1 2020-02-25 Martin Sebor <msebor@redhat.com>
3 * doc/extend.texi (attribute access): Correct an example.
5 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
7 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
9 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
11 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
12 (VD): Enable for V4BF.
14 (VQ): Enable for V8BF.
17 (VDBL, Vdbl): Add V4BF.
18 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
19 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
20 (bfloat16x8x2_t): Likewise.
21 (bfloat16x4x3_t): Likewise.
22 (bfloat16x8x3_t): Likewise.
23 (bfloat16x4x4_t): Likewise.
24 (bfloat16x8x4_t): Likewise.
26 (vld1_bf16, vld1_bf16_x2): New.
27 (vld1_bf16_x3, vld1_bf16_x4): New.
28 (vld1q_bf16, vld1q_bf16_x2): New.
29 (vld1q_bf16_x3, vld1q_bf16_x4): New.
30 (vld1_lane_bf16): New.
31 (vld1q_lane_bf16): New.
33 (vld1q_dup_bf16): New.
37 (vld2q_dup_bf16): New.
41 (vld3q_dup_bf16): New.
45 (vld4q_dup_bf16): New.
46 (vst1_bf16, vst1_bf16_x2): New.
47 (vst1_bf16_x3, vst1_bf16_x4): New.
48 (vst1q_bf16, vst1q_bf16_x2): New.
49 (vst1q_bf16_x3, vst1q_bf16_x4): New.
50 (vst1_lane_bf16): New.
51 (vst1q_lane_bf16): New.
59 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
61 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
63 (VALLDI_F16): Likewise.
66 (vswap_width_name): Likewise.
67 (VSWAP_WIDTH): Likewise.
71 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
72 (vget_lane_bf16, vgetq_lane_bf16): New.
74 (vdup_n_bf16, vdupq_n_bf16): New.
75 (vdup_lane_bf16, vdup_laneq_bf16): New.
76 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
77 (vduph_lane_bf16, vduph_laneq_bf16): New.
78 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
79 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
80 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
81 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
82 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
83 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
84 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
85 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
86 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
87 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
88 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
89 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
90 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
91 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
92 (vreinterpretq_bf16_p128): New.
93 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
94 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
95 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
96 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
97 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
98 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
99 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
100 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
101 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
102 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
103 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
104 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
105 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
106 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
107 (vreinterpretq_p128_bf16): New.
109 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
111 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
112 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
113 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
114 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
115 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
116 * config/arm/iterators.md (VSF2BF): New attribute.
117 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
118 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
119 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
121 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
123 * config/arm/arm.md (required_for_purecode): New attribute.
124 (enabled): Handle required_for_purecode.
125 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
126 work with -mpure-code.
128 2020-02-25 Jakub Jelinek <jakub@redhat.com>
130 PR rtl-optimization/93908
131 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
134 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
136 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
138 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
140 * doc/install.texi (--enable-checking): Adjust wording.
142 2020-02-25 Richard Biener <rguenther@suse.de>
144 PR tree-optimization/93868
145 * tree-vect-slp.c (slp_copy_subtree): New function.
146 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
147 re-arranging stmts in it.
149 2020-02-25 Jakub Jelinek <jakub@redhat.com>
152 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
153 dummy function and remove it at the end.
156 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
157 paramter -> parameter.
158 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
159 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
161 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
163 * doc/install.texi (--enable-checking): Properly document current
165 (--enable-stage1-checking): Minor clarification about bootstrap.
167 2020-02-24 David Malcolm <dmalcolm@redhat.com>
170 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
171 -fanalyzer-checker=taint is also required.
172 (-fanalyzer-checker=): Note that providing this option enables the
173 given checker, and doing so may be required for checkers that are
176 2020-02-24 David Malcolm <dmalcolm@redhat.com>
178 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
179 significant control flow events; add a "3" which shows all
180 control flow events; the old "3" becomes "4".
182 2020-02-24 Jakub Jelinek <jakub@redhat.com>
184 PR tree-optimization/93582
185 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
186 pd.offset and pd.size to be counted in bits rather than bytes, add
187 support for maxsizei that is not a multiple of BITS_PER_UNIT and
188 handle bitfield stores and loads.
189 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
190 uncomparable quantities - bytes vs. bits. Allow push_partial_def
191 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
192 pd.offset/pd.size to be counted in bits rather than bytes.
193 Formatting fix. Rename shadowed len variable to buflen.
195 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
196 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
199 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
200 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
201 * opts-common.c (parse_options_from_collect_gcc_options): New function.
202 (prepend_xassembler_to_collect_as_options): Likewise.
203 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
204 (prepend_xassembler_to_collect_as_options): Likewise.
205 * lto-opts.c (lto_write_options): Stream assembler options
206 in COLLECT_AS_OPTIONS.
207 * lto-wrapper.c (xassembler_options_error): New static variable.
208 (get_options_from_collect_gcc_options): Move parsing options code to
209 parse_options_from_collect_gcc_options and call it.
210 (merge_and_complain): Validate -Xassembler options.
211 (append_compiler_options): Handle OPT_Xassembler.
212 (run_gcc): Append command line -Xassembler options to
214 * doc/invoke.texi: Add documentation about using Xassembler
217 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
219 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
221 (riscv_rtx_costs): Update cost model for LTGT.
223 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
225 PR rtl-optimization/93564
226 * ira-color.c (struct update_cost_queue_elem): New member start.
227 (queue_update_cost, get_next_update_cost): Add new arg start.
228 (allocnos_conflict_p): New function.
229 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
230 Add checking conflicts with allocnos_conflict_p.
231 (update_costs_from_prefs, restore_costs_from_copies): Adjust
232 update_costs_from_allocno calls.
233 (update_conflict_hard_regno_costs): Add checking conflicts with
234 allocnos_conflict_p. Adjust calls of queue_update_cost and
235 get_next_update_cost.
236 (assign_hard_reg): Adjust calls of queue_update_cost. Add
238 (bucket_allocno_compare_func): Restore previous version.
240 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
242 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
243 double-word size when handling aggregate return values.
244 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
245 that homogeneous SFmode and DFmode aggregates are passed and returned
246 in general registers.
248 2020-02-21 Jakub Jelinek <jakub@redhat.com>
251 * opts.c (print_filtered_help): Translate help before appending
252 messages to it rather than after that.
254 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
256 PR rtl-optimization/PR92989
257 * lra-lives.c (process_bb_lives): Restore the original order
258 of the bb liveness update. Call make_hard_regno_dead for each
259 register clobbered at the start of an EH receiver.
261 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
264 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
265 self-recursively generated.
267 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
270 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
273 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
275 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
276 Document new target supports option.
278 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
280 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
281 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
282 * config/arm/iterators.md (MATMUL): New iterator.
283 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
284 (mmla_sfx): New attribute.
285 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
286 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
287 (UNSPEC_MATMUL_US): New.
289 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
291 * config/arm/arm.md: Prevent scalar shifts from being used when big
294 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
295 Richard Biener <rguenther@suse.de>
297 PR tree-optimization/93586
298 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
299 after mismatched array refs; do not sure type size information to
300 recover from unmatched referneces with !flag_strict_aliasing_p.
302 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
304 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
305 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
306 (scatter_store<mode>): Rename to ...
307 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
308 (scatter<mode>_exec): Delete. Move contents ...
309 (mask_scatter_store<mode>): ... here, and rename that to ...
310 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
311 Remove mode conversion.
312 (mask_gather_load<mode>): Rename to ...
313 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
314 Remove mode conversion.
315 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
317 2020-02-21 Martin Jambor <mjambor@suse.cz>
319 PR tree-optimization/93845
320 * tree-sra.c (verify_sra_access_forest): Only test access size of
323 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
325 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
326 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
327 (addv64di3_exec): Likewise.
328 (subv64di3): Likewise.
329 (subv64di3_exec): Likewise.
330 (addv64di3_zext): Likewise.
331 (addv64di3_zext_exec): Likewise.
332 (addv64di3_zext_dup): Likewise.
333 (addv64di3_zext_dup_exec): Likewise.
334 (addv64di3_zext_dup2): Likewise.
335 (addv64di3_zext_dup2_exec): Likewise.
336 (addv64di3_sext_dup2): Likewise.
337 (addv64di3_sext_dup2_exec): Likewise.
338 (<expander>v64di3): Likewise.
339 (<expander>v64di3_exec): Likewise.
340 (*<reduc_op>_dpp_shr_v64di): Likewise.
341 (*plus_carry_dpp_shr_v64di): Likewise.
342 * config/gcn/gcn.md (adddi3): Likewise.
343 (addptrdi3): Likewise.
344 (<expander>di3): Likewise.
346 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
348 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
350 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
352 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
353 support. Use aarch64_emit_mult instead of emitting multiplication
354 instructions directly.
355 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
356 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
358 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
360 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
361 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
362 instead of emitting multiplication instructions directly.
363 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
364 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
365 (@aarch64_frecps<mode>): New expanders.
367 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
369 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
370 on and produce uint64_ts rather than ints.
371 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
372 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
374 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
376 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
377 an unused xmsk register when handling approximate rsqrt.
379 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
381 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
382 flag_finite_math_only condition.
384 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
387 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
388 to destination operand for shufps alternative.
389 (*vec_extractv2si_1): Ditto.
391 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
394 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
397 2020-02-20 Martin Liska <mliska@suse.cz>
400 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
402 2020-02-20 Martin Liska <mliska@suse.cz>
405 * common/config/avr/avr-common.c: Remote trailing "|".
407 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
409 * collect2.c (maybe_run_lto_and_relink): Fix typo in
412 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
414 PR tree-optimization/93767
415 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
416 access-size bias from the offset calculations for negative strides.
418 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
420 * collect2.c (c_file, o_file): Make const again.
421 (ldout,lderrout, dump_ld_file): Remove.
422 (tool_cleanup): Avoid calling not signal-safe functions.
423 (maybe_run_lto_and_relink): Avoid possible signal handler
424 access to unintialzed memory (lto_o_files).
425 (main): Avoid leaking temp files in $TMPDIR.
426 Initialize c_file/o_file with concat, which avoids exposing
427 uninitialized memory to signal handler, which calls unlink(!).
428 Avoid calling maybe_unlink when the main function returns,
429 since the atexit handler is already doing this.
430 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
432 2020-02-19 Martin Jambor <mjambor@suse.cz>
434 PR tree-optimization/93776
435 * tree-sra.c (create_access): Do not create zero size accesses.
436 (get_access_for_expr): Do not search for zero sized accesses.
438 2020-02-19 Martin Jambor <mjambor@suse.cz>
440 PR tree-optimization/93667
441 * tree-sra.c (scalarizable_type_p): Return false if record fields
442 do not follow wach other.
444 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
446 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
447 rather than fmv.x.s/fmv.s.x.
449 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
451 * config/aarch64/aarch64-simd-builtins.def
452 (intrinsic_vec_smult_lo_): New.
453 (intrinsic_vec_umult_lo_): Likewise.
454 (vec_widen_smult_hi_): Likewise.
455 (vec_widen_umult_hi_): Likewise.
456 * config/aarch64/aarch64-simd.md
457 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
458 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
459 (vmull_high_s16): Likewise.
460 (vmull_high_s32): Likewise.
461 (vmull_high_u8): Likewise.
462 (vmull_high_u16): Likewise.
463 (vmull_high_u32): Likewise.
464 (vmull_s8): Likewise.
465 (vmull_s16): Likewise.
466 (vmull_s32): Likewise.
467 (vmull_u8): Likewise.
468 (vmull_u16): Likewise.
469 (vmull_u32): Likewise.
471 2020-02-18 Martin Liska <mliska@suse.cz>
473 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
474 bootstrap by missing removal of invalid sanity check.
476 2020-02-18 Martin Liska <mliska@suse.cz>
479 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
480 Always compare LHS of gimple_assign.
482 2020-02-18 Martin Liska <mliska@suse.cz>
485 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
486 and return type of functions.
487 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
488 Drop MALLOC attribute for void functions.
489 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
490 malloc_state for a new VOID clone.
492 2020-02-18 Martin Liska <mliska@suse.cz>
495 * common.opt: Add -fprofile-reproducibility.
496 * doc/invoke.texi: Document it.
497 * value-prof.c (dump_histogram_value):
498 Document and support behavior for counters[0]
499 being a negative value.
500 (get_nth_most_common_value): Handle negative
501 counters[0] in respect to flag_profile_reproducible.
503 2020-02-18 Jakub Jelinek <jakub@redhat.com>
506 * cgraph.c (verify_speculative_call): Use speculative_id instead of
507 speculative_uid in messages. Remove trailing whitespace from error
508 message. Use num_speculative_call_targets instead of
509 num_speculative_targets in a message.
510 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
511 edge messages and stmt instead of cal_stmt in reference message.
513 PR tree-optimization/93780
514 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
515 before calling build_vector_type.
516 (execute_update_addresses_taken): Likewise.
519 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
520 typo, functoin -> function.
521 * tree.c (free_lang_data_in_decl): Fix comment typo,
522 functoin -> function.
523 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
525 2020-02-17 David Malcolm <dmalcolm@redhat.com>
527 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
529 (print_option_information): Don't call get_option_url if URLs
532 2020-02-17 Alexandre Oliva <oliva@adacore.com>
534 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
535 handling of register_common-less targets.
537 2020-02-17 Martin Liska <mliska@suse.cz>
540 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
542 2020-02-17 Martin Liska <mliska@suse.cz>
545 * config/rs6000/rs6000.c (rs6000_option_override_internal):
548 2020-02-17 Martin Liska <mliska@suse.cz>
551 * config/rx/elf.opt: Fix typo.
553 2020-02-17 Richard Biener <rguenther@suse.de>
556 * opts-global.c (print_ignored_options): Use inform and
559 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
562 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
564 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
567 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
568 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
570 2020-02-15 Jason Merrill <jason@redhat.com>
572 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
574 2020-02-15 Jakub Jelinek <jakub@redhat.com>
576 PR tree-optimization/93744
577 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
578 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
579 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
580 sure @2 in the first and @1 in the other patterns has no side-effects.
582 2020-02-15 David Malcolm <dmalcolm@redhat.com>
583 Bernd Edlinger <bernd.edlinger@hotmail.de>
587 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
588 * configure.ac (--with-diagnostics-urls): New configuration
589 option, based on --with-diagnostics-color.
590 (DIAGNOSTICS_URLS_DEFAULT): New define.
591 * config.h: Regenerate.
592 * configure: Regenerate.
593 * diagnostic.c (diagnostic_urls_init): Handle -1 for
594 DIAGNOSTICS_URLS_DEFAULT from configure-time
595 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
596 and TERM_URLS environment variable.
597 * diagnostic-url.h (diagnostic_url_format): New enum type.
598 (diagnostic_urls_enabled_p): rename to...
599 (determine_url_format): ... this, and change return type.
600 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
601 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
602 the linux console, and mingw.
603 (diagnostic_urls_enabled_p): rename to...
604 (determine_url_format): ... this, and adjust.
605 * pretty-print.h (pretty_printer::show_urls): rename to...
606 (pretty_printer::url_format): ... this, and change to enum.
607 * pretty-print.c (pretty_printer::pretty_printer,
608 pp_begin_url, pp_end_url, test_urls): Adjust.
609 * doc/install.texi (--with-diagnostics-urls): Document the new
610 configuration option.
611 (--with-diagnostics-color): Document the existing interaction
612 with GCC_COLORS better.
613 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
614 vindex reference. Update description of defaults based on the above.
615 (-fdiagnostics-color): Update description of how -fdiagnostics-color
616 interacts with GCC_COLORS.
618 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
621 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
622 conjunction with TARGET_GNU_TLS in early return.
624 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
626 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
627 the mode is not wider than UNITS_PER_WORD.
629 2020-02-14 Martin Jambor <mjambor@suse.cz>
631 PR tree-optimization/93516
632 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
633 access of the same type as the parent.
634 (propagate_subaccesses_from_lhs): Likewise.
636 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
639 * config/i386/avx512vbmi2intrin.h
640 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
641 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
642 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
643 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
644 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
645 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
646 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
647 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
648 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
649 of lacking a closing parenthesis.
650 * config/i386/avx512vbmi2vlintrin.h
651 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
652 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
653 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
654 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
655 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
656 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
657 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
658 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
659 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
660 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
661 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
662 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
663 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
664 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
665 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
666 _mm_shldi_epi32, _mm_mask_shldi_epi32,
667 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
668 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
670 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
673 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
674 the target function entry.
676 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
678 * common/config/arc/arc-common.c (arc_option_optimization_table):
679 Disable if-conversion step when optimized for size.
681 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
683 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
684 R12-R15 are always in ARCOMPACT16_REGS register class.
685 * config/arc/arc.opt (mq-class): Deprecate.
686 * config/arc/constraint.md ("q"): Remove dependency on mq-class
688 * doc/invoke.texi (mq-class): Update text.
689 * common/config/arc/arc-common.c (arc_option_optimization_table):
692 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
694 * config/arc/arc.c (arc_insn_cost): New function.
695 (TARGET_INSN_COST): Define.
696 * config/arc/arc.md (cost): New attribute.
697 (add_n): Use arc_nonmemory_operand.
698 (ashlsi3_insn): Likewise, also update constraints.
699 (ashrsi3_insn): Likewise.
701 (add_shift): Likewise.
702 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
704 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
706 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
708 (umulsidi_600): Likewise.
710 2020-02-13 Jakub Jelinek <jakub@redhat.com>
713 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
714 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
715 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
716 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
717 pass __A to the builtin followed by __W instead of __A followed by
719 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
720 _mm512_mask_popcnt_epi64): Likewise.
721 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
722 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
723 _mm256_mask_popcnt_epi64): Likewise.
725 PR tree-optimization/93582
726 * fold-const.h (shift_bytes_in_array_left,
727 shift_bytes_in_array_right): Declare.
728 * fold-const.c (shift_bytes_in_array_left,
729 shift_bytes_in_array_right): New function, moved from
730 gimple-ssa-store-merging.c, no longer static.
731 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
732 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
733 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
734 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
735 shift_bytes_in_array.
736 (verify_shift_bytes_in_array): Rename to ...
737 (verify_shift_bytes_in_array_left): ... this. Use
738 shift_bytes_in_array_left instead of shift_bytes_in_array.
739 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
740 instead of verify_shift_bytes_in_array.
741 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
742 / native_interpret_expr where the store covers all needed bits,
743 punt on PDP-endian, otherwise allow all involved offsets and sizes
744 not to be byte-aligned.
747 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
748 use const_0_to_255_operand predicate instead of immediate_operand.
749 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
750 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
751 vgf2p8affineinvqb_<mode><mask_name>,
752 vgf2p8affineqb_<mode><mask_name>): Drop mode from
753 const_0_to_255_operand predicated operands.
755 2020-02-12 Jeff Law <law@redhat.com>
757 * config/h8300/h8300.md (comparison shortening peepholes): Use
758 a mode iterator to merge the HImode and SImode peepholes.
760 2020-02-12 Jakub Jelinek <jakub@redhat.com>
763 * real.c (is_even): Make static. Function comment fix.
764 (is_halfway_below): Make static, don't assert R is not inf/nan,
765 instead return false for those. Small formatting fixes.
767 2020-02-12 Martin Sebor <msebor@redhat.com>
770 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
771 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
772 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
773 (strlen_check_and_optimize_call): Adjust callee name.
775 2020-02-12 Jeff Law <law@redhat.com>
777 * config/h8300/h8300.md (comparison shortening peepholes): Drop
778 (and (xor)) variant. Combine other two into single peephole.
780 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
782 PR rtl-optimization/93565
783 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
785 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
787 * config/aarch64/aarch64-simd.md
788 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
789 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
790 generating separate ADDV and zero_extend patterns.
791 * config/aarch64/iterators.md (VDQV_E): New iterator.
793 2020-02-12 Jeff Law <law@redhat.com>
795 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
796 expanders, splits, etc.
797 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
798 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
799 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
800 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
801 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
802 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
804 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
806 2020-02-12 Jakub Jelinek <jakub@redhat.com>
809 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
810 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
811 TARGET_AVX512DQ from condition.
812 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
813 instead of <mask_mode512bit_condition> in condition. If
814 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
816 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
819 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
822 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
824 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
826 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
827 where strlen is more legible.
828 (rs6000_builtin_vectorized_libmass): Ditto.
829 (rs6000_print_options_internal): Ditto.
831 2020-02-11 Martin Sebor <msebor@redhat.com>
833 PR tree-optimization/93683
834 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
836 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
838 * config/rs6000/predicates.md (cint34_operand): Rename the
839 -mprefixed-addr option to be -mprefixed.
840 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
841 the -mprefixed-addr option to be -mprefixed.
842 (OTHER_FUTURE_MASKS): Likewise.
843 (POWERPC_MASKS): Likewise.
844 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
845 the -mprefixed-addr option to be -mprefixed. Change error
846 messages to refer to -mprefixed.
847 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
849 (rs6000_legitimate_offset_address_p): Likewise.
850 (rs6000_mode_dependent_address): Likewise.
851 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
852 "-mprefixed" for target attributes and pragmas.
853 (address_to_insn_form): Rename the -mprefixed-addr option to be
855 (rs6000_adjust_insn_length): Likewise.
856 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
857 -mprefixed-addr option to be -mprefixed.
858 (ASM_OUTPUT_OPCODE): Likewise.
859 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
860 -mprefixed-addr option to be -mprefixed.
861 * config/rs6000/rs6000.opt (-mprefixed): Rename the
862 -mprefixed-addr option to be prefixed. Change the option from
863 being undocumented to being documented.
864 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
865 -mprefixed option. Update the -mpcrel documentation to mention
868 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
870 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
871 including FIRST_PSEUDO_REGISTER - 1.
872 * ira-color.c (print_hard_reg_set): Ditto.
874 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
876 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
877 (USTERNOP_QUALIFIERS): New define.
878 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
879 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
880 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
881 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
882 * config/arm/arm_neon.h (vusdot_s32): New.
883 (vusdot_lane_s32): New.
884 (vusdotq_lane_s32): New.
885 (vsudot_lane_s32): New.
886 (vsudotq_lane_s32): New.
887 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
888 * config/arm/iterators.md (DOTPROD_I8MM): New.
889 (sup, opsuffix): Add <us/su>.
890 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
891 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
893 2020-02-11 Richard Biener <rguenther@suse.de>
895 PR tree-optimization/93661
896 PR tree-optimization/93662
897 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
899 * tree-sra.c (get_access_for_expr): Likewise.
901 2020-02-10 Jakub Jelinek <jakub@redhat.com>
904 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
905 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
906 Change condition from TARGET_AVX2 to TARGET_AVX.
908 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
911 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
914 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
916 Try to generate zero-based comparisons.
917 * config/cris/cris.c (cris_reduce_compare): New function.
918 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
919 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
920 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
922 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
925 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
926 in Thumb state and also as a destination in Arm state. Add T16
929 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
931 * md.texi (Define Subst): Match closing paren in example.
933 2020-02-10 Jakub Jelinek <jakub@redhat.com>
937 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
938 arguments of strncmp.
940 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
943 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
944 but different source value.
945 (adjust_callers_for_value_intersection): New function.
946 (gather_edges_for_value): Adjust order of callers to let a
947 non-self-recursive caller be the first element.
948 (self_recursive_pass_through_p): Add a new parameter "simple", and
949 check generalized self-recursive pass-through jump function.
950 (self_recursive_agg_pass_through_p): Likewise.
951 (find_more_scalar_values_for_callers_subset): Compute value from
952 pass-through jump function for self-recursive.
953 (intersect_with_plats): Cleanup previous implementation code for value
954 itersection with self-recursive call edge.
955 (intersect_with_agg_replacements): Likewise.
956 (intersect_aggregates_with_edge): Deduce value from pass-through jump
957 function for self-recursive call edge. Cleanup previous implementation
958 code for value intersection with self-recursive call edge.
959 (decide_whether_version_node): Remove dead callers and adjust order
960 to let a non-self-recursive caller be the first element.
962 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
964 * recog.c: Move pass_split_before_sched2 code in front of
965 pass_split_before_regstack.
966 (pass_data_split_before_sched2): Rename pass to split3 from split4.
967 (pass_data_split_before_regstack): Rename pass to split4 from split3.
968 (rest_of_handle_split_before_sched2): Remove.
969 (pass_split_before_sched2::execute): Unconditionally call
971 (enable_split_before_sched2): New function.
972 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
973 (pass_split_before_regstack::gate): Ditto.
974 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
975 Update name check for renamed split4 pass.
976 * config/sh/sh.c (register_sh_passes): Update pass insertion
977 point for renamed split4 pass.
979 2020-02-09 Jakub Jelinek <jakub@redhat.com>
981 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
982 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
983 copying them around between host and target.
985 2020-02-08 Andrew Pinski <apinski@marvell.com>
988 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
989 STRICT_ALIGNMENT also.
991 2020-02-08 Jim Wilson <jimw@sifive.com>
994 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
996 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
997 Jakub Jelinek <jakub@redhat.com>
1000 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1001 xmm16-xmm31 call-used even in 64-bit ms-abi.
1003 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1005 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1006 (simd_ummla, simd_usmmla): Likewise.
1007 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1008 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1009 (vusmmlaq_s32): New.
1011 2020-02-07 Richard Biener <rguenther@suse.de>
1014 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1015 skipping unreachable regions.
1016 (optimize_inline_calls): Skip folding stmts when we didn't
1019 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1022 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1023 Don't return aggregates with only SFmode and DFmode in SSE
1025 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1027 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1030 * config/rs6000/rs6000-logue.c
1031 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1032 if it fails, move rs into end_addr and retry. Add
1033 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1034 the insn pattern doesn't describe well what exactly happens to
1038 * config/i386/predicates.md (avx_identity_operand): Remove.
1039 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1040 (avx_<castmode><avxsizesuffix>_<castmode>,
1041 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1042 a VEC_CONCAT of the operand and UNSPEC_CAST.
1043 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1044 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1048 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1049 recog_data.insn if distance_non_agu_define changed it.
1051 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1054 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1055 we only had X-FORM (reg+reg) addressing for vectors. Also before
1056 ISA 3.0, we only had X-FORM addressing for scalars in the
1057 traditional Altivec registers.
1059 2020-02-06 <zhongyunde@huawei.com>
1060 Vladimir Makarov <vmakarov@redhat.com>
1062 PR rtl-optimization/93561
1063 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1064 hard register range.
1066 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1068 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1071 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1073 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1074 where the low and the high 32 bits are equal to each other specially,
1075 with an rldimi instruction.
1077 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1079 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1081 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1083 * config/arm/arm-tables.opt: Regenerate.
1085 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1088 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1089 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1090 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1092 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1094 PR rtl-optimization/87763
1095 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1097 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1099 * config/aarch64/aarch64-simd-builtins.def
1100 (bfmlaq): New built-in function.
1101 (bfmlalb): New built-in function.
1102 (bfmlalt): New built-in function.
1103 (bfmlalb_lane): New built-in function.
1104 (bfmlalt_lane): New built-in function.
1105 * config/aarch64/aarch64-simd.md
1106 (aarch64_bfmmlaqv4sf): New pattern.
1107 (aarch64_bfmlal<bt>v4sf): New pattern.
1108 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1109 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1110 (vbfmlalbq_f32): New intrinsic.
1111 (vbfmlaltq_f32): New intrinsic.
1112 (vbfmlalbq_lane_f32): New intrinsic.
1113 (vbfmlaltq_lane_f32): New intrinsic.
1114 (vbfmlalbq_laneq_f32): New intrinsic.
1115 (vbfmlaltq_laneq_f32): New intrinsic.
1116 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1117 (bt): New int attribute.
1119 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1121 * config/i386/i386.md (*pushtf): Emit "#" instead of
1122 calling gcc_unreachable in insn output.
1125 (*pushsf_rex64): Ditto for alternatives other than 1.
1126 (*pushsf): Ditto for alternatives other than 1.
1128 2020-02-06 Martin Liska <mliska@suse.cz>
1130 PR gcov-profile/91971
1131 PR gcov-profile/93466
1132 * coverage.c (coverage_init): Revert mangling of
1133 path into filename. It can lead to huge filename length.
1134 Creation of subfolders seem more natural.
1136 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1139 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1140 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1141 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1143 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1146 * config/i386/predicates.md (avx_identity_operand): New predicate.
1147 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1148 define_insn_and_split.
1151 * omp-low.c (use_pointer_for_field): For nested constructs, also
1152 look for map clauses on target construct.
1153 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1154 taskreg_nesting_level.
1157 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1158 shared clause, call omp_notice_variable on outer context if any.
1160 2020-02-05 Jason Merrill <jason@redhat.com>
1163 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1164 non-zero address even if weak and not yet defined.
1166 2020-02-05 Martin Sebor <msebor@redhat.com>
1168 PR tree-optimization/92765
1169 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1170 * tree-ssa-strlen.c (compute_string_length): Remove.
1171 (determine_min_objsize): Remove.
1172 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1173 Avoid using type size as the upper bound on string length.
1174 (handle_builtin_string_cmp): Add an argument. Adjust.
1175 (strlen_check_and_optimize_call): Pass additional argument to
1176 handle_builtin_string_cmp.
1178 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1180 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1181 (*pushdi2_rex64 peephole2): Unconditionally split after
1183 (*ashl<mode>3_doubleword): Ditto.
1184 (*<shift_insn><mode>3_doubleword): Ditto.
1186 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1189 * config/rs6000/rs6000.c (get_vector_offset): Fix
1191 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1193 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1195 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1198 (Special Functions for Debugging the Analyzer): Update description
1199 of __analyzer_dump_exploded_nodes.
1201 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1204 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1205 include sets and not clobbers in the vzeroupper pattern.
1206 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1207 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1208 (*avx_vzeroupper_1): New define_insn_and_split.
1211 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1212 don't run when !optimize.
1213 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1216 2020-02-05 Richard Biener <rguenther@suse.de>
1219 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1220 checks before matching calls.
1222 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1224 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1225 function comment typo.
1228 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1229 simd_clone_create failed when i == 0, adjust clone->nargs by
1232 2020-02-05 Martin Liska <mliska@suse.cz>
1235 * doc/invoke.texi: Document that one should
1236 not combine ASLR and -fpch.
1238 2020-02-04 Richard Biener <rguenther@suse.de>
1240 PR tree-optimization/93538
1241 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1243 2020-02-04 Richard Biener <rguenther@suse.de>
1245 PR tree-optimization/91123
1246 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1247 (vn_walk_cb_data::last_vuse): New member.
1248 (vn_walk_cb_data::saved_operands): Likewsie.
1249 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1250 (vn_walk_cb_data::push_partial_def): Use finish.
1251 (vn_reference_lookup_2): Update last_vuse and use finish if
1252 we've saved operands.
1253 (vn_reference_lookup_3): Use finish and update calls to
1254 push_partial_defs everywhere. When translating through
1255 memcpy or aggregate copies save off operands and alias-set.
1256 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1257 operation for redundant store removal.
1259 2020-02-04 Richard Biener <rguenther@suse.de>
1261 PR tree-optimization/92819
1262 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1263 generating more stmts than before.
1265 2020-02-04 Martin Liska <mliska@suse.cz>
1267 * config/arm/arm.c (arm_gen_far_branch): Move the function
1268 outside of selftests.
1270 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1272 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1273 function to adjust PC-relative vector addresses.
1274 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1275 handle vectors with PC-relative addresses.
1277 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1279 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1281 (hard_reg_and_mode_to_addr_mask): Delete.
1282 (rs6000_adjust_vec_address): If the original vector address
1283 was REG+REG or REG+OFFSET and the element is not zero, do the add
1284 of the elements in the original address before adding the offset
1285 for the vector element. Use address_to_insn_form to validate the
1286 address using the register being loaded, rather than guessing
1287 whether the address is a DS-FORM or DQ-FORM address.
1289 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1291 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1292 to calculate the offset in memory from the start of a vector of a
1293 particular element. Add code to keep the element number in
1294 bounds if the element number is variable.
1295 (rs6000_adjust_vec_address): Move calculation of offset of the
1296 vector element to get_vector_offset.
1297 (rs6000_split_vec_extract_var): Do not do the initial AND of
1298 element here, move the code to get_vector_offset.
1300 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1302 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1305 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1307 * config/rs6000/constraints.md: Improve documentation.
1309 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1312 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1313 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1315 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1317 * config.gcc: Remove "carrizo" support.
1318 * config/gcn/gcn-opts.h (processor_type): Likewise.
1319 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1320 * config/gcn/gcn.opt (gpu_type): Likewise.
1321 * config/gcn/t-omp-device: Likewise.
1323 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1326 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1327 * config/arm/arm.c (arm_gen_far_branch): New function
1329 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1331 2020-02-03 Julian Brown <julian@codesourcery.com>
1332 Tobias Burnus <tobias@codesourcery.com>
1334 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1336 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1339 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1340 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1343 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1345 PR rtl-optimization/91333
1346 * ira-color.c (struct allocno_color_data): Add member
1348 (init_allocno_threads): Set the member up.
1349 (bucket_allocno_compare_func): Add compare hard reg
1352 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1354 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1356 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1357 * config.in: Regenerated.
1358 * configure: Regenerated.
1359 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1360 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1361 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1363 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1365 * configure: Regenerate.
1367 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1369 PR rtl-optimization/91333
1370 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1371 reg preferences comparison up.
1373 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1375 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1376 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1377 aarch64-sve-builtins-base.h.
1378 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1379 aarch64-sve-builtins-base.cc.
1380 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1381 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1383 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1384 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1385 (svcvtnt): New functions.
1386 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1387 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1388 (svcvtnt): New functions.
1389 (svcvt): Add a form that converts f32 to bf16.
1390 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1391 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1393 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1394 Treat B as bfloat16_t.
1395 (ternary_bfloat_lane_base): New class.
1396 (ternary_bfloat_def): Likewise.
1397 (ternary_bfloat): New shape.
1398 (ternary_bfloat_lane_def): New class.
1399 (ternary_bfloat_lane): New shape.
1400 (ternary_bfloat_lanex2_def): New class.
1401 (ternary_bfloat_lanex2): New shape.
1402 (ternary_bfloat_opt_n_def): New class.
1403 (ternary_bfloat_opt_n): New shape.
1404 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1405 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1406 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1407 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1408 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1409 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1410 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1411 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1412 the pattern off the narrow mode instead of the wider one.
1413 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1414 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1415 (sve_fp_op): Handle them.
1416 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1417 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1419 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1421 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1422 * config/aarch64/aarch64-modes.def (BF): Move definition before
1423 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1424 (SVE_MODES): Handle BF modes.
1425 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1427 (aarch64_full_sve_mode): Likewise.
1428 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1430 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1431 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1432 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1433 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1435 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1437 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1438 (TYPES_all_data): Add bf16.
1439 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1440 (register_tuple_type): Increase buffer size.
1441 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1442 (bf16): New type suffix.
1443 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1444 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1445 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1446 Change type from all_data to all_arith.
1447 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1450 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1451 Matthew Malcomson <matthew.malcomson@arm.com>
1452 Richard Sandiford <richard.sandiford@arm.com>
1454 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1455 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1456 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1457 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1458 __ARM_FEATURE_MATMUL_FP64.
1459 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1460 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1461 be disabled at the same time.
1462 (f32mm): New extension.
1463 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1464 (AARCH64_FL_F64MM): Bump to the next bit up.
1465 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1466 (TARGET_SVE_F64MM): New macros.
1467 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1468 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1469 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1470 (UNSPEC_ZIP2Q): New unspeccs.
1471 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1472 (optab, sur, perm_insn): Handle the new unspecs.
1473 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1474 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1475 TARGET_SVE_F64MM instead of separate tests.
1476 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1477 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1478 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1479 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1480 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1481 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1482 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1483 (TYPES_s_signed): New macro.
1484 (TYPES_s_integer): Use it.
1485 (TYPES_d_float): New macro.
1486 (TYPES_d_data): Use it.
1487 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1488 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1489 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1490 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1491 (svmmla): New shape.
1492 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1493 template parameters.
1494 (ternary_resize2_lane_base): Likewise.
1495 (ternary_resize2_base): New class.
1496 (ternary_qq_lane_base): Likewise.
1497 (ternary_intq_uintq_lane_def): Likewise.
1498 (ternary_intq_uintq_lane): New shape.
1499 (ternary_intq_uintq_opt_n_def): New class
1500 (ternary_intq_uintq_opt_n): New shape.
1501 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1502 (ternary_uintq_intq_def): New class.
1503 (ternary_uintq_intq): New shape.
1504 (ternary_uintq_intq_lane_def): New class.
1505 (ternary_uintq_intq_lane): New shape.
1506 (ternary_uintq_intq_opt_n_def): New class.
1507 (ternary_uintq_intq_opt_n): New shape.
1508 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1509 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1510 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1511 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1513 (svdotprod_lane_impl): ...this new class.
1514 (svmmla_impl, svusdot_impl): New classes.
1515 (svdot_lane): Update to use svdotprod_lane_impl.
1516 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1517 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1519 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1520 function, with no types defined.
1521 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1522 AARCH64_FL_I8MM functions.
1523 (svmmla): New AARCH64_FL_F32MM function.
1524 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1525 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1526 AARCH64_FL_F64MM function.
1527 (REQUIRED_EXTENSIONS):
1529 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1531 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1534 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1536 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1537 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1538 (*movti_internal): Do not check for
1539 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1540 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1541 just after check for TARGET_AVX.
1542 (*movdf_internal): Ditto.
1543 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1544 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1545 * config/i386/sse.md (mov<mode>_internal): Only check
1546 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1547 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1548 (<sse>_andnot<mode>3<mask_name>): Move check for
1549 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1550 (<code><mode>3<mask_name>): Ditto.
1551 (*andnot<mode>3): Ditto.
1552 (*andnottf3): Ditto.
1553 (*<code><mode>3): Ditto.
1554 (*<code>tf3): Ditto.
1555 (*andnot<VI:mode>3): Remove
1556 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1557 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1558 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1559 (sse4_1_blendv<ssemodesuffix>): Ditto.
1560 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1561 Explain that tune applies to 128bit instructions only.
1563 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1565 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1566 to definition of hsa_kernel_description. Parse assembly to find SGPR
1567 and VGPR count of kernel and store in hsa_kernel_description.
1569 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1571 PR rtl-optimization/91838
1572 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1573 to truncate if allowed or reject combination.
1575 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1577 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1578 (find_inv_vars_cb): Likewise.
1580 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1582 * calls.c (special_function_p): Split out the check for DECL_NAME
1583 being non-NULL and fndecl being extern at file scope into a
1584 new maybe_special_function_p and call it. Drop check for fndecl
1585 being non-NULL that was after a usage of DECL_NAME (fndecl).
1586 * tree.h (maybe_special_function_p): New inline function.
1588 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1590 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1591 (mask_gather_load<mode>): ... here, and zero-initialize the
1593 (maskload<mode>di): Zero-initialize the destination.
1596 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1599 * doc/analyzer.texi (Limitations): Note that constraints on
1600 floating-point values are currently ignored.
1602 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1605 * symtab.c (symtab_node::noninterposable_alias): If localalias
1606 already exists, but is not usable, append numbers after it until
1607 a unique name is found. Formatting fix.
1610 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1613 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1615 * config/gcn/gcn.c (print_operand): Handle LTGT.
1616 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1618 2020-01-30 Richard Biener <rguenther@suse.de>
1620 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1621 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1623 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1625 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1626 without a DECL in .data.rel.ro.local.
1628 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1631 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1635 * config/i386/sse.md
1636 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1637 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1638 any_extend code iterator instead of always zero_extend.
1639 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1640 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1641 Use any_extend code iterator instead of always zero_extend.
1642 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1643 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1644 Use any_extend code iterator instead of always zero_extend.
1645 (*sse2_pmovmskb_ext): New define_insn.
1646 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1649 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1650 (*popcountsi2_zext_falsedep): New define_insn.
1652 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1654 * config.in: Regenerated.
1655 * configure: Regenerated.
1657 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1660 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1661 LLVM's assembler changed the default in version 9.
1663 2020-01-24 Jeff Law <law@redhat.com>
1665 PR tree-optimization/89689
1666 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1668 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1672 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1674 PR rtl-optimization/87763
1675 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1676 simplification to handle subregs as well as bare regs.
1677 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1679 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1682 * ira.c (ira): Revert use of simplified LRA algorithm.
1684 2020-01-29 Martin Jambor <mjambor@suse.cz>
1686 PR tree-optimization/92706
1687 * tree-sra.c (struct access): Fields first_link, last_link,
1688 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1689 next_rhs_queued and grp_rhs_queued respectively, new fields
1690 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1691 (struct assign_link): Field next renamed to next_rhs, new field
1692 next_lhs. Updated comment.
1693 (work_queue_head): Renamed to rhs_work_queue_head.
1694 (lhs_work_queue_head): New variable.
1695 (add_link_to_lhs): New function.
1696 (relink_to_new_repr): Also relink LHS lists.
1697 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1698 (add_access_to_lhs_work_queue): New function.
1699 (pop_access_from_work_queue): Renamed to
1700 pop_access_from_rhs_work_queue.
1701 (pop_access_from_lhs_work_queue): New function.
1702 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1704 (child_would_conflict_in_lacc): Renamed to
1705 child_would_conflict_in_acc. Adjusted parameter names.
1706 (create_artificial_child_access): New parameter set_grp_read, use it.
1707 (subtree_mark_written_and_enqueue): Renamed to
1708 subtree_mark_written_and_rhs_enqueue.
1709 (propagate_subaccesses_across_link): Renamed to
1710 propagate_subaccesses_from_rhs.
1711 (propagate_subaccesses_from_lhs): New function.
1712 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1715 2020-01-29 Martin Jambor <mjambor@suse.cz>
1717 PR tree-optimization/92706
1718 * tree-sra.c (struct access): Adjust comment of
1719 grp_total_scalarization.
1720 (find_access_in_subtree): Look for single children spanning an entire
1722 (scalarizable_type_p): Allow register accesses, adjust callers.
1723 (completely_scalarize): Remove function.
1724 (scalarize_elem): Likewise.
1725 (create_total_scalarization_access): Likewise.
1726 (sort_and_splice_var_accesses): Do not track total scalarization
1728 (analyze_access_subtree): New parameter totally, adjust to new meaning
1729 of grp_total_scalarization.
1730 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1731 (can_totally_scalarize_forest_p): New function.
1732 (create_total_scalarization_access): Likewise.
1733 (create_total_access_and_reshape): Likewise.
1734 (total_should_skip_creating_access): Likewise.
1735 (totally_scalarize_subtree): Likewise.
1736 (analyze_all_variable_accesses): Perform total scalarization after
1737 subaccess propagation using the new functions above.
1738 (initialize_constant_pool_replacements): Output initializers by
1739 traversing the access tree.
1741 2020-01-29 Martin Jambor <mjambor@suse.cz>
1743 * tree-sra.c (verify_sra_access_forest): New function.
1744 (verify_all_sra_access_forests): Likewise.
1745 (create_artificial_child_access): Set parent.
1746 (analyze_all_variable_accesses): Call the verifier.
1748 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1750 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1751 if called on indirect edge.
1752 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1753 speculative call if needed.
1755 2020-01-29 Richard Biener <rguenther@suse.de>
1757 PR tree-optimization/93428
1758 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1759 permutation when the load node is created.
1760 (vect_analyze_slp_instance): Re-use it here.
1762 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1764 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1766 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1768 PR rtl-optimization/93272
1769 * ira-lives.c (process_out_of_region_eh_regs): New function.
1770 (process_bb_node_lives): Call it.
1772 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1774 * coverage.c (read_counts_file): Make error message lowercase.
1776 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1778 * profile-count.c (profile_quality_display_names): Fix ordering.
1780 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1783 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1784 hash only when edge is first within the sequence.
1785 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1786 (symbol_table::create_edge): Do not set target_prob.
1787 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1789 (cgraph_edge::make_speculative): Drop target_prob parameter.
1790 (cgraph_edge::speculative_call_info): Remove.
1791 (cgraph_edge::first_speculative_call_target): New member function.
1792 (update_call_stmt_hash_for_removing_direct_edge): New function.
1793 (cgraph_edge::resolve_speculation): Rewrite to new API.
1794 (cgraph_edge::speculative_call_for_target): New member function.
1795 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1796 multiple speculation targets.
1797 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1799 (verify_speculative_call): Verify that targets form an interval.
1800 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1801 (cgraph_edge::first_speculative_call_target): New member function.
1802 (cgraph_edge::next_speculative_call_target): New member function.
1803 (cgraph_edge::speculative_call_target_ref): New member function.
1804 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1805 (cgraph_edge): Remove target_prob.
1806 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1807 Fix handling of speculative calls.
1808 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1809 * ipa-fnsummary.c (analyze_function_body): Likewise.
1810 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1811 * ipa-profile.c (dump_histogram): Fix formating.
1812 (ipa_profile_generate_summary): Watch for overflows.
1813 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1814 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1815 (update_indirect_edges_after_inlining): Update to new API.
1816 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1818 * profile-count.h: (profile_probability::adjusted): New.
1819 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1820 updating of profile.
1821 * value-prof.c (gimple_ic_transform): Rename to ...
1822 (dump_ic_profile): ... this one; update dumping.
1823 (stream_in_histogram_value): Fix formating.
1824 (gimple_value_profile_transformations): Update.
1826 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1829 * config/i386/i386.md (*movoi_internal_avx): Remove
1830 TARGET_SSE_TYPELESS_STORES check.
1831 (*movti_internal): Prefer TARGET_AVX over
1832 TARGET_SSE_TYPELESS_STORES.
1833 (*movtf_internal): Likewise.
1834 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1835 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1836 from TARGET_SSE_TYPELESS_STORES.
1838 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1840 * diagnostic-core.h (warning_at): Rename overload to...
1841 (warning_meta): ...this.
1842 (emit_diagnostic_valist): Delete decl of overload taking
1843 diagnostic_metadata.
1844 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1845 (warning_at): Rename overload taking diagnostic_metadata to...
1846 (warning_meta): ...this.
1848 2020-01-28 Richard Biener <rguenther@suse.de>
1850 PR tree-optimization/93439
1851 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1852 * tree-cfg.c (move_sese_region_to_fn): ... here.
1853 (verify_types_in_gimple_reference): Verify used cliques are
1856 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1859 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1860 argument of a pointer to struct gcc_options and pass it to
1861 parse_mtune_ctrl_str.
1862 (ix86_function_specific_restore): Pass opts to
1863 set_ix86_tune_features.
1864 (ix86_option_override_internal): Likewise.
1865 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1866 gcc_options and use it for x_ix86_tune_ctrl_string.
1868 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1870 PR rtl-optimization/87763
1871 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1872 simplification to handle subregs as well as bare regs.
1873 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1875 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1877 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1878 for reduction chains that (now) include a call.
1880 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1882 PR tree-optimization/92822
1883 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1884 out the don't-care elements of a vector whose significant elements
1885 are duplicates, make the don't-care elements duplicates too.
1887 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1889 PR tree-optimization/93434
1890 * tree-predcom.c (split_data_refs_to_components): Record which
1891 components have had aliasing loads removed. Prevent store-store
1892 commoning for all such components.
1894 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1897 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1898 -1 or is_vshift is true, use new_vector with number of elts npatterns
1899 rather than new_unary_operation.
1901 PR tree-optimization/93454
1902 * gimple-fold.c (fold_array_ctor_reference): Perform
1903 elt_size.to_uhwi () just once, instead of calling it in every
1904 iteration. Punt if that value is above size of the temporary
1905 buffer. Decrease third native_encode_expr argument when
1906 bufoff + elt_sz is above size of buf.
1908 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1910 * config/mips/mips.c (mips_declare_object_name)
1911 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1913 2020-01-27 Martin Liska <mliska@suse.cz>
1915 PR gcov-profile/93403
1916 * tree-profile.c (gimple_init_gcov_profiler): Generate
1917 both __gcov_indirect_call_profiler_v4 and
1918 __gcov_indirect_call_profiler_v4_atomic.
1920 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1923 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1925 (@aarch64_split_simd_mov<mode>): Use it.
1926 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1927 Leave the vec_extract patterns to handle 2-element vectors.
1928 (aarch64_simd_mov_from_<mode>high): Likewise.
1929 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1930 (vec_extractv2dfv1df): Likewise.
1932 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1934 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1935 jump conditions for *compare_condjump<GPI:mode>.
1937 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1940 * digraph.cc (test_edge::test_edge): Specify template for base
1943 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1945 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1947 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1949 * config/arc/arc-protos.h (gen_mlo): Remove.
1950 (gen_mhi): Likewise.
1951 * config/arc/arc.c (AUX_MULHI): Define.
1952 (arc_must_save_reister): Special handling for r58/59.
1953 (arc_compute_frame_size): Consider mlo/mhi registers.
1954 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1956 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1957 mlo/mhi name selection.
1958 (arc_restore_callee_saves): Don't early restore blink when ISR.
1959 (arc_expand_prologue): Add mlo/mhi saving.
1960 (arc_expand_epilogue): Add mlo/mhi restoring.
1963 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1964 numbering when MUL64 option is used.
1965 (DWARF2_FRAME_REG_OUT): Define.
1966 * config/arc/arc.md (arc600_stall): New pattern.
1967 (VUNSPEC_ARC_ARC600_STALL): Define.
1968 (mulsi64): Use correct mlo/mhi registers.
1969 (mulsi_600): Clean it up.
1970 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1972 (mhi_operand): Likewise.
1974 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1975 Petro Karashchenko <petro.karashchenko@ring.com>
1977 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1978 attributes if needed.
1979 (prepare_move_operands): Generate special unspec instruction for
1981 (arc_isuncached_mem_p): Propagate uncached attribute to each
1983 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1984 (VUNSPEC_ARC_STDI): Likewise.
1985 (ALLI): New mode iterator.
1986 (mALLI): New mode attribute.
1987 (lddi): New instruction pattern.
1989 (stdidi_split): Split instruction for architectures which are not
1990 supporting ll64 option.
1991 (lddidi_split): Likewise.
1993 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1995 PR rtl-optimization/92989
1996 * lra-lives.c (process_bb_lives): Update the live-in set before
1997 processing additional clobbers.
1999 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2001 PR rtl-optimization/93170
2002 * cselib.c (cselib_invalidate_regno_val): New function, split out
2004 (cselib_invalidate_regno): ...here.
2005 (cselib_invalidated_by_call_p): New function.
2006 (cselib_process_insn): Iterate over all the hard-register entries in
2007 REG_VALUES and invalidate any that cross call-clobbered registers.
2009 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2011 * dojump.c (split_comparison): Use HONOR_NANS rather than
2012 HONOR_SNANS when splitting LTGT.
2014 2020-01-27 Martin Liska <mliska@suse.cz>
2017 * opts.c (print_filtered_help): Exclude language-specific
2018 options from --help=common unless enabled in all FEs.
2020 2020-01-27 Martin Liska <mliska@suse.cz>
2022 * opts.c (print_help): Exclude params from
2023 all except --help=param.
2025 2020-01-27 Martin Liska <mliska@suse.cz>
2028 * config/i386/i386-features.c (make_resolver_func):
2029 Align the code with ppc64 target implementation.
2030 Do not generate a unique name for resolver function.
2032 2020-01-27 Richard Biener <rguenther@suse.de>
2034 PR tree-optimization/93397
2035 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2036 converted reduction chain SLP graph adjustment.
2038 2020-01-26 Marek Polacek <polacek@redhat.com>
2041 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2044 2020-01-26 Jason Merrill <jason@redhat.com>
2047 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2050 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2052 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2053 (rx_setmem): Likewise.
2055 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2058 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2059 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2060 drop <di> from constraint of last operand.
2063 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2064 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2065 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2067 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2070 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2073 2020-01-24 Jeff Law <law@redhat.com>
2075 PR tree-optimization/92788
2076 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2079 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2082 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2083 *avx_vperm_broadcast_<mode>,
2084 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2085 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2086 Move before avx2_perm<mode>/avx512f_perm<mode>.
2089 * simplify-rtx.c (simplify_const_unary_operation,
2090 simplify_const_binary_operation): Punt for mode precision above
2091 MAX_BITSIZE_MODE_ANY_INT.
2093 2020-01-24 Andrew Pinski <apinski@marvell.com>
2095 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2098 2020-01-24 Jeff Law <law@redhat.com>
2101 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2102 for REGs. Call output_operand_lossage to get more reasonable
2105 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2107 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2108 gcn_fp_compare_operator.
2109 (vec_cmpu<mode>di): Use gcn_compare_operator.
2110 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2111 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2112 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2113 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2114 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2115 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2116 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2117 gcn_fp_compare_operator.
2118 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2119 gcn_fp_compare_operator.
2120 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2121 gcn_fp_compare_operator.
2122 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2123 gcn_fp_compare_operator.
2125 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2127 * doc/install.texi (Cross-Compiler-Specific Options): Document
2128 `--with-toolexeclibdir' option.
2130 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2132 * target.def (flags_regnum): Also mention effect on delay slot filling.
2133 * doc/tm.texi: Regenerate.
2135 2020-01-23 Jeff Law <law@redhat.com>
2137 PR translation/90162
2138 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2140 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2143 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2146 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2148 PR rtl-optimization/93402
2149 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2152 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2154 * config.in: Regenerated.
2155 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2156 for TARGET_LIBC_GNUSTACK.
2157 * configure: Regenerated.
2158 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2159 found to be 2.31 or greater.
2161 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2163 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2165 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2166 (mips_asm_file_end): New function. Delegate to
2167 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2168 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2170 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2173 * config/i386/i386-modes.def (POImode): New mode.
2174 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2175 * config/i386/i386.md (DPWI): New mode attribute.
2176 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2178 (QPWI): ... this. Use POI instead of OI for TImode.
2179 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2180 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2183 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2186 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2188 (speculation_tracker_rev): New pattern.
2189 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2190 Use speculation_tracker_rev to track the inverse condition.
2192 2020-01-23 Richard Biener <rguenther@suse.de>
2194 PR tree-optimization/93381
2195 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2196 alias-set of the def as argument and record the first one.
2197 (vn_walk_cb_data::first_set): New member.
2198 (vn_reference_lookup_3): Pass the alias-set of the current def
2199 to push_partial_def. Fix alias-set used in the aggregate copy
2201 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2202 * real.c (clear_significand_below): Fix out-of-bound access.
2204 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2207 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2208 New define_insn patterns.
2210 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2212 * doc/sourcebuild.texi (check-function-bodies): Add an
2213 optional target/xfail selector.
2215 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2217 PR rtl-optimization/93124
2218 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2219 bare USE and CLOBBER insns.
2221 2020-01-22 Andrew Pinski <apinski@marvell.com>
2223 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2225 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2228 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2229 diagnostic_manager into "ana" namespace.
2230 * selftest-run-tests.c (selftest::run_tests): Update for move of
2231 selftest::run_analyzer_selftests to
2232 ana::selftest::run_analyzer_selftests.
2234 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2236 * cfgexpand.c (union_stack_vars): Update the size.
2238 2020-01-22 Richard Biener <rguenther@suse.de>
2240 PR tree-optimization/93381
2241 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2242 throughout, handle all conversions the same.
2244 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2247 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2248 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2249 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2250 Call force_reg on high_in2 unconditionally.
2252 2020-01-22 Martin Liska <mliska@suse.cz>
2254 PR tree-optimization/92924
2255 * profile.c (compute_value_histograms): Divide
2258 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2261 * output.h (assemble_name_resolve): Declare.
2262 * varasm.c (assemble_name_resolve): New function.
2263 (assemble_name): Use it.
2264 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2266 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2268 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2269 update_web_docs_git instead of update_web_docs_svn.
2271 2020-01-21 Andrew Pinski <apinski@marvell.com>
2274 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2275 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2276 (*tlsgd_small_<mode>): Likewise.
2277 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2278 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2279 register. Convert that register back to dest using convert_mode.
2281 2020-01-21 Jim Wilson <jimw@sifive.com>
2283 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2286 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2287 Uros Bizjak <ubizjak@gmail.com>
2290 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2292 (legitimize_tls_address): Do GNU2 TLS address computation in
2293 ptr_mode and zero-extend result to Pmode.
2294 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2295 :P with :PTR and Pmode with ptr_mode.
2296 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2297 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2298 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2300 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2303 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2304 the last two operands are CONST_INT_P before using them as such.
2306 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2308 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2309 to get the integer element types.
2311 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2313 * config/aarch64/aarch64-sve-builtins.h
2314 (function_expander::convert_to_pmode): Declare.
2315 * config/aarch64/aarch64-sve-builtins.cc
2316 (function_expander::convert_to_pmode): New function.
2317 (function_expander::get_contiguous_base): Use it.
2318 (function_expander::prepare_gather_address_operands): Likewise.
2319 * config/aarch64/aarch64-sve-builtins-sve2.cc
2320 (svwhilerw_svwhilewr_impl::expand): Likewise.
2322 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2325 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2326 cfun->machine->label_is_assembled.
2327 (aarch64_print_patchable_function_entry): New.
2328 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2329 * config/aarch64/aarch64.h (struct machine_function): New field,
2332 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2335 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2338 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2341 * cgraph.c (cgraph_edge::resolve_speculation,
2342 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2343 call_stmt_site_hash.
2345 2020-01-21 Martin Liska <mliska@suse.cz>
2347 * config/rs6000/rs6000.c (common_mode_defined): Remove
2350 2020-01-21 Richard Biener <rguenther@suse.de>
2352 PR tree-optimization/92328
2353 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2354 type when value-numbering same-sized store by inserting a
2356 (eliminate_dom_walker::eliminate_stmt): When eliminating
2357 a redundant store handle bit-reinterpretation of the same value.
2359 2020-01-21 Andrew Pinski <apinski@marvel.com>
2362 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2364 (prepare_block_for_update): This. Use a worklist instead of
2367 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2369 * gcc/config/arm/arm.c (clear_operation_p):
2370 Initialise last_regno, skip first iteration
2371 based on the first_set value and use ints instead
2372 of the unnecessary HOST_WIDE_INTs.
2374 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2377 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2378 compare_mode other than SFmode or DFmode.
2380 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2383 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2384 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2385 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2387 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2389 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2391 2020-01-20 Andrew Pinski <apinski@marvell.com>
2394 * targhooks.c (default_print_patchable_function_entry): Use
2395 output_asm_insn to emit the nop instruction.
2397 2020-01-20 Fangrui Song <maskray@google.com>
2400 * targhooks.c (default_print_patchable_function_entry): Align to
2403 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2406 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2407 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2408 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2409 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2410 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2411 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2412 Remove the {q} suffix from lea.
2413 (*tls_dynamic_gnu2_call_64): Renamed to ...
2414 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2415 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2416 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2417 Pass Pmode to gen_tls_dynamic_gnu2_64.
2419 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2421 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2423 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2425 * config/aarch64/aarch64-sve-builtins-base.cc
2426 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2428 2020-01-20 Richard Biener <rguenther@suse.de>
2431 * dwarf2out.c (prune_unused_types): Unconditionally mark
2432 called function DIEs.
2434 2020-01-20 Martin Liska <mliska@suse.cz>
2436 PR tree-optimization/93199
2437 * tree-eh.c (struct leh_state): Add
2438 new field outer_non_cleanup.
2439 (cleanup_is_dead_in): Pass leh_state instead
2440 of eh_region. Add a checking that state->outer_non_cleanup
2441 points to outer non-clean up region.
2442 (lower_try_finally): Record outer_non_cleanup
2444 (lower_catch): Likewise.
2445 (lower_eh_filter): Likewise.
2446 (lower_eh_must_not_throw): Likewise.
2447 (lower_cleanup): Likewise.
2449 2020-01-20 Richard Biener <rguenther@suse.de>
2451 PR tree-optimization/93094
2452 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2453 (vect_transform_loop): Likewise.
2454 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2455 loop_vectorized_call to vect_transform_loop.
2456 * tree-vect-loop.c (vect_transform_loop): Pass down
2457 loop_vectorized_call to vect_loop_versioning.
2458 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2459 the earlier discovered loop_vectorized_call.
2461 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2463 * doc/contribute.texi: Update for SVN -> Git transition.
2464 * doc/install.texi: Likewise.
2466 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2469 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2470 speculative targets.
2471 (verify_speculative_call): New function
2472 (cgraph_node::verify_node): Use it.
2473 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2476 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2479 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2480 (cgraph_edge::make_direct): Remove all indirect targets.
2481 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2482 (cgraph_node::verify_node): Verify that only one call_stmt or
2483 lto_stmt_uid is set.
2484 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2486 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2487 (lto_output_ref): Simplify streaming of stmt.
2488 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2490 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2492 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2493 Mark parameter unused.
2495 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2497 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2499 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2501 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2503 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2505 * Makefile.in: Add coroutine-passes.o.
2506 * builtin-types.def (BT_CONST_SIZE): New.
2507 (BT_FN_BOOL_PTR): New.
2508 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2509 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2510 * coroutine-builtins.def: New file.
2511 * coroutine-passes.cc: New file.
2512 * function.h (struct GTY function): Add a bit to indicate that the
2513 function is a coroutine component.
2514 * internal-fn.c (expand_CO_FRAME): New.
2515 (expand_CO_YIELD): New.
2516 (expand_CO_SUSPN): New.
2517 (expand_CO_ACTOR): New.
2518 * internal-fn.def (CO_ACTOR): New.
2522 * passes.def: Add pass_coroutine_lower_builtins,
2523 pass_coroutine_early_expand_ifns.
2524 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2525 (make_pass_coroutine_early_expand_ifns): New.
2526 * doc/invoke.texi: Document the fcoroutines command line
2529 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2531 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2534 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2535 after checking the argument is a REG. Don't use REGNO (reg)
2536 again to set last_regno, reuse regno variable instead.
2538 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2540 * doc/analyzer.texi (Limitations): Add note about NaN.
2542 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2543 Sudakshina Das <sudi.das@arm.com>
2545 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2546 and valid immediate.
2547 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2548 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2549 * config/arm/constraints.md (Pg): New.
2550 * config/arm/predicates.md (long_shift_imm): New.
2551 (arm_reg_or_long_shift_imm): Likewise.
2552 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2553 (thumb2_lsll): Likewise.
2556 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2557 Sudakshina Das <sudi.das@arm.com>
2559 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2560 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2561 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2562 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2563 * config/arm/thumb2.md (thumb2_asrl): New.
2564 (thumb2_lsll): Likewise.
2566 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2568 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2571 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2573 * gdbinit.in (help-gcc-hooks): New command.
2574 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2575 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2578 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2580 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2581 correct target macro.
2583 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2585 * config/aarch64/aarch64-protos.h
2586 (aarch64_sve_ld1ro_operand_p): New.
2587 * config/aarch64/aarch64-sve-builtins-base.cc
2588 (class load_replicate): New.
2589 (class svld1ro_impl): New.
2590 (class svld1rq_impl): Change to inherit from load_replicate.
2591 (svld1ro): New sve intrinsic function base.
2592 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2593 New DEF_SVE_FUNCTION.
2594 * config/aarch64/aarch64-sve-builtins-base.h
2595 (svld1ro): New decl.
2596 * config/aarch64/aarch64-sve-builtins.cc
2597 (function_expander::add_mem_operand): Modify assert to allow
2599 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2601 * config/aarch64/aarch64.c
2602 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2603 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2604 (aarch64_sve_ld1ro_operand_p): New.
2605 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2606 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2607 * config/aarch64/predicates.md
2608 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2610 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2612 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2613 Introduce this ACLE specified predefined macro.
2614 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2615 (fp): Disabling this disables f64mm.
2616 (simd): Disabling this disables f64mm.
2617 (fp16): Disabling this disables f64mm.
2618 (sve): Disabling this disables f64mm.
2619 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2620 (AARCH64_ISA_F64MM): New.
2621 (TARGET_F64MM): New.
2622 * doc/invoke.texi (f64mm): Document new option.
2624 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2626 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2627 (neoversen1_tunings): Likewise.
2629 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2632 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2633 Add assert to ensure prolog has been emitted.
2634 (aarch64_split_atomic_op): Likewise.
2635 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2636 Use epilogue_completed rather than reload_completed.
2637 (aarch64_atomic_exchange<mode>): Likewise.
2638 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2639 (atomic_nand<mode>): Likewise.
2640 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2641 (atomic_fetch_nand<mode>): Likewise.
2642 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2643 (atomic_nand_fetch<mode>): Likewise.
2645 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2648 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2650 (REVERSE_CONDITION): Delete.
2651 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2652 (CCFP_CCFPE): Likewise.
2653 (e): New mode attribute.
2654 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2655 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2656 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2657 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2658 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2659 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2660 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2661 name of generator from gen_ccmpdi to gen_ccmpccdi.
2662 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2663 the previous comparison but aren't able to, use the new ccmp_rev
2666 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2668 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2669 than testing directly for INTEGER_CST.
2670 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2672 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2674 PR tree-optimization/93292
2675 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2676 get_vectype_for_scalar_type returns NULL.
2678 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2680 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2681 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2683 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2685 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2687 * params.opt: (max-predicted-iterations): Set bounds.
2688 * predict.c (real_almost_one, real_br_prob_base,
2689 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2690 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2691 probabilities; do not truncate to reg_br_prob_bases.
2692 (estimate_loops_at_level): Pass max_cyclic_prob.
2693 (estimate_loops): Compute max_cyclic_prob.
2694 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2696 * profile-count.c (profile_probability::to_sreal): New.
2697 * profile-count.h (class sreal): Move up in file.
2698 (profile_probability::to_sreal): Declare.
2700 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2703 (arm_invalid_conversion): New function for target hook.
2704 (arm_invalid_unary_op): New function for target hook.
2705 (arm_invalid_binary_op): New function for target hook.
2707 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2709 * config.gcc: Add arm_bf16.h.
2710 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2711 (arm_simd_builtin_std_type): Add BFmode.
2712 (arm_init_simd_builtin_types): Define element types for vector types.
2713 (arm_init_bf16_types): New function.
2714 (arm_init_builtins): Add arm_init_bf16_types function call.
2715 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2716 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2717 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2718 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2719 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2720 (arm_mangle_type): Add __bf16.
2721 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2722 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2723 arm_bf16_ptr_type_node.
2724 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2725 define_split between ARM registers.
2726 * config/arm/arm_bf16.h: New file.
2727 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2728 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2730 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2731 * config/arm/vfp.md: Add BFmode to movhf patterns.
2733 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2734 Andre Vieira <andre.simoesdiasvieira@arm.com>
2736 * config/arm/arm-cpus.in (mve, mve_float): New features.
2737 (dsp, mve, mve.fp): New options.
2738 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2739 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2740 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2742 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2743 Thomas Preud'homme <thomas.preudhomme@arm.com>
2745 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2747 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2748 error for using -mcmse when targeting Armv8.1-M Mainline.
2750 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2751 Thomas Preud'homme <thomas.preudhomme@arm.com>
2753 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2754 address in r4 when targeting Armv8.1-M Mainline.
2755 (nonsecure_call_value_internal): Likewise.
2756 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2757 a register match_operand again. Emit BLXNS when targeting
2759 (nonsecure_call_value_reg_thumb2): Likewise.
2761 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2762 Thomas Preud'homme <thomas.preudhomme@arm.com>
2764 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2765 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2766 variable as true when floating-point ABI is not hard. Replace
2767 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2768 Generate VLSTM and VLLDM instruction respectively before and
2769 after a function call to cmse_nonsecure_call function.
2770 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2771 (VUNSPEC_VLLDM): Likewise.
2772 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2773 (lazy_load_multiple_insn): Likewise.
2775 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2776 Thomas Preud'homme <thomas.preudhomme@arm.com>
2778 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2779 (arm_emit_vfp_multi_reg_pop): Likewise.
2780 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2781 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2782 restore callee-saved VFP registers.
2784 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2785 Thomas Preud'homme <thomas.preudhomme@arm.com>
2787 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2788 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2789 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2790 callee-saved GPRs as well as clear ip register before doing a nonsecure
2791 call then restore callee-saved GPRs after it when targeting
2793 (arm_reorg): Adapt to function rename.
2795 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2796 Thomas Preud'homme <thomas.preudhomme@arm.com>
2798 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2799 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2800 clear_vfp_multiple pattern based on a new vfp parameter.
2801 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2802 targeting Armv8.1-M Mainline.
2803 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2804 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2805 whether VFP registers are available before looking call_used_regs for a
2807 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2808 of prototype of clear_operation_p.
2809 (clear_vfp_multiple_operation): New predicate.
2810 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2811 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2813 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2814 Thomas Preud'homme <thomas.preudhomme@arm.com>
2816 * config/arm/arm-protos.h (clear_operation_p): Declare.
2817 * config/arm/arm.c (clear_operation_p): New function.
2818 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2819 targeting Armv8.1-M Mainline or successor.
2820 (output_return_instruction): Only output APSR register clearing if
2821 Armv8.1-M Mainline instructions not available.
2822 (thumb_exit): Likewise.
2823 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2824 * config/arm/thumb2.md (clear_apsr): New define_insn.
2825 (clear_multiple): Likewise.
2826 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2828 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2829 Thomas Preud'homme <thomas.preudhomme@arm.com>
2831 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2832 (use_return_insn): Also return false for Armv8.1-M Mainline.
2833 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2834 Mainline instructions are available.
2835 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2836 when targeting Armv8.1-M Mainline Security Extensions.
2837 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2838 Mainline entry function.
2839 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2840 targeting Armv8.1-M Mainline or successor.
2841 (arm_expand_epilogue): Fix indentation of caller-saved register
2842 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2844 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2845 (FP_SYSREGS): Likewise.
2846 (enum vfp_sysregs_encoding): Define enum.
2847 (fp_sysreg_names): Declare.
2848 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2849 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2850 (pop_fpsysreg_insn): Likewise.
2852 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2853 Thomas Preud'homme <thomas.preudhomme@arm.com>
2855 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2856 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2857 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2858 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2859 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2860 (ARMv8_1m_main): New feature group.
2861 (armv8.1-m.main): New architecture.
2862 * config/arm/arm-tables.opt: Regenerate.
2863 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2864 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2865 (arm_options_perform_arch_sanity_checks): Error out when targeting
2866 Armv8.1-M Mainline Security Extensions.
2867 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2869 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2871 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2872 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2873 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2874 aarch64_bfdot_laneq): New.
2875 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2876 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2877 vbfdotq_laneq_f32): New.
2878 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2879 VBFMLA_W, VBF): New.
2880 (isquadop): Add V4BF, V8BF.
2882 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2884 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2885 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2886 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2887 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2888 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2889 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2890 usdot_laneq, sudot_lane,sudot_laneq): New.
2891 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2892 (aarch64_<sur>dot_lane): New.
2893 * config/aarch64/arm_neon.h (vusdot_s32): New.
2895 (vusdot_lane_s32): New.
2896 (vsudot_lane_s32): New.
2897 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2898 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2900 2020-01-16 Martin Liska <mliska@suse.cz>
2902 * value-prof.c (dump_histogram_value): Fix
2903 obvious spacing issue.
2905 2020-01-16 Andrew Pinski <apinski@marvell.com>
2907 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2908 !storage_order_barrier_p.
2910 2020-01-16 Andrew Pinski <apinski@marvell.com>
2912 * sched-int.h (_dep): Add unused bit-field field for the padding.
2913 * sched-deps.c (init_dep_1): Init unused field.
2915 2020-01-16 Andrew Pinski <apinski@marvell.com>
2917 * optabs.h (create_expand_operand): Initialize target field also.
2919 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2921 PR tree-optimization/92429
2922 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2923 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2925 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2928 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2930 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2931 aarch64_sve_int_mode to each mode.
2933 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2935 * doc/analyzer.texi (Overview): Add note about
2936 -fdump-ipa-analyzer.
2938 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2940 PR tree-optimization/93231
2941 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2942 input_type is unsigned. Use tree_to_shwi for shift constant.
2943 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2944 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2945 inputs more efficiently.
2947 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2949 * config/i386/i386.md (*movsf_internal): Do not require
2950 SSE2 ISA for alternatives 14 and 15.
2952 2020-01-15 Richard Biener <rguenther@suse.de>
2955 * tree-eh.c (sink_clobbers): If we already visited the destination
2956 block do not defer insertion.
2957 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2958 the purpose of defered insertion.
2960 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2962 * BASE-VER: Bump to 10.0.1.
2964 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2966 PR tree-optimization/93247
2967 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2968 type of the stmt that we're going to vectorize.
2970 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2972 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2973 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2976 2020-01-15 Martin Liska <mliska@suse.cz>
2978 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2979 2 calls of streamer_read_hwi in a function call.
2981 2020-01-15 Richard Biener <rguenther@suse.de>
2983 * alias.c (record_alias_subset): Avoid redundant work when
2984 subset is already recorded.
2986 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2988 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2989 the analyzer options provide CWE identifiers.
2991 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2993 * tree-diagnostic-path.cc (path_summary::event_range::print):
2994 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2995 using get_pure_location.
2997 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2999 PR tree-optimization/93262
3000 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3001 perform head trimming only if the last argument is constant,
3002 either all ones, or larger or equal to head trim, in the latter
3003 case decrease the last argument by head_trim.
3005 PR tree-optimization/93249
3006 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3007 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3008 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3009 perform head trim unless we can prove there are no '\0' chars
3010 from the source among the first head_trim chars.
3012 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3014 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3016 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3019 * config/i386/sse.md
3020 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3021 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3022 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3023 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3024 just a single alternative instead of two, make operands 1 and 2
3027 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3030 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3033 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3035 * Makefile.in (lang_opt_files): Add analyzer.opt.
3036 (ANALYZER_OBJS): New.
3037 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3038 tristate.o and ANALYZER_OBJS.
3039 (TEXI_GCCINT_FILES): Add analyzer.texi.
3040 * common.opt (-fanalyzer): New driver option.
3041 * config.in: Regenerate.
3042 * configure: Regenerate.
3043 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3044 (gccdepdir): Also create depdir for "analyzer" subdir.
3045 * digraph.cc: New file.
3046 * digraph.h: New file.
3047 * doc/analyzer.texi: New file.
3048 * doc/gccint.texi ("Static Analyzer") New menu item.
3049 (analyzer.texi): Include it.
3050 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3051 ("Warning Options"): Add static analysis warnings to the list.
3052 (-Wno-analyzer-double-fclose): New option.
3053 (-Wno-analyzer-double-free): New option.
3054 (-Wno-analyzer-exposure-through-output-file): New option.
3055 (-Wno-analyzer-file-leak): New option.
3056 (-Wno-analyzer-free-of-non-heap): New option.
3057 (-Wno-analyzer-malloc-leak): New option.
3058 (-Wno-analyzer-possible-null-argument): New option.
3059 (-Wno-analyzer-possible-null-dereference): New option.
3060 (-Wno-analyzer-null-argument): New option.
3061 (-Wno-analyzer-null-dereference): New option.
3062 (-Wno-analyzer-stale-setjmp-buffer): New option.
3063 (-Wno-analyzer-tainted-array-index): New option.
3064 (-Wno-analyzer-use-after-free): New option.
3065 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3066 (-Wno-analyzer-use-of-uninitialized-value): New option.
3067 (-Wanalyzer-too-complex): New option.
3068 (-fanalyzer-call-summaries): New warning.
3069 (-fanalyzer-checker=): New warning.
3070 (-fanalyzer-fine-grained): New warning.
3071 (-fno-analyzer-state-merge): New warning.
3072 (-fno-analyzer-state-purge): New warning.
3073 (-fanalyzer-transitivity): New warning.
3074 (-fanalyzer-verbose-edges): New warning.
3075 (-fanalyzer-verbose-state-changes): New warning.
3076 (-fanalyzer-verbosity=): New warning.
3077 (-fdump-analyzer): New warning.
3078 (-fdump-analyzer-callgraph): New warning.
3079 (-fdump-analyzer-exploded-graph): New warning.
3080 (-fdump-analyzer-exploded-nodes): New warning.
3081 (-fdump-analyzer-exploded-nodes-2): New warning.
3082 (-fdump-analyzer-exploded-nodes-3): New warning.
3083 (-fdump-analyzer-supergraph): New warning.
3084 * doc/sourcebuild.texi (dg-require-dot): New.
3085 (dg-check-dot): New.
3086 * gdbinit.in (break-on-saved-diagnostic): New command.
3087 * graphviz.cc: New file.
3088 * graphviz.h: New file.
3089 * ordered-hash-map-tests.cc: New file.
3090 * ordered-hash-map.h: New file.
3091 * passes.def (pass_analyzer): Add before
3092 pass_ipa_whole_program_visibility.
3093 * selftest-run-tests.c (selftest::run_tests): Call
3094 selftest::ordered_hash_map_tests_cc_tests.
3095 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3097 * shortest-paths.h: New file.
3098 * timevar.def (TV_ANALYZER): New timevar.
3099 (TV_ANALYZER_SUPERGRAPH): Likewise.
3100 (TV_ANALYZER_STATE_PURGE): Likewise.
3101 (TV_ANALYZER_PLAN): Likewise.
3102 (TV_ANALYZER_SCC): Likewise.
3103 (TV_ANALYZER_WORKLIST): Likewise.
3104 (TV_ANALYZER_DUMP): Likewise.
3105 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3106 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3107 * tree-pass.h (make_pass_analyzer): New decl.
3108 * tristate.cc: New file.
3109 * tristate.h: New file.
3111 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3114 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3115 alternatives 9 and 10.
3117 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3119 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3120 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3121 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3122 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3123 (selftest::hash_map_tests_c_tests): Call it.
3124 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3125 New static constant, using the value of = H::empty_zero_p.
3126 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3127 from default_hash_traits <Value>.
3128 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3130 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3131 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3132 calls to mark_empty with !Descriptor::empty_zero_p.
3133 (hash_table::empty_slow): Conditionalize the memset call with a
3134 check that Descriptor::empty_zero_p; otherwise, loop through the
3135 entries calling mark_empty on them.
3136 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3137 (pointer_hash::empty_zero_p): Likewise.
3138 (pair_hash::empty_zero_p): Likewise.
3139 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3141 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3142 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3143 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3144 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3145 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3146 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3147 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3148 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3150 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3153 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3155 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3156 fix typo on return value.
3158 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3161 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3163 (cgraph_edge::make_speculative): Add param for setting speculative_id
3165 (cgraph_edge::speculative_call_info): Update comments and find reference
3166 by speculative_id for multiple indirect targets.
3167 (cgraph_edge::resolve_speculation): Decrease the speculations
3168 for indirect edge, drop it's speculative if not direct target
3169 left. Update comments.
3170 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3171 (cgraph_node::dump): Print num_speculative_call_targets.
3172 (cgraph_node::verify_node): Don't report error if speculative
3173 edge not include statement.
3174 (cgraph_edge::num_speculative_call_targets_p): New function.
3175 * cgraph.h (int common_target_id): Remove.
3176 (int common_target_probability): Remove.
3177 (num_speculative_call_targets): New variable.
3178 (make_speculative): Add param for setting speculative_id.
3179 (cgraph_edge::num_speculative_call_targets_p): New declare.
3180 (target_prob): New variable.
3181 (speculative_id): New variable.
3182 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3183 call summaries for multiple speculative call targets.
3184 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3185 * ipa-profile.c (struct speculative_call_target): New struct.
3186 (class speculative_call_summary): New class.
3187 (class speculative_call_summaries): New class.
3188 (call_sums): New variable.
3189 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3190 (ipa_profile_write_edge_summary): New function.
3191 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3192 (ipa_profile_dump_all_summaries): New function.
3193 (ipa_profile_read_edge_summary): New function.
3194 (ipa_profile_read_summary_section): New function.
3195 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3196 (ipa_profile): Generate num_speculative_call_targets from
3198 * ipa-ref.h (speculative_id): New variable.
3199 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3200 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3201 common_target_probability. Stream out speculative_id and
3202 num_speculative_call_targets.
3203 (input_edge): Likewise.
3204 * predict.c (dump_prediction): Remove edges count assert to be
3206 * symtab.c (symtab_node::create_reference): Init speculative_id.
3207 (symtab_node::clone_references): Clone speculative_id.
3208 (symtab_node::clone_referring): Clone speculative_id.
3209 (symtab_node::clone_reference): Clone speculative_id.
3210 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3211 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3212 if indirect call contains multiple speculative targets.
3213 * value-prof.h (check_ic_target): Remove.
3214 * value-prof.c (gimple_value_profile_transformations):
3215 Use void function gimple_ic_transform.
3216 * value-prof.c (gimple_ic_transform): Handle topn case.
3217 Fix comment typos. Change it to a void function.
3219 2020-01-13 Andrew Pinski <apinski@marvell.com>
3221 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3222 (octeontx2t98): New define.
3223 (octeontx2t96): New define.
3224 (octeontx2t93): New define.
3225 (octeontx2f95): New define.
3226 (octeontx2f95n): New define.
3227 (octeontx2f95mm): New define.
3228 * config/aarch64/aarch64-tune.md: Regenerate.
3229 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3231 2020-01-13 Jason Merrill <jason@redhat.com>
3233 PR c++/33799 - destroy return value if local cleanup throws.
3234 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3236 2020-01-13 Martin Liska <mliska@suse.cz>
3238 * ipa-cp.c (get_max_overall_size): Use newly
3239 renamed param param_ipa_cp_unit_growth.
3240 * params.opt: Remove legacy param name.
3242 2020-01-13 Martin Sebor <msebor@redhat.com>
3244 PR tree-optimization/93213
3245 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3246 stores to be eliminated.
3248 2020-01-13 Martin Liska <mliska@suse.cz>
3250 * opts.c (print_help): Do not print CL_PARAM
3251 and CL_WARNING for CL_OPTIMIZATION.
3253 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3256 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3257 depending on optimization settings.
3259 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3261 PR tree-optimization/90838
3262 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3263 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3264 argument rather than to initialize temporary for targets that
3265 don't use the mode argument at all. Initialize ctzval to avoid
3268 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3270 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3271 * tree-core.h: Document it.
3272 * gimplify.c (gimplify_omp_workshare): Set it.
3273 * omp-low.c (lower_omp_target): Use it.
3274 * tree-pretty-print.c (dump_omp_clause): Print it.
3276 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3277 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3279 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3281 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3282 * common.opt (fdiagnostics-path-format=): New option.
3283 (diagnostic_path_format): New enum.
3284 (fdiagnostics-show-path-depths): New option.
3285 * coretypes.h (diagnostic_event_id_t): New forward decl.
3286 * diagnostic-color.c (color_dict): Add "path".
3287 * diagnostic-event-id.h: New file.
3288 * diagnostic-format-json.cc (json_from_expanded_location): Make
3290 (json_end_diagnostic): Call context->make_json_for_path if it
3291 exists and the diagnostic has a path.
3292 (diagnostic_output_format_init): Clear context->print_path.
3293 * diagnostic-path.h: New file.
3294 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3295 when printing a run of events in a diagnostic_path so that they
3296 all get the same color.
3297 (layout::m_diagnostic_path_p): New field.
3298 (layout::layout): Initialize it.
3299 (layout::print_any_labels): Don't colorize the label text for an
3300 event in a diagnostic_path.
3301 (gcc_rich_location::add_location_if_nearby): Add
3302 "restrict_to_current_line_spans" and "label" params. Pass the
3303 former to layout.maybe_add_location_range; pass the latter
3304 when calling add_range.
3305 * diagnostic.c: Include "diagnostic-path.h".
3306 (diagnostic_initialize): Initialize context->path_format and
3307 context->show_path_depths.
3308 (diagnostic_show_any_path): New function.
3309 (diagnostic_path::interprocedural_p): New function.
3310 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3311 (simple_diagnostic_path::num_events): New function.
3312 (simple_diagnostic_path::get_event): New function.
3313 (simple_diagnostic_path::add_event): New function.
3314 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3315 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3316 (debug): New overload taking a diagnostic_path *.
3317 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3318 * diagnostic.h (enum diagnostic_path_format): New enum.
3319 (json::value): New forward decl.
3320 (diagnostic_context::path_format): New field.
3321 (diagnostic_context::show_path_depths): New field.
3322 (diagnostic_context::print_path): New callback field.
3323 (diagnostic_context::make_json_for_path): New callback field.
3324 (diagnostic_show_any_path): New decl.
3325 (json_from_expanded_location): New decl.
3326 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3327 (-fdiagnostics-show-path-depths): New option.
3328 (-fdiagnostics-color): Add "path" to description of default
3329 GCC_COLORS; describe it.
3330 (-fdiagnostics-format=json): Document how diagnostic paths are
3331 represented in the JSON output format.
3332 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3333 Add optional params "restrict_to_current_line_spans" and "label".
3334 * opts.c (common_handle_option): Handle
3335 OPT_fdiagnostics_path_format_ and
3336 OPT_fdiagnostics_show_path_depths.
3337 * pretty-print.c: Include "diagnostic-event-id.h".
3338 (pp_format): Implement "%@" format code for printing
3339 diagnostic_event_id_t *.
3340 (selftest::test_pp_format): Add tests for "%@".
3341 * selftest-run-tests.c (selftest::run_tests): Call
3342 selftest::tree_diagnostic_path_cc_tests.
3343 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3344 * toplev.c (general_init): Initialize global_dc->path_format and
3345 global_dc->show_path_depths.
3346 * tree-diagnostic-path.cc: New file.
3347 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3348 non-static. Drop "diagnostic" param in favor of storing the
3349 original value of "where" and re-using it.
3350 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3351 maybe_unwind_expanded_macro_loc.
3352 (tree_diagnostics_defaults): Initialize context->print_path and
3353 context->make_json_for_path.
3354 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3356 (default_tree_make_json_for_path): New decl.
3357 (maybe_unwind_expanded_macro_loc): New decl.
3359 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3361 PR tree-optimization/93210
3362 * fold-const.h (native_encode_initializer,
3363 can_native_interpret_type_p): Declare.
3364 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3366 (native_encode_initializer): New function, moved from dwarf2out.c.
3367 Adjust to native_encode_expr compatible arguments, including dry-run
3368 and partial extraction modes. Don't handle STRING_CST.
3369 (can_native_interpret_type_p): No longer static.
3370 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3371 offset / BITS_PER_UNIT fits into int and don't call it if
3372 can_native_interpret_type_p fails. If suboff is NULL and for
3373 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3374 native_encode_initializer.
3375 (fold_const_aggregate_ref_1): Formatting fix.
3376 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3377 (tree_add_const_value_attribute): Adjust caller.
3379 PR tree-optimization/90838
3380 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3381 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3382 CTZ_DEFINED_VALUE_AT_ZERO.
3384 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3387 * lra-constraints.c (match_reload): Permit input operands have the
3388 same mode as output while other input operands have a different
3391 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3393 PR tree-optimization/90838
3394 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3395 (check_ctz_string): Likewise.
3396 (optimize_count_trailing_zeroes): Likewise.
3397 (simplify_count_trailing_zeroes): Likewise.
3398 (pass_forwprop::execute): Try ctz simplification.
3399 * match.pd: Add matching for ctz idioms.
3401 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3403 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3405 (aarch64_invalid_unary_op): New function for target hook.
3406 (aarch64_invalid_binary_op): New function for target hook.
3408 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3410 * config.gcc: Add arm_bf16.h.
3411 * config/aarch64/aarch64-builtins.c
3412 (aarch64_simd_builtin_std_type): Add BFmode.
3413 (aarch64_init_simd_builtin_types): Define element types for vector
3415 (aarch64_init_bf16_types): New function.
3416 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3417 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3419 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3420 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3422 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3423 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3424 * config/aarch64/aarch64.c
3425 (aarch64_classify_vector_mode): Add support for BF types.
3426 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3427 (aarch64_vq_mode): Add support for BF types.
3428 (aarch64_simd_container_mode): Add support for BF types.
3429 (aarch64_mangle_type): Add support for BF scalar type.
3430 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3431 * config/aarch64/arm_bf16.h: New file.
3432 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3433 * config/aarch64/iterators.md: Add BF types to mode attributes.
3434 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3436 2020-01-10 Jason Merrill <jason@redhat.com>
3438 PR c++/93173 - incorrect tree sharing.
3439 * gimplify.c (copy_if_shared): No longer static.
3440 * gimplify.h: Declare it.
3442 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3444 * doc/invoke.texi (-msve-vector-bits=): Document that
3445 -msve-vector-bits=128 now generates VL-specific code for
3446 little-endian targets.
3447 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3448 build_vector_type_for_mode to construct the data vector types.
3449 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3450 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3451 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3452 for 128-bit vectors.
3454 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3456 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3459 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3461 * config/aarch64/aarch64-builtins.c
3462 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3463 rather than checking the number of elements and the element mode.
3465 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3467 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3468 get_related_vectype_for_scalar_type rather than build_vector_type
3469 to create the index type for a conditional reduction.
3471 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3473 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3474 for any type of gather or scatter, including strided accesses.
3476 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3478 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3481 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3483 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3485 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3486 parameter and its use to reset DR_OFFSET's.
3487 (vect_transform_loop): Remove orig_drs_init argument.
3488 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3489 member of dr_vec_info rather than the offset of the associated
3490 data_reference's innermost_loop_behavior.
3491 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3492 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3493 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3494 get_dr_vinfo_offset.
3495 (vectorizable_store): Likewise.
3496 (vectorizable_load): Likewise.
3498 2020-01-10 Richard Biener <rguenther@suse.de>
3500 * gimple-ssa-store-merging
3501 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3503 2020-01-10 Martin Liska <mliska@suse.cz>
3506 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3507 encapsulation that was there before r280040.
3509 2020-01-10 Richard Biener <rguenther@suse.de>
3512 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3513 sequences to avoid walking them again for secondary opportunities.
3514 (pass_lower_eh_dispatch::execute): Instead actually insert
3517 2020-01-10 Richard Biener <rguenther@suse.de>
3520 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3521 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3522 avoid quadraticness.
3524 2020-01-10 Martin Jambor <mjambor@suse.cz>
3526 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3527 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3528 to get param_ipa_sra_max_replacements.
3529 (param_splitting_across_edge): Pass the caller to
3530 pull_accesses_from_callee.
3532 2020-01-10 Martin Jambor <mjambor@suse.cz>
3534 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3535 * ipa-cp.c (max_new_size): Removed.
3536 (orig_overall_size): New variable.
3537 (get_max_overall_size): New function.
3538 (estimate_local_effects): Use it. Adjust dump.
3539 (decide_about_value): Likewise.
3540 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3541 orig_overall_size. Adjust dump.
3542 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3544 2020-01-10 Martin Jambor <mjambor@suse.cz>
3546 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3547 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3548 instead of param_ipa_max_agg_items.
3549 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3550 optimization info for the callee.
3552 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3554 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3555 markers if debug_inline_points is false.
3557 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3559 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3561 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3562 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3563 aarch64-sve-builtins-sve2.h.
3564 (aarch64-sve-builtins-sve2.o): New rule.
3565 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3566 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3567 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3568 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3569 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3570 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3572 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3573 instructions that are handled here.
3574 (@cond_asrd<mode>): Generalize to...
3575 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3576 (*cond_asrd<mode>_2): Generalize to...
3577 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3578 (*cond_asrd<mode>_z): Generalize to...
3579 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3580 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3581 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3582 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3583 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3585 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3586 (@aarch64_scatter_stnt<mode>): Likewise.
3587 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3588 (@aarch64_mul_lane_<mode>): Likewise.
3589 (@aarch64_sve_suqadd<mode>_const): Likewise.
3590 (*<sur>h<addsub><mode>): Generalize to...
3591 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3593 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3594 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3595 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3596 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3597 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3598 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3599 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3600 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3601 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3602 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3603 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3604 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3605 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3606 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3607 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3608 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3609 (@aarch64_sve2_xar<mode>): Likewise.
3610 (@aarch64_sve2_bcax<mode>): Likewise.
3611 (*aarch64_sve2_eor3<mode>): Rename to...
3612 (@aarch64_sve2_eor3<mode>): ...this.
3613 (@aarch64_sve2_bsl<mode>): New expander.
3614 (@aarch64_sve2_nbsl<mode>): Likewise.
3615 (@aarch64_sve2_bsl1n<mode>): Likewise.
3616 (@aarch64_sve2_bsl2n<mode>): Likewise.
3617 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3618 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3619 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3620 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3621 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3622 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3623 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3624 (<su>mull<bt><Vwide>): Generalize to...
3625 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3627 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3628 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3629 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3630 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3631 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3632 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3633 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3634 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3635 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3636 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3637 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3638 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3639 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3640 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3641 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3642 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3643 (<SHRNB:r>shrnb<mode>): Generalize to...
3644 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3646 (<SHRNT:r>shrnt<mode>): Generalize to...
3647 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3649 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3650 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3651 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3652 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3653 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3654 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3655 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3656 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3657 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3658 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3659 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3660 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3661 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3662 (@aarch64_sve2_cvtnt<mode>): Likewise.
3663 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3664 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3665 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3666 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3667 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3668 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3669 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3670 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3671 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3672 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3673 (@aarch64_sve2_pmul<mode>): Likewise.
3674 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3675 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3676 (@aarch64_sve2_tbl2<mode>): Likewise.
3677 (@aarch64_sve2_tbx<mode>): Likewise.
3678 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3679 (@aarch64_sve2_histcnt<mode>): Likewise.
3680 (@aarch64_sve2_histseg<mode>): Likewise.
3681 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3682 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3683 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3684 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3685 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3686 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3687 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3688 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3689 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3690 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3691 (SVE2_PMULL_PAIR_I): New mode iterators.
3692 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3693 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3694 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3695 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3696 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3697 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3698 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3699 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3700 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3701 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3702 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3703 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3704 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3705 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3706 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3707 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3708 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3709 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3710 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3711 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3712 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3713 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3714 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3715 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3716 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3717 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3718 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3719 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3720 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3721 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3722 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3724 (VNARROW, Ventype): New mode attributes.
3725 (Vewtype): Handle VNx2DI. Fix typo in comment.
3726 (VDOUBLE): New mode attribute.
3727 (sve_lane_con): Handle VNx8HI.
3728 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3729 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3730 (sve_int_op, sve_int_op_rev): Handle the above codes.
3731 (sve_pred_int_rhs2_operand): Likewise.
3732 (MULLBT, SHRNB, SHRNT): Delete.
3733 (SVE_INT_SHIFT_IMM): New int iterator.
3734 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3735 and UNSPEC_WHILEHS for TARGET_SVE2.
3736 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3737 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3738 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3739 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3740 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3741 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3742 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3743 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3744 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3745 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3746 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3747 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3748 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3749 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3750 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3751 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3752 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3753 (optab): Handle the new unspecs.
3754 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3756 (lr): Handle the new unspecs.
3758 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3759 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3760 (sve_int_qsub_op): New int attributes.
3761 (sve_fp_op, rot): Handle the new unspecs.
3762 * config/aarch64/aarch64-sve-builtins.h
3763 (function_resolver::require_matching_pointer_type): Declare.
3764 (function_resolver::resolve_unary): Add an optional boolean argument.
3765 (function_resolver::finish_opt_n_resolution): Add an optional
3766 type_suffix_index argument.
3767 (gimple_folder::redirect_call): Declare.
3768 (gimple_expander::prepare_gather_address_operands): Add an optional
3770 * config/aarch64/aarch64-sve-builtins.cc: Include
3771 aarch64-sve-builtins-sve2.h.
3772 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3773 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3774 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3775 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3776 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3777 (TYPES_s_unsigned): Likewise.
3778 (TYPES_s_integer): Use TYPES_s_unsigned.
3779 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3780 (TYPES_sd_integer): Use them.
3781 (TYPES_d_unsigned): New macro.
3782 (TYPES_d_integer): Use it.
3783 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3784 (TYPES_cvt_narrow): Likewise.
3785 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3786 (preds_mx): New variable.
3787 (function_builder::add_overloaded_function): Allow the new feature
3788 set to be more restrictive than the original one.
3789 (function_resolver::infer_pointer_type): Remove qualifiers from
3790 the pointer type before printing it.
3791 (function_resolver::require_matching_pointer_type): New function.
3792 (function_resolver::resolve_sv_displacement): Handle functions
3793 that don't support 32-bit vector indices or svint32_t vector offsets.
3794 (function_resolver::finish_opt_n_resolution): Take the inferred type
3795 as a separate argument.
3796 (function_resolver::resolve_unary): Optionally treat all forms in
3797 the same way as normal merging functions.
3798 (gimple_folder::redirect_call): New function.
3799 (function_expander::prepare_gather_address_operands): Add an argument
3800 that says whether scaled forms are available. If they aren't,
3801 handle scaling of vector indices and don't add the extension and
3803 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3804 fall back to using cond_* instead.
3805 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3806 Split out the member variables into...
3807 (rtx_code_function_base): ...this new base class.
3808 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3809 (unspec_based_function): Split out the member variables into...
3810 (unspec_based_function_base): ...this new base class.
3811 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3812 (unspec_based_function_exact_insn): New class.
3813 (unspec_based_add_function, unspec_based_add_lane_function)
3814 (unspec_based_lane_function, unspec_based_pred_function)
3815 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3816 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3817 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3819 (unspec_based_fused_function): New class.
3820 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3821 (unspec_based_fused_lane_function): New class.
3822 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3824 (CODE_FOR_MODE1): New macro.
3825 (fixed_insn_function): New class.
3826 (while_comparison): Likewise.
3827 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3828 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3829 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3830 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3831 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3832 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3833 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3834 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3835 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3836 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3837 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3838 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3839 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3840 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3841 Also add an initial argument for unary_convert_narrowt, regardless
3842 of the predication type.
3843 (build_32_64): Allow loads and stores to specify MODE_none.
3844 (build_sv_index64, build_sv_uint_offset): New functions.
3845 (long_type_suffix): New function.
3846 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3847 (binary_imm_long_base, load_gather_sv_base): Likewise.
3848 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3849 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3850 (unary_narrowb_base, unary_narrowt_base): Likewise.
3851 (binary_long_lane_def, binary_long_lane): New shape.
3852 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3853 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3854 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3855 (binary_to_uint_def, binary_to_uint): Likewise.
3856 (binary_wide_def, binary_wide): Likewise.
3857 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3858 (compare_def, compare): Likewise.
3859 (compare_ptr_def, compare_ptr): Likewise.
3860 (load_ext_gather_index_restricted_def,
3861 load_ext_gather_index_restricted): Likewise.
3862 (load_ext_gather_offset_restricted_def,
3863 load_ext_gather_offset_restricted): Likewise.
3864 (load_gather_sv_def): Inherit from load_gather_sv_base.
3865 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3866 (shift_left_imm_def, shift_left_imm): Likewise.
3867 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3868 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3869 (store_scatter_index_restricted_def,
3870 store_scatter_index_restricted): Likewise.
3871 (store_scatter_offset_restricted_def,
3872 store_scatter_offset_restricted): Likewise.
3873 (tbl_tuple_def, tbl_tuple): Likewise.
3874 (ternary_long_lane_def, ternary_long_lane): Likewise.
3875 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3876 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3877 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3878 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3879 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3880 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3881 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3882 (ternary_uint_def, ternary_uint): Likewise.
3883 (unary_convert): Fix typo in comment.
3884 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3885 (unary_long_def, unary_long): Likewise.
3886 (unary_narrowb_def, unary_narrowb): Likewise.
3887 (unary_narrowt_def, unary_narrowt): Likewise.
3888 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3889 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3890 (unary_to_int_def, unary_to_int): Likewise.
3891 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3892 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3893 (svasrd_impl): Delete.
3894 (svcadd_impl::expand): Handle integer operations too.
3895 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3896 new functions to derive the unspec numbers.
3897 (svmla_svmls_lane_impl): Replace with...
3898 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3899 integer operations too.
3900 (svwhile_impl): Rename to...
3901 (svwhilelx_impl): ...this and inherit from while_comparison.
3902 (svasrd): Use unspec_based_function.
3903 (svmla_lane): Use svmla_lane_impl.
3904 (svmls_lane): Use svmls_lane_impl.
3905 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3906 (svwhilele, svwhilelt): Use svwhilelx_impl.
3907 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3908 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3909 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3910 * config/aarch64/aarch64-sve-builtins.def: Include
3911 aarch64-sve-builtins-sve2.def.
3913 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3915 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3916 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3917 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3918 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3919 immediates as well as vector ones.
3920 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3921 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3922 (aarch64_sve_qsub_immediate): Update calls accordingly.
3924 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3926 * config/aarch64/aarch64-sve2.md: Add banner comments.
3927 (<su>mulh<r>s<mode>3): Move further up file.
3928 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3929 (*aarch64_sve2_sra<mode>): Move further down file.
3930 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3932 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3934 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3936 (while_optab_cmp): Handle them.
3937 * config/aarch64/aarch64-sve.md
3938 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3939 and add a "@" marker.
3940 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3941 instead of gen_aarch64_sve2_while_ptest.
3942 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3944 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3946 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3947 (UNSPEC_WHILELE): ...this.
3948 (UNSPEC_WHILE_LO): Rename to...
3949 (UNSPEC_WHILELO): ...this.
3950 (UNSPEC_WHILE_LS): Rename to...
3951 (UNSPEC_WHILELS): ...this.
3952 (UNSPEC_WHILE_LT): Rename to...
3953 (UNSPEC_WHILELT): ...this.
3954 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3955 (cmp_op, while_optab_cmp): Likewise.
3956 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3957 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3958 (svwhilelt): Likewise.
3960 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3962 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3963 (unary_to_uint): Define.
3964 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3965 (unary_count): Rename to...
3966 (unary_to_uint_def, unary_to_uint): ...this.
3967 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3969 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3971 * config/aarch64/aarch64-sve-builtins-functions.h
3972 (code_for_mode_function): New class.
3973 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3974 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3975 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3976 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3977 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3979 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3981 * config/aarch64/iterators.md (addsub): New code attribute.
3982 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3984 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3985 in the asm string and attributes. Fix indentation.
3986 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3988 (@aarch64_sve_<optab><mode>): ...this.
3989 * config/aarch64/aarch64-sve-builtins.h
3990 (function_expander::expand_signed_unpred_op): Delete.
3991 * config/aarch64/aarch64-sve-builtins.cc
3992 (function_expander::expand_signed_unpred_op): Likewise.
3993 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3994 try using code_for_aarch64_sve instead.
3995 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3996 (svqsub_impl): Likewise.
3997 (svqadd, svqsub): Use rtx_code_function instead.
3999 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4001 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4002 (HADDSUB, sur, addsub): Remove them.
4004 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4006 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4007 functions the same way, rather than singling out those that
4008 aren't mapped directly to optabs.
4010 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4012 * target.def (compatible_vector_types_p): New target hook.
4013 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4014 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4015 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4016 * doc/tm.texi: Regenerate.
4017 * gimple-expr.c: Include target.h.
4018 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4019 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4021 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4022 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4023 Use the original predicate if it already has a suitable type.
4025 2020-01-09 Martin Jambor <mjambor@suse.cz>
4027 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4028 resolve_speculation and redirect_call_stmt_to_callee static. Change
4029 return type of set_call_stmt to cgraph_edge *.
4030 * auto-profile.c (afdo_indirect_call): Adjust call to
4031 redirect_call_stmt_to_callee.
4032 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4033 make the this pointer explicit, adjust self-recursive calls and the
4034 call top make_direct. Return the resulting edge.
4035 (cgraph_edge::remove): Make this pointer explicit.
4036 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4037 (cgraph_edge::make_direct): Likewise, adjust call to
4038 resolve_speculation.
4039 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4040 call to set_call_stmt.
4041 (cgraph_update_edges_for_call_stmt_node): Update call to
4042 set_call_stmt and remove.
4043 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4044 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4045 (cgraph_node::create_edge_including_clones): Moved "first" definition
4046 of edge to the block where it was used. Adjusted calls to
4048 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4049 cgraph_edge::remove.
4050 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4051 make_direct and redirect_call_stmt_to_callee.
4052 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4053 resolve_speculation and make_direct.
4054 * ipa-inline-transform.c (inline_transform): Adjust call to
4055 redirect_call_stmt_to_callee.
4056 (check_speculations_1):: Adjust call to resolve_speculation.
4057 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4058 resolve-speculation.
4059 (inline_small_functions): Adjust call to resolve_speculation.
4060 (ipa_inline): Likewise.
4061 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4063 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4064 safe with regards to edge removal, adjust calls to
4065 redirect_call_stmt_to_callee.
4066 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4067 and redirect_call_stmt_to_callee.
4068 * multiple_target.c (create_dispatcher_calls): Adjust call to
4069 redirect_call_stmt_to_callee
4070 (redirect_to_specific_clone): Likewise.
4071 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4072 Adjust calls to cgraph_edge::remove.
4073 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4074 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4075 (expand_call_inline): Adjust call to cgraph_edge::remove.
4077 2020-01-09 Martin Liska <mliska@suse.cz>
4079 * params.opt: Set Optimization for
4080 param_max_speculative_devirt_maydefs.
4082 2020-01-09 Martin Sebor <msebor@redhat.com>
4086 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4088 2020-01-09 Martin Liska <mliska@suse.cz>
4090 * auto-profile.c (auto_profile): Use opt_for_fn
4092 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4093 (propagate_vals_across_arith_jfunc): Likewise.
4094 (hint_time_bonus): Likewise.
4095 (incorporate_penalties): Likewise.
4096 (good_cloning_opportunity_p): Likewise.
4097 (perform_estimation_of_a_value): Likewise.
4098 (estimate_local_effects): Likewise.
4099 (ipcp_propagate_stage): Likewise.
4100 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4101 (set_switch_stmt_execution_predicate): Likewise.
4102 (analyze_function_body): Likewise.
4103 * ipa-inline-analysis.c (offline_size): Likewise.
4104 * ipa-inline.c (early_inliner): Likewise.
4105 * ipa-prop.c (ipa_analyze_node): Likewise.
4106 (ipcp_transform_function): Likewise.
4107 * ipa-sra.c (process_scan_results): Likewise.
4108 (ipa_sra_summarize_function): Likewise.
4109 * params.opt: Rename ipcp-unit-growth to
4110 ipa-cp-unit-growth. Add Optimization for various
4111 IPA-related parameters.
4113 2020-01-09 Richard Biener <rguenther@suse.de>
4116 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4118 2020-01-09 Richard Biener <rguenther@suse.de>
4120 PR tree-optimization/93040
4121 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4123 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4125 * common/config/avr/avr-common.c (avr_option_optimization_table)
4126 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4128 2020-01-09 Martin Liska <mliska@suse.cz>
4130 * cgraphclones.c (symbol_table::materialize_all_clones):
4131 Use cgraph_node::dump_name.
4133 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4136 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4137 output_operand_lossage instead of gcc_unreachable.
4138 * doc/md.texi (riscv f constraint): Fix typo.
4141 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4142 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4143 CONST_SCALAR_INT_P instead of CONST_INT_P.
4144 (*subv<mode>4_1): Rename to ...
4145 (subv<mode>4_1): ... this.
4146 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4147 define_insn_and_split patterns.
4148 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4151 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4153 * vec.c (class selftest::count_dtor): New class.
4154 (selftest::test_auto_delete_vec): New test.
4155 (selftest::vec_c_tests): Call it.
4156 * vec.h (class auto_delete_vec): New class template.
4157 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4159 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4161 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4163 2020-01-08 Jim Wilson <jimw@sifive.com>
4165 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4166 use of TLS_MODEL_LOCAL_EXEC when not pic.
4168 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4170 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4173 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4176 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4177 *stack_protect_set_3 peephole2): Also check that the second
4178 insns source is general_operand.
4181 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4182 predicate for output operand instead of register_operand.
4183 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4184 memory destination and non-memory operands[2].
4186 2020-01-08 Martin Liska <mliska@suse.cz>
4188 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4189 ::dump_asm_name instead of (::name or ::asm_name).
4190 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4191 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4192 (analyze_functions): Likewise.
4193 (expand_all_functions): Likewise.
4194 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4195 (propagate_bits_across_jump_function): Likewise.
4196 (dump_profile_updates): Likewise.
4197 (ipcp_store_bits_results): Likewise.
4198 (ipcp_store_vr_results): Likewise.
4199 * ipa-devirt.c (dump_targets): Likewise.
4200 * ipa-fnsummary.c (analyze_function_body): Likewise.
4201 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4202 (process_hsa_functions): Likewise.
4203 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4204 (set_alias_uids): Likewise.
4205 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4206 * ipa-inline.c (recursive_inlining): Likewise.
4207 (inline_to_all_callers_1): Likewise.
4208 (ipa_inline): Likewise.
4209 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4210 (ipa_propagate_frequency): Likewise.
4211 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4212 (remove_described_reference): Likewise.
4213 * ipa-pure-const.c (worse_state): Likewise.
4214 (check_retval_uses): Likewise.
4215 (analyze_function): Likewise.
4216 (propagate_pure_const): Likewise.
4217 (propagate_nothrow): Likewise.
4218 (dump_malloc_lattice): Likewise.
4219 (propagate_malloc): Likewise.
4220 (pass_local_pure_const::execute): Likewise.
4221 * ipa-visibility.c (optimize_weakref): Likewise.
4222 (function_and_variable_visibility): Likewise.
4223 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4224 (ipa_discover_variable_flags): Likewise.
4225 * lto-streamer-out.c (output_function): Likewise.
4226 (output_constructor): Likewise.
4227 * tree-inline.c (copy_bb): Likewise.
4228 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4229 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4231 2020-01-08 Richard Biener <rguenther@suse.de>
4234 * tree-eh.c (sink_clobbers): Update virtual operands for
4235 the first and last stmt only. Add a dry-run capability.
4236 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4237 after CFG manipulations and in RPO order to catch all
4238 secondary opportunities reliably.
4240 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4243 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4245 2019-01-08 Richard Biener <rguenther@suse.de>
4248 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4249 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4250 virtual operand, also updating SSA use.
4251 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4252 Update stmt after resetting virtual operand.
4253 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4254 * gimple-iterator.c (gsi_remove): When not removing the stmt
4255 permanently do not delink immediate uses or mark the stmt modified.
4257 2020-01-08 Martin Liska <mliska@suse.cz>
4259 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4260 (ipa_call_context::estimate_size_and_time): Likewise.
4261 (inline_analyze_function): Likewise.
4263 2020-01-08 Martin Liska <mliska@suse.cz>
4265 * cgraph.c (cgraph_node::dump): Use systematically
4268 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4270 Add -nodevicespecs option for avr.
4273 * config/avr/avr.opt (-nodevicespecs): New driver option.
4274 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4275 "-specs=device-specs/..." if that option is not set.
4276 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4278 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4280 Implement 64-bit double functions for avr.
4283 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4284 --with-double-comparison.
4285 * doc/install.texi: Document them.
4286 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4287 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4288 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4289 * doc/invoke.texi (AVR Built-in Macros): Document them.
4290 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4291 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4292 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4294 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4297 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4298 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4299 when only building rm-profile multilibs.
4301 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4304 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4305 lattice for a value to check.
4306 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4307 finite propagation in self-recursive scc.
4309 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4311 * ipa-inline.c (caller_growth_limits): Restore the AND.
4313 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4315 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4316 (VEC_ALLREG_ALT): New iterator.
4317 (VEC_ALLREG_INT_MODE): New iterator.
4318 (VCMP_MODE): New iterator.
4319 (VCMP_MODE_INT): New iterator.
4320 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4321 (vec_cmp<u>v64qidi): New define_expand.
4322 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4323 (vec_cmpu<mode>di_exec): New define_expand.
4324 (vec_cmp<u>v64qidi_exec): New define_expand.
4325 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4326 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4327 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4328 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4329 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4330 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4331 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4332 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4333 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4334 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4336 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4337 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4339 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4341 * config/gcn/constraints.md (DA): Update description and match.
4343 (Db): New constraint.
4344 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4346 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4347 Implement 'Db' mixed immediate type.
4348 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4349 (addcv64si3_dup<exec_vcc>): Delete.
4350 (subcv64si3<exec_vcc>): Rework constraints.
4351 (addv64di3): Rework constraints.
4352 (addv64di3_exec): Rework constraints.
4353 (subv64di3): Rework constraints.
4354 (addv64di3_dup): Delete.
4355 (addv64di3_dup_exec): Delete.
4356 (addv64di3_zext): Rework constraints.
4357 (addv64di3_zext_exec): Rework constraints.
4358 (addv64di3_zext_dup): Rework constraints.
4359 (addv64di3_zext_dup_exec): Rework constraints.
4360 (addv64di3_zext_dup2): Rework constraints.
4361 (addv64di3_zext_dup2_exec): Rework constraints.
4362 (addv64di3_sext_dup2): Rework constraints.
4363 (addv64di3_sext_dup2_exec): Rework constraints.
4365 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4367 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4368 existing target checks.
4370 2020-01-07 Richard Biener <rguenther@suse.de>
4372 * doc/install.texi: Bump minimal supported MPC version.
4374 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4376 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4377 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4378 * langhooks.c: Include stor-layout.h.
4379 (lhd_simulate_enum_decl): New function.
4380 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4381 handle_arm_sve_h for the LTO frontend.
4382 (register_vector_type): Cope with null returns from pushdecl.
4384 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4386 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4387 (aarch64_sve::nvectors_if_data_type): Replace with...
4388 (aarch64_sve::builtin_type_p): ...this.
4389 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4390 (find_vector_type): Delete.
4391 (add_sve_type_attribute): New function.
4392 (lookup_sve_type_attribute): Likewise.
4393 (register_builtin_types): Add an "SVE type" attribute to each type.
4394 (register_tuple_type): Likewise.
4395 (svbool_type_p, nvectors_if_data_type): Delete.
4396 (mangle_builtin_type): Use lookup_sve_type_attribute.
4397 (builtin_type_p): Likewise. Add an overload that returns the
4398 number of constituent vector and predicate registers.
4399 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4400 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4401 instead of aarch64_sve_argument_p.
4402 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4403 (aarch64_pass_by_reference): Likewise.
4404 (aarch64_function_value_1): Likewise.
4405 (aarch64_return_in_memory): Likewise.
4406 (aarch64_layout_arg): Likewise.
4408 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4410 PR tree-optimization/93156
4411 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4412 least significant bit is always clear.
4414 PR tree-optimization/93118
4415 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4416 simplifier with two intermediate conversions.
4418 2020-01-07 Martin Liska <mliska@suse.cz>
4420 * params.opt: Add Optimization for various parameters.
4422 2020-01-07 Martin Liska <mliska@suse.cz>
4425 * doc/extend.texi: Explain cloning for target_clone
4428 2020-01-07 Martin Liska <mliska@suse.cz>
4430 PR tree-optimization/92860
4431 * common.opt: Make in Optimization option
4432 as it is affected by -O0, which is an Optimization
4434 * tree-inline.c (tree_inlinable_function_p):
4435 Use opt_for_fn for warn_inline.
4436 (expand_call_inline): Likewise.
4438 2020-01-07 Martin Liska <mliska@suse.cz>
4440 PR tree-optimization/92860
4441 * common.opt: Make flag_ree as optimization
4444 2020-01-07 Martin Liska <mliska@suse.cz>
4446 PR optimization/92860
4447 * params.opt: Mark param_min_crossjump_insns with Optimization
4450 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4452 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4453 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4455 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4457 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4458 helper function to return the valid addressing formats for a given
4459 hard register and mode.
4460 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4462 * config/rs6000/constraints.md (Q constraint): Update
4464 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4467 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4468 Use 'Q' for doing vector extract from memory.
4469 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4471 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4472 doing vector extract from memory.
4473 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4474 extract from memory.
4476 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4477 for the offset being 34-bits when -mcpu=future is used.
4479 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4481 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4482 instead of cmpib_comparison_operator in cmpib patterns.
4483 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4484 of cmpib_comparison_operator. Revise comment.
4486 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4488 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4489 in an IFN_DIV_POW2 node to be equal.
4491 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4493 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4494 (vect_check_scalar_mask): ...this.
4495 (vectorizable_store, vectorizable_load): Update call accordingly.
4496 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4497 argument in calls to conditional internal functions.
4499 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4501 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4502 '0' matching inputs.
4503 (subv64di3_exec): Likewise.
4505 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4507 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4508 * doc/md.texi (movstr): Likewise.
4510 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4512 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4515 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4517 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4519 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4520 to a temporary file and use move-if-change to update the real
4521 file where necessary.
4523 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4525 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4526 rather than Upa for CPY /M.
4528 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4530 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4533 2020-01-06 Martin Liska <mliska@suse.cz>
4535 PR tree-optimization/92860
4536 * params.opt: Mark param_max_combine_insns with Optimization
4539 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4542 * config/i386/i386.md (SWIDWI): New mode iterator.
4543 (DWI, dwi): Add TImode variants.
4544 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4545 <general_hilo_operand> instead of <general_operand>. Use
4546 CONST_SCALAR_INT_P instead of CONST_INT_P.
4547 (*addv<mode>4_1): Rename to ...
4548 (addv<mode>4_1): ... this.
4549 (QWI): New mode attribute.
4550 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4551 define_insn_and_split patterns.
4552 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4554 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4555 <general_hilo_operand> instead of <general_operand>.
4556 (*addcarry<mode>_1): New define_insn.
4557 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4559 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4561 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4562 Use "call" instead of "set".
4564 2020-01-03 Martin Jambor <mjambor@suse.cz>
4567 * ipa-cp.c (print_all_lattices): Skip functions without info.
4569 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4572 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4573 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4574 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4575 for 'e' simd clones.
4578 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4580 (mprefer-vector-width=): Add Save.
4581 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4582 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4583 (ix86_debug_options, ix86_function_specific_print): Adjust
4584 ix86_target_string callers.
4585 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4586 (ix86_valid_target_attribute_tree): Likewise.
4587 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4588 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4589 ix86_target_string caller.
4592 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4593 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4594 instead of gen_int_shift_amount + convert_modes.
4596 PR rtl-optimization/93088
4597 * loop-iv.c (find_single_def_src): Punt after looking through
4598 128 reg copies for regs with single definitions. Move definitions
4601 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4603 * config/arm/arm-c.c (arm_cpu_builtins): Define
4604 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4605 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4606 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4607 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4608 * config/arm/arm-tables.opt: Regenerated.
4609 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4610 arm_arch_i8mm and arm_arch_bf16 when enabled.
4611 * config/arm/arm.h (TARGET_I8MM): New macro.
4612 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4613 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4614 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4615 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4616 (v8_6_a_simd_variants): New.
4617 (v8_*_a_simd_variants): Add i8mm and bf16.
4618 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4620 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4623 * predict.c (compute_function_frequency): Don't call
4624 warn_function_cold on functions that already have cold attribute.
4626 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4629 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4630 COMDAT group function labels in .data.rel.ro.local section.
4631 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4634 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4635 comparison_operator in B and S integer comparisons. Likewise, use
4636 ordered_comparison_operator instead of cmpib_comparison_operator in
4638 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4640 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4642 Update copyright years.
4644 * gcc.c (process_command): Update copyright notice dates.
4645 * gcov-dump.c (print_version): Ditto.
4646 * gcov.c (print_version): Ditto.
4647 * gcov-tool.c (print_version): Ditto.
4648 * gengtype.c (create_file): Ditto.
4649 * doc/cpp.texi: Bump @copying's copyright year.
4650 * doc/cppinternals.texi: Ditto.
4651 * doc/gcc.texi: Ditto.
4652 * doc/gccint.texi: Ditto.
4653 * doc/gcov.texi: Ditto.
4654 * doc/install.texi: Ditto.
4655 * doc/invoke.texi: Ditto.
4657 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4659 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4662 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4664 PR tree-optimization/93098
4665 * match.pd (popcount): For shift amounts, use integer_onep
4666 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4667 tests. Make sure that precision is power of two larger than or equal
4668 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4669 instead of ULL suffixed constants. Formatting fixes.
4671 Copyright (C) 2020 Free Software Foundation, Inc.
4673 Copying and distribution of this file, with or without modification,
4674 are permitted in any medium without royalty provided the copyright
4675 notice and this notice are preserved.