use them.
3864 (TYPES_d_unsigned): New macro.
3865 (TYPES_d_integer): Use it.
3866 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3867 (TYPES_cvt_narrow): Likewise.
3868 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3869 (preds_mx): New variable.
3870 (function_builder::add_overloaded_function): Allow the new feature
3871 set to be more restrictive than the original one.
3872 (function_resolver::infer_pointer_type): Remove qualifiers from
3873 the pointer type before printing it.
3874 (function_resolver::require_matching_pointer_type): New function.
3875 (function_resolver::resolve_sv_displacement): Handle functions
3876 that don't support 32-bit vector indices or svint32_t vector offsets.
3877 (function_resolver::finish_opt_n_resolution): Take the inferred type
3878 as a separate argument.
3879 (function_resolver::resolve_unary): Optionally treat all forms in
3880 the same way as normal merging functions.
3881 (gimple_folder::redirect_call): New function.
3882 (function_expander::prepare_gather_address_operands): Add an argument
3883 that says whether scaled forms are available. If they aren't,
3884 handle scaling of vector indices and don't add the extension and
3886 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3887 fall back to using cond_* instead.
3888 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3889 Split out the member variables into...
3890 (rtx_code_function_base): ...this new base class.
3891 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3892 (unspec_based_function): Split out the member variables into...
3893 (unspec_based_function_base): ...this new base class.
3894 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3895 (unspec_based_function_exact_insn): New class.
3896 (unspec_based_add_function, unspec_based_add_lane_function)
3897 (unspec_based_lane_function, unspec_based_pred_function)
3898 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3899 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3900 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3902 (unspec_based_fused_function): New class.
3903 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3904 (unspec_based_fused_lane_function): New class.
3905 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3907 (CODE_FOR_MODE1): New macro.
3908 (fixed_insn_function): New class.
3909 (while_comparison): Likewise.
3910 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3911 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3912 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3913 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3914 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3915 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3916 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3917 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3918 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3919 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3920 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3921 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3922 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3923 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3924 Also add an initial argument for unary_convert_narrowt, regardless
3925 of the predication type.
3926 (build_32_64): Allow loads and stores to specify MODE_none.
3927 (build_sv_index64, build_sv_uint_offset): New functions.
3928 (long_type_suffix): New function.
3929 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3930 (binary_imm_long_base, load_gather_sv_base): Likewise.
3931 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3932 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3933 (unary_narrowb_base, unary_narrowt_base): Likewise.
3934 (binary_long_lane_def, binary_long_lane): New shape.
3935 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3936 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3937 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3938 (binary_to_uint_def, binary_to_uint): Likewise.
3939 (binary_wide_def, binary_wide): Likewise.
3940 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3941 (compare_def, compare): Likewise.
3942 (compare_ptr_def, compare_ptr): Likewise.
3943 (load_ext_gather_index_restricted_def,
3944 load_ext_gather_index_restricted): Likewise.
3945 (load_ext_gather_offset_restricted_def,
3946 load_ext_gather_offset_restricted): Likewise.
3947 (load_gather_sv_def): Inherit from load_gather_sv_base.
3948 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3949 (shift_left_imm_def, shift_left_imm): Likewise.
3950 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3951 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3952 (store_scatter_index_restricted_def,
3953 store_scatter_index_restricted): Likewise.
3954 (store_scatter_offset_restricted_def,
3955 store_scatter_offset_restricted): Likewise.
3956 (tbl_tuple_def, tbl_tuple): Likewise.
3957 (ternary_long_lane_def, ternary_long_lane): Likewise.
3958 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3959 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3960 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3961 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3962 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3963 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3964 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3965 (ternary_uint_def, ternary_uint): Likewise.
3966 (unary_convert): Fix typo in comment.
3967 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3968 (unary_long_def, unary_long): Likewise.
3969 (unary_narrowb_def, unary_narrowb): Likewise.
3970 (unary_narrowt_def, unary_narrowt): Likewise.
3971 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3972 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3973 (unary_to_int_def, unary_to_int): Likewise.
3974 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3975 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3976 (svasrd_impl): Delete.
3977 (svcadd_impl::expand): Handle integer operations too.
3978 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3979 new functions to derive the unspec numbers.
3980 (svmla_svmls_lane_impl): Replace with...
3981 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3982 integer operations too.
3983 (svwhile_impl): Rename to...
3984 (svwhilelx_impl): ...this and inherit from while_comparison.
3985 (svasrd): Use unspec_based_function.
3986 (svmla_lane): Use svmla_lane_impl.
3987 (svmls_lane): Use svmls_lane_impl.
3988 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3989 (svwhilele, svwhilelt): Use svwhilelx_impl.
3990 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3991 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3992 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3993 * config/aarch64/aarch64-sve-builtins.def: Include
3994 aarch64-sve-builtins-sve2.def.
3996 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3998 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3999 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4000 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4001 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4002 immediates as well as vector ones.
4003 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4004 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4005 (aarch64_sve_qsub_immediate): Update calls accordingly.
4007 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4009 * config/aarch64/aarch64-sve2.md: Add banner comments.
4010 (<su>mulh<r>s<mode>3): Move further up file.
4011 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4012 (*aarch64_sve2_sra<mode>): Move further down file.
4013 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4015 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4017 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4019 (while_optab_cmp): Handle them.
4020 * config/aarch64/aarch64-sve.md
4021 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4022 and add a "@" marker.
4023 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4024 instead of gen_aarch64_sve2_while_ptest.
4025 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4027 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4029 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4030 (UNSPEC_WHILELE): ...this.
4031 (UNSPEC_WHILE_LO): Rename to...
4032 (UNSPEC_WHILELO): ...this.
4033 (UNSPEC_WHILE_LS): Rename to...
4034 (UNSPEC_WHILELS): ...this.
4035 (UNSPEC_WHILE_LT): Rename to...
4036 (UNSPEC_WHILELT): ...this.
4037 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4038 (cmp_op, while_optab_cmp): Likewise.
4039 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4040 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4041 (svwhilelt): Likewise.
4043 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4045 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4046 (unary_to_uint): Define.
4047 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4048 (unary_count): Rename to...
4049 (unary_to_uint_def, unary_to_uint): ...this.
4050 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4052 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4054 * config/aarch64/aarch64-sve-builtins-functions.h
4055 (code_for_mode_function): New class.
4056 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4057 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4058 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4059 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4060 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4062 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4064 * config/aarch64/iterators.md (addsub): New code attribute.
4065 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4067 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4068 in the asm string and attributes. Fix indentation.
4069 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4071 (@aarch64_sve_<optab><mode>): ...this.
4072 * config/aarch64/aarch64-sve-builtins.h
4073 (function_expander::expand_signed_unpred_op): Delete.
4074 * config/aarch64/aarch64-sve-builtins.cc
4075 (function_expander::expand_signed_unpred_op): Likewise.
4076 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4077 try using code_for_aarch64_sve instead.
4078 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4079 (svqsub_impl): Likewise.
4080 (svqadd, svqsub): Use rtx_code_function instead.
4082 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4084 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4085 (HADDSUB, sur, addsub): Remove them.
4087 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4089 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4090 functions the same way, rather than singling out those that
4091 aren't mapped directly to optabs.
4093 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4095 * target.def (compatible_vector_types_p): New target hook.
4096 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4097 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4098 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4099 * doc/tm.texi: Regenerate.
4100 * gimple-expr.c: Include target.h.
4101 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4102 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4104 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4105 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4106 Use the original predicate if it already has a suitable type.
4108 2020-01-09 Martin Jambor <mjambor@suse.cz>
4110 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4111 resolve_speculation and redirect_call_stmt_to_callee static. Change
4112 return type of set_call_stmt to cgraph_edge *.
4113 * auto-profile.c (afdo_indirect_call): Adjust call to
4114 redirect_call_stmt_to_callee.
4115 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4116 make the this pointer explicit, adjust self-recursive calls and the
4117 call top make_direct. Return the resulting edge.
4118 (cgraph_edge::remove): Make this pointer explicit.
4119 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4120 (cgraph_edge::make_direct): Likewise, adjust call to
4121 resolve_speculation.
4122 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4123 call to set_call_stmt.
4124 (cgraph_update_edges_for_call_stmt_node): Update call to
4125 set_call_stmt and remove.
4126 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4127 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4128 (cgraph_node::create_edge_including_clones): Moved "first" definition
4129 of edge to the block where it was used. Adjusted calls to
4131 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4132 cgraph_edge::remove.
4133 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4134 make_direct and redirect_call_stmt_to_callee.
4135 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4136 resolve_speculation and make_direct.
4137 * ipa-inline-transform.c (inline_transform): Adjust call to
4138 redirect_call_stmt_to_callee.
4139 (check_speculations_1):: Adjust call to resolve_speculation.
4140 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4141 resolve-speculation.
4142 (inline_small_functions): Adjust call to resolve_speculation.
4143 (ipa_inline): Likewise.
4144 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4146 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4147 safe with regards to edge removal, adjust calls to
4148 redirect_call_stmt_to_callee.
4149 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4150 and redirect_call_stmt_to_callee.
4151 * multiple_target.c (create_dispatcher_calls): Adjust call to
4152 redirect_call_stmt_to_callee
4153 (redirect_to_specific_clone): Likewise.
4154 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4155 Adjust calls to cgraph_edge::remove.
4156 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4157 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4158 (expand_call_inline): Adjust call to cgraph_edge::remove.
4160 2020-01-09 Martin Liska <mliska@suse.cz>
4162 * params.opt: Set Optimization for
4163 param_max_speculative_devirt_maydefs.
4165 2020-01-09 Martin Sebor <msebor@redhat.com>
4169 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4171 2020-01-09 Martin Liska <mliska@suse.cz>
4173 * auto-profile.c (auto_profile): Use opt_for_fn
4175 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4176 (propagate_vals_across_arith_jfunc): Likewise.
4177 (hint_time_bonus): Likewise.
4178 (incorporate_penalties): Likewise.
4179 (good_cloning_opportunity_p): Likewise.
4180 (perform_estimation_of_a_value): Likewise.
4181 (estimate_local_effects): Likewise.
4182 (ipcp_propagate_stage): Likewise.
4183 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4184 (set_switch_stmt_execution_predicate): Likewise.
4185 (analyze_function_body): Likewise.
4186 * ipa-inline-analysis.c (offline_size): Likewise.
4187 * ipa-inline.c (early_inliner): Likewise.
4188 * ipa-prop.c (ipa_analyze_node): Likewise.
4189 (ipcp_transform_function): Likewise.
4190 * ipa-sra.c (process_scan_results): Likewise.
4191 (ipa_sra_summarize_function): Likewise.
4192 * params.opt: Rename ipcp-unit-growth to
4193 ipa-cp-unit-growth. Add Optimization for various
4194 IPA-related parameters.
4196 2020-01-09 Richard Biener <rguenther@suse.de>
4199 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4201 2020-01-09 Richard Biener <rguenther@suse.de>
4203 PR tree-optimization/93040
4204 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4206 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4208 * common/config/avr/avr-common.c (avr_option_optimization_table)
4209 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4211 2020-01-09 Martin Liska <mliska@suse.cz>
4213 * cgraphclones.c (symbol_table::materialize_all_clones):
4214 Use cgraph_node::dump_name.
4216 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4219 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4220 output_operand_lossage instead of gcc_unreachable.
4221 * doc/md.texi (riscv f constraint): Fix typo.
4224 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4225 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4226 CONST_SCALAR_INT_P instead of CONST_INT_P.
4227 (*subv<mode>4_1): Rename to ...
4228 (subv<mode>4_1): ... this.
4229 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4230 define_insn_and_split patterns.
4231 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4234 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4236 * vec.c (class selftest::count_dtor): New class.
4237 (selftest::test_auto_delete_vec): New test.
4238 (selftest::vec_c_tests): Call it.
4239 * vec.h (class auto_delete_vec): New class template.
4240 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4242 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4244 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4246 2020-01-08 Jim Wilson <jimw@sifive.com>
4248 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4249 use of TLS_MODEL_LOCAL_EXEC when not pic.
4251 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4253 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4256 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4259 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4260 *stack_protect_set_3 peephole2): Also check that the second
4261 insns source is general_operand.
4264 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4265 predicate for output operand instead of register_operand.
4266 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4267 memory destination and non-memory operands[2].
4269 2020-01-08 Martin Liska <mliska@suse.cz>
4271 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4272 ::dump_asm_name instead of (::name or ::asm_name).
4273 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4274 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4275 (analyze_functions): Likewise.
4276 (expand_all_functions): Likewise.
4277 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4278 (propagate_bits_across_jump_function): Likewise.
4279 (dump_profile_updates): Likewise.
4280 (ipcp_store_bits_results): Likewise.
4281 (ipcp_store_vr_results): Likewise.
4282 * ipa-devirt.c (dump_targets): Likewise.
4283 * ipa-fnsummary.c (analyze_function_body): Likewise.
4284 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4285 (process_hsa_functions): Likewise.
4286 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4287 (set_alias_uids): Likewise.
4288 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4289 * ipa-inline.c (recursive_inlining): Likewise.
4290 (inline_to_all_callers_1): Likewise.
4291 (ipa_inline): Likewise.
4292 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4293 (ipa_propagate_frequency): Likewise.
4294 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4295 (remove_described_reference): Likewise.
4296 * ipa-pure-const.c (worse_state): Likewise.
4297 (check_retval_uses): Likewise.
4298 (analyze_function): Likewise.
4299 (propagate_pure_const): Likewise.
4300 (propagate_nothrow): Likewise.
4301 (dump_malloc_lattice): Likewise.
4302 (propagate_malloc): Likewise.
4303 (pass_local_pure_const::execute): Likewise.
4304 * ipa-visibility.c (optimize_weakref): Likewise.
4305 (function_and_variable_visibility): Likewise.
4306 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4307 (ipa_discover_variable_flags): Likewise.
4308 * lto-streamer-out.c (output_function): Likewise.
4309 (output_constructor): Likewise.
4310 * tree-inline.c (copy_bb): Likewise.
4311 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4312 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4314 2020-01-08 Richard Biener <rguenther@suse.de>
4317 * tree-eh.c (sink_clobbers): Update virtual operands for
4318 the first and last stmt only. Add a dry-run capability.
4319 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4320 after CFG manipulations and in RPO order to catch all
4321 secondary opportunities reliably.
4323 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4326 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4328 2019-01-08 Richard Biener <rguenther@suse.de>
4331 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4332 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4333 virtual operand, also updating SSA use.
4334 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4335 Update stmt after resetting virtual operand.
4336 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4337 * gimple-iterator.c (gsi_remove): When not removing the stmt
4338 permanently do not delink immediate uses or mark the stmt modified.
4340 2020-01-08 Martin Liska <mliska@suse.cz>
4342 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4343 (ipa_call_context::estimate_size_and_time): Likewise.
4344 (inline_analyze_function): Likewise.
4346 2020-01-08 Martin Liska <mliska@suse.cz>
4348 * cgraph.c (cgraph_node::dump): Use systematically
4351 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4353 Add -nodevicespecs option for avr.
4356 * config/avr/avr.opt (-nodevicespecs): New driver option.
4357 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4358 "-specs=device-specs/..." if that option is not set.
4359 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4361 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4363 Implement 64-bit double functions for avr.
4366 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4367 --with-double-comparison.
4368 * doc/install.texi: Document them.
4369 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4370 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4371 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4372 * doc/invoke.texi (AVR Built-in Macros): Document them.
4373 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4374 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4375 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4377 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4380 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4381 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4382 when only building rm-profile multilibs.
4384 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4387 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4388 lattice for a value to check.
4389 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4390 finite propagation in self-recursive scc.
4392 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4394 * ipa-inline.c (caller_growth_limits): Restore the AND.
4396 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4398 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4399 (VEC_ALLREG_ALT): New iterator.
4400 (VEC_ALLREG_INT_MODE): New iterator.
4401 (VCMP_MODE): New iterator.
4402 (VCMP_MODE_INT): New iterator.
4403 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4404 (vec_cmp<u>v64qidi): New define_expand.
4405 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4406 (vec_cmpu<mode>di_exec): New define_expand.
4407 (vec_cmp<u>v64qidi_exec): New define_expand.
4408 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4409 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4410 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4411 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4412 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4413 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4414 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4415 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4416 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4417 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4419 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4420 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4422 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4424 * config/gcn/constraints.md (DA): Update description and match.
4426 (Db): New constraint.
4427 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4429 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4430 Implement 'Db' mixed immediate type.
4431 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4432 (addcv64si3_dup<exec_vcc>): Delete.
4433 (subcv64si3<exec_vcc>): Rework constraints.
4434 (addv64di3): Rework constraints.
4435 (addv64di3_exec): Rework constraints.
4436 (subv64di3): Rework constraints.
4437 (addv64di3_dup): Delete.
4438 (addv64di3_dup_exec): Delete.
4439 (addv64di3_zext): Rework constraints.
4440 (addv64di3_zext_exec): Rework constraints.
4441 (addv64di3_zext_dup): Rework constraints.
4442 (addv64di3_zext_dup_exec): Rework constraints.
4443 (addv64di3_zext_dup2): Rework constraints.
4444 (addv64di3_zext_dup2_exec): Rework constraints.
4445 (addv64di3_sext_dup2): Rework constraints.
4446 (addv64di3_sext_dup2_exec): Rework constraints.
4448 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4450 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4451 existing target checks.
4453 2020-01-07 Richard Biener <rguenther@suse.de>
4455 * doc/install.texi: Bump minimal supported MPC version.
4457 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4459 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4460 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4461 * langhooks.c: Include stor-layout.h.
4462 (lhd_simulate_enum_decl): New function.
4463 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4464 handle_arm_sve_h for the LTO frontend.
4465 (register_vector_type): Cope with null returns from pushdecl.
4467 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4469 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4470 (aarch64_sve::nvectors_if_data_type): Replace with...
4471 (aarch64_sve::builtin_type_p): ...this.
4472 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4473 (find_vector_type): Delete.
4474 (add_sve_type_attribute): New function.
4475 (lookup_sve_type_attribute): Likewise.
4476 (register_builtin_types): Add an "SVE type" attribute to each type.
4477 (register_tuple_type): Likewise.
4478 (svbool_type_p, nvectors_if_data_type): Delete.
4479 (mangle_builtin_type): Use lookup_sve_type_attribute.
4480 (builtin_type_p): Likewise. Add an overload that returns the
4481 number of constituent vector and predicate registers.
4482 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4483 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4484 instead of aarch64_sve_argument_p.
4485 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4486 (aarch64_pass_by_reference): Likewise.
4487 (aarch64_function_value_1): Likewise.
4488 (aarch64_return_in_memory): Likewise.
4489 (aarch64_layout_arg): Likewise.
4491 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4493 PR tree-optimization/93156
4494 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4495 least significant bit is always clear.
4497 PR tree-optimization/93118
4498 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4499 simplifier with two intermediate conversions.
4501 2020-01-07 Martin Liska <mliska@suse.cz>
4503 * params.opt: Add Optimization for various parameters.
4505 2020-01-07 Martin Liska <mliska@suse.cz>
4508 * doc/extend.texi: Explain cloning for target_clone
4511 2020-01-07 Martin Liska <mliska@suse.cz>
4513 PR tree-optimization/92860
4514 * common.opt: Make in Optimization option
4515 as it is affected by -O0, which is an Optimization
4517 * tree-inline.c (tree_inlinable_function_p):
4518 Use opt_for_fn for warn_inline.
4519 (expand_call_inline): Likewise.
4521 2020-01-07 Martin Liska <mliska@suse.cz>
4523 PR tree-optimization/92860
4524 * common.opt: Make flag_ree as optimization
4527 2020-01-07 Martin Liska <mliska@suse.cz>
4529 PR optimization/92860
4530 * params.opt: Mark param_min_crossjump_insns with Optimization
4533 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4535 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4536 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4538 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4540 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4541 helper function to return the valid addressing formats for a given
4542 hard register and mode.
4543 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4545 * config/rs6000/constraints.md (Q constraint): Update
4547 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4550 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4551 Use 'Q' for doing vector extract from memory.
4552 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4554 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4555 doing vector extract from memory.
4556 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4557 extract from memory.
4559 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4560 for the offset being 34-bits when -mcpu=future is used.
4562 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4564 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4565 instead of cmpib_comparison_operator in cmpib patterns.
4566 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4567 of cmpib_comparison_operator. Revise comment.
4569 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4571 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4572 in an IFN_DIV_POW2 node to be equal.
4574 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4576 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4577 (vect_check_scalar_mask): ...this.
4578 (vectorizable_store, vectorizable_load): Update call accordingly.
4579 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4580 argument in calls to conditional internal functions.
4582 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4584 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4585 '0' matching inputs.
4586 (subv64di3_exec): Likewise.
4588 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4590 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4591 * doc/md.texi (movstr): Likewise.
4593 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4595 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4598 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4600 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4602 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4603 to a temporary file and use move-if-change to update the real
4604 file where necessary.
4606 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4608 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4609 rather than Upa for CPY /M.
4611 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4613 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4616 2020-01-06 Martin Liska <mliska@suse.cz>
4618 PR tree-optimization/92860
4619 * params.opt: Mark param_max_combine_insns with Optimization
4622 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4625 * config/i386/i386.md (SWIDWI): New mode iterator.
4626 (DWI, dwi): Add TImode variants.
4627 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4628 <general_hilo_operand> instead of <general_operand>. Use
4629 CONST_SCALAR_INT_P instead of CONST_INT_P.
4630 (*addv<mode>4_1): Rename to ...
4631 (addv<mode>4_1): ... this.
4632 (QWI): New mode attribute.
4633 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4634 define_insn_and_split patterns.
4635 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4637 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4638 <general_hilo_operand> instead of <general_operand>.
4639 (*addcarry<mode>_1): New define_insn.
4640 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4642 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4644 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4645 Use "call" instead of "set".
4647 2020-01-03 Martin Jambor <mjambor@suse.cz>
4650 * ipa-cp.c (print_all_lattices): Skip functions without info.
4652 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4655 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4656 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4657 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4658 for 'e' simd clones.
4661 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4663 (mprefer-vector-width=): Add Save.
4664 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4665 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4666 (ix86_debug_options, ix86_function_specific_print): Adjust
4667 ix86_target_string callers.
4668 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4669 (ix86_valid_target_attribute_tree): Likewise.
4670 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4671 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4672 ix86_target_string caller.
4675 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4676 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4677 instead of gen_int_shift_amount + convert_modes.
4679 PR rtl-optimization/93088
4680 * loop-iv.c (find_single_def_src): Punt after looking through
4681 128 reg copies for regs with single definitions. Move definitions
4684 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4686 * config/arm/arm-c.c (arm_cpu_builtins): Define
4687 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4688 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4689 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4690 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4691 * config/arm/arm-tables.opt: Regenerated.
4692 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4693 arm_arch_i8mm and arm_arch_bf16 when enabled.
4694 * config/arm/arm.h (TARGET_I8MM): New macro.
4695 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4696 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4697 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4698 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4699 (v8_6_a_simd_variants): New.
4700 (v8_*_a_simd_variants): Add i8mm and bf16.
4701 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4703 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4706 * predict.c (compute_function_frequency): Don't call
4707 warn_function_cold on functions that already have cold attribute.
4709 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4712 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4713 COMDAT group function labels in .data.rel.ro.local section.
4714 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4717 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4718 comparison_operator in B and S integer comparisons. Likewise, use
4719 ordered_comparison_operator instead of cmpib_comparison_operator in
4721 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4723 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4725 Update copyright years.
4727 * gcc.c (process_command): Update copyright notice dates.
4728 * gcov-dump.c (print_version): Ditto.
4729 * gcov.c (print_version): Ditto.
4730 * gcov-tool.c (print_version): Ditto.
4731 * gengtype.c (create_file): Ditto.
4732 * doc/cpp.texi: Bump @copying's copyright year.
4733 * doc/cppinternals.texi: Ditto.
4734 * doc/gcc.texi: Ditto.
4735 * doc/gccint.texi: Ditto.
4736 * doc/gcov.texi: Ditto.
4737 * doc/install.texi: Ditto.
4738 * doc/invoke.texi: Ditto.
4740 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4742 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4745 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4747 PR tree-optimization/93098
4748 * match.pd (popcount): For shift amounts, use integer_onep
4749 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4750 tests. Make sure that precision is power of two larger than or equal
4751 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4752 instead of ULL suffixed constants. Formatting fixes.
4754 Copyright (C) 2020 Free Software Foundation, Inc.
4756 Copying and distribution of this file, with or without modification,
4757 are permitted in any medium without royalty provided the copyright
4758 notice and this notice are preserved.