1 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
3 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
5 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
7 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
10 (VQ): Enable for V8BF.
13 (VDBL, Vdbl): Add V4BF.
14 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
15 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
16 (bfloat16x8x2_t): Likewise.
17 (bfloat16x4x3_t): Likewise.
18 (bfloat16x8x3_t): Likewise.
19 (bfloat16x4x4_t): Likewise.
20 (bfloat16x8x4_t): Likewise.
22 (vld1_bf16, vld1_bf16_x2): New.
23 (vld1_bf16_x3, vld1_bf16_x4): New.
24 (vld1q_bf16, vld1q_bf16_x2): New.
25 (vld1q_bf16_x3, vld1q_bf16_x4): New.
26 (vld1_lane_bf16): New.
27 (vld1q_lane_bf16): New.
29 (vld1q_dup_bf16): New.
33 (vld2q_dup_bf16): New.
37 (vld3q_dup_bf16): New.
41 (vld4q_dup_bf16): New.
42 (vst1_bf16, vst1_bf16_x2): New.
43 (vst1_bf16_x3, vst1_bf16_x4): New.
44 (vst1q_bf16, vst1q_bf16_x2): New.
45 (vst1q_bf16_x3, vst1q_bf16_x4): New.
46 (vst1_lane_bf16): New.
47 (vst1q_lane_bf16): New.
55 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
57 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
59 (VALLDI_F16): Likewise.
62 (vswap_width_name): Likewise.
63 (VSWAP_WIDTH): Likewise.
67 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
68 (vget_lane_bf16, vgetq_lane_bf16): New.
70 (vdup_n_bf16, vdupq_n_bf16): New.
71 (vdup_lane_bf16, vdup_laneq_bf16): New.
72 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
73 (vduph_lane_bf16, vduph_laneq_bf16): New.
74 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
75 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
76 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
77 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
78 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
79 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
80 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
81 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
82 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
83 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
84 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
85 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
86 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
87 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
88 (vreinterpretq_bf16_p128): New.
89 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
90 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
91 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
92 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
93 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
94 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
95 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
96 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
97 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
98 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
99 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
100 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
101 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
102 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
103 (vreinterpretq_p128_bf16): New.
105 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
107 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
108 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
109 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
110 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
111 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
112 * config/arm/iterators.md (VSF2BF): New attribute.
113 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
114 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
115 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
117 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
119 * config/arm/arm.md (required_for_purecode): New attribute.
120 (enabled): Handle required_for_purecode.
121 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
122 work with -mpure-code.
124 2020-02-25 Jakub Jelinek <jakub@redhat.com>
126 PR rtl-optimization/93908
127 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
130 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
132 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
134 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
136 * doc/install.texi (--enable-checking): Adjust wording.
138 2020-02-25 Richard Biener <rguenther@suse.de>
140 PR tree-optimization/93868
141 * tree-vect-slp.c (slp_copy_subtree): New function.
142 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
143 re-arranging stmts in it.
145 2020-02-25 Jakub Jelinek <jakub@redhat.com>
148 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
149 dummy function and remove it at the end.
152 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
153 paramter -> parameter.
154 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
155 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
157 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
159 * doc/install.texi (--enable-checking): Properly document current
161 (--enable-stage1-checking): Minor clarification about bootstrap.
163 2020-02-24 David Malcolm <dmalcolm@redhat.com>
166 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
167 -fanalyzer-checker=taint is also required.
168 (-fanalyzer-checker=): Note that providing this option enables the
169 given checker, and doing so may be required for checkers that are
172 2020-02-24 David Malcolm <dmalcolm@redhat.com>
174 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
175 significant control flow events; add a "3" which shows all
176 control flow events; the old "3" becomes "4".
178 2020-02-24 Jakub Jelinek <jakub@redhat.com>
180 PR tree-optimization/93582
181 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
182 pd.offset and pd.size to be counted in bits rather than bytes, add
183 support for maxsizei that is not a multiple of BITS_PER_UNIT and
184 handle bitfield stores and loads.
185 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
186 uncomparable quantities - bytes vs. bits. Allow push_partial_def
187 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
188 pd.offset/pd.size to be counted in bits rather than bytes.
189 Formatting fix. Rename shadowed len variable to buflen.
191 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
192 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
195 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
196 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
197 * opts-common.c (parse_options_from_collect_gcc_options): New function.
198 (prepend_xassembler_to_collect_as_options): Likewise.
199 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
200 (prepend_xassembler_to_collect_as_options): Likewise.
201 * lto-opts.c (lto_write_options): Stream assembler options
202 in COLLECT_AS_OPTIONS.
203 * lto-wrapper.c (xassembler_options_error): New static variable.
204 (get_options_from_collect_gcc_options): Move parsing options code to
205 parse_options_from_collect_gcc_options and call it.
206 (merge_and_complain): Validate -Xassembler options.
207 (append_compiler_options): Handle OPT_Xassembler.
208 (run_gcc): Append command line -Xassembler options to
210 * doc/invoke.texi: Add documentation about using Xassembler
213 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
215 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
217 (riscv_rtx_costs): Update cost model for LTGT.
219 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
221 PR rtl-optimization/93564
222 * ira-color.c (struct update_cost_queue_elem): New member start.
223 (queue_update_cost, get_next_update_cost): Add new arg start.
224 (allocnos_conflict_p): New function.
225 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
226 Add checking conflicts with allocnos_conflict_p.
227 (update_costs_from_prefs, restore_costs_from_copies): Adjust
228 update_costs_from_allocno calls.
229 (update_conflict_hard_regno_costs): Add checking conflicts with
230 allocnos_conflict_p. Adjust calls of queue_update_cost and
231 get_next_update_cost.
232 (assign_hard_reg): Adjust calls of queue_update_cost. Add
234 (bucket_allocno_compare_func): Restore previous version.
236 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
238 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
239 double-word size when handling aggregate return values.
240 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
241 that homogeneous SFmode and DFmode aggregates are passed and returned
242 in general registers.
244 2020-02-21 Jakub Jelinek <jakub@redhat.com>
247 * opts.c (print_filtered_help): Translate help before appending
248 messages to it rather than after that.
250 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
252 PR rtl-optimization/PR92989
253 * lra-lives.c (process_bb_lives): Restore the original order
254 of the bb liveness update. Call make_hard_regno_dead for each
255 register clobbered at the start of an EH receiver.
257 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
260 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
261 self-recursively generated.
263 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
266 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
269 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
271 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
272 Document new target supports option.
274 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
276 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
277 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
278 * config/arm/iterators.md (MATMUL): New iterator.
279 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
280 (mmla_sfx): New attribute.
281 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
282 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
283 (UNSPEC_MATMUL_US): New.
285 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
287 * config/arm/arm.md: Prevent scalar shifts from being used when big
290 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
291 Richard Biener <rguenther@suse.de>
293 PR tree-optimization/93586
294 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
295 after mismatched array refs; do not sure type size information to
296 recover from unmatched referneces with !flag_strict_aliasing_p.
298 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
300 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
301 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
302 (scatter_store<mode>): Rename to ...
303 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
304 (scatter<mode>_exec): Delete. Move contents ...
305 (mask_scatter_store<mode>): ... here, and rename that to ...
306 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
307 Remove mode conversion.
308 (mask_gather_load<mode>): Rename to ...
309 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
310 Remove mode conversion.
311 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
313 2020-02-21 Martin Jambor <mjambor@suse.cz>
315 PR tree-optimization/93845
316 * tree-sra.c (verify_sra_access_forest): Only test access size of
319 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
321 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
322 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
323 (addv64di3_exec): Likewise.
324 (subv64di3): Likewise.
325 (subv64di3_exec): Likewise.
326 (addv64di3_zext): Likewise.
327 (addv64di3_zext_exec): Likewise.
328 (addv64di3_zext_dup): Likewise.
329 (addv64di3_zext_dup_exec): Likewise.
330 (addv64di3_zext_dup2): Likewise.
331 (addv64di3_zext_dup2_exec): Likewise.
332 (addv64di3_sext_dup2): Likewise.
333 (addv64di3_sext_dup2_exec): Likewise.
334 (<expander>v64di3): Likewise.
335 (<expander>v64di3_exec): Likewise.
336 (*<reduc_op>_dpp_shr_v64di): Likewise.
337 (*plus_carry_dpp_shr_v64di): Likewise.
338 * config/gcn/gcn.md (adddi3): Likewise.
339 (addptrdi3): Likewise.
340 (<expander>di3): Likewise.
342 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
344 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
346 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
348 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
349 support. Use aarch64_emit_mult instead of emitting multiplication
350 instructions directly.
351 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
352 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
354 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
356 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
357 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
358 instead of emitting multiplication instructions directly.
359 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
360 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
361 (@aarch64_frecps<mode>): New expanders.
363 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
365 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
366 on and produce uint64_ts rather than ints.
367 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
368 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
370 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
372 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
373 an unused xmsk register when handling approximate rsqrt.
375 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
377 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
378 flag_finite_math_only condition.
380 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
383 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
384 to destination operand for shufps alternative.
385 (*vec_extractv2si_1): Ditto.
387 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
390 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
393 2020-02-20 Martin Liska <mliska@suse.cz>
396 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
398 2020-02-20 Martin Liska <mliska@suse.cz>
401 * common/config/avr/avr-common.c: Remote trailing "|".
403 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
405 * collect2.c (maybe_run_lto_and_relink): Fix typo in
408 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
410 PR tree-optimization/93767
411 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
412 access-size bias from the offset calculations for negative strides.
414 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
416 * collect2.c (c_file, o_file): Make const again.
417 (ldout,lderrout, dump_ld_file): Remove.
418 (tool_cleanup): Avoid calling not signal-safe functions.
419 (maybe_run_lto_and_relink): Avoid possible signal handler
420 access to unintialzed memory (lto_o_files).
421 (main): Avoid leaking temp files in $TMPDIR.
422 Initialize c_file/o_file with concat, which avoids exposing
423 uninitialized memory to signal handler, which calls unlink(!).
424 Avoid calling maybe_unlink when the main function returns,
425 since the atexit handler is already doing this.
426 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
428 2020-02-19 Martin Jambor <mjambor@suse.cz>
430 PR tree-optimization/93776
431 * tree-sra.c (create_access): Do not create zero size accesses.
432 (get_access_for_expr): Do not search for zero sized accesses.
434 2020-02-19 Martin Jambor <mjambor@suse.cz>
436 PR tree-optimization/93667
437 * tree-sra.c (scalarizable_type_p): Return false if record fields
438 do not follow wach other.
440 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
442 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
443 rather than fmv.x.s/fmv.s.x.
445 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
447 * config/aarch64/aarch64-simd-builtins.def
448 (intrinsic_vec_smult_lo_): New.
449 (intrinsic_vec_umult_lo_): Likewise.
450 (vec_widen_smult_hi_): Likewise.
451 (vec_widen_umult_hi_): Likewise.
452 * config/aarch64/aarch64-simd.md
453 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
454 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
455 (vmull_high_s16): Likewise.
456 (vmull_high_s32): Likewise.
457 (vmull_high_u8): Likewise.
458 (vmull_high_u16): Likewise.
459 (vmull_high_u32): Likewise.
460 (vmull_s8): Likewise.
461 (vmull_s16): Likewise.
462 (vmull_s32): Likewise.
463 (vmull_u8): Likewise.
464 (vmull_u16): Likewise.
465 (vmull_u32): Likewise.
467 2020-02-18 Martin Liska <mliska@suse.cz>
469 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
470 bootstrap by missing removal of invalid sanity check.
472 2020-02-18 Martin Liska <mliska@suse.cz>
475 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
476 Always compare LHS of gimple_assign.
478 2020-02-18 Martin Liska <mliska@suse.cz>
481 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
482 and return type of functions.
483 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
484 Drop MALLOC attribute for void functions.
485 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
486 malloc_state for a new VOID clone.
488 2020-02-18 Martin Liska <mliska@suse.cz>
491 * common.opt: Add -fprofile-reproducibility.
492 * doc/invoke.texi: Document it.
493 * value-prof.c (dump_histogram_value):
494 Document and support behavior for counters[0]
495 being a negative value.
496 (get_nth_most_common_value): Handle negative
497 counters[0] in respect to flag_profile_reproducible.
499 2020-02-18 Jakub Jelinek <jakub@redhat.com>
502 * cgraph.c (verify_speculative_call): Use speculative_id instead of
503 speculative_uid in messages. Remove trailing whitespace from error
504 message. Use num_speculative_call_targets instead of
505 num_speculative_targets in a message.
506 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
507 edge messages and stmt instead of cal_stmt in reference message.
509 PR tree-optimization/93780
510 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
511 before calling build_vector_type.
512 (execute_update_addresses_taken): Likewise.
515 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
516 typo, functoin -> function.
517 * tree.c (free_lang_data_in_decl): Fix comment typo,
518 functoin -> function.
519 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
521 2020-02-17 David Malcolm <dmalcolm@redhat.com>
523 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
525 (print_option_information): Don't call get_option_url if URLs
528 2020-02-17 Alexandre Oliva <oliva@adacore.com>
530 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
531 handling of register_common-less targets.
533 2020-02-17 Martin Liska <mliska@suse.cz>
536 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
538 2020-02-17 Martin Liska <mliska@suse.cz>
541 * config/rs6000/rs6000.c (rs6000_option_override_internal):
544 2020-02-17 Martin Liska <mliska@suse.cz>
547 * config/rx/elf.opt: Fix typo.
549 2020-02-17 Richard Biener <rguenther@suse.de>
552 * opts-global.c (print_ignored_options): Use inform and
555 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
558 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
560 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
563 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
564 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
566 2020-02-15 Jason Merrill <jason@redhat.com>
568 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
570 2020-02-15 Jakub Jelinek <jakub@redhat.com>
572 PR tree-optimization/93744
573 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
574 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
575 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
576 sure @2 in the first and @1 in the other patterns has no side-effects.
578 2020-02-15 David Malcolm <dmalcolm@redhat.com>
579 Bernd Edlinger <bernd.edlinger@hotmail.de>
583 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
584 * configure.ac (--with-diagnostics-urls): New configuration
585 option, based on --with-diagnostics-color.
586 (DIAGNOSTICS_URLS_DEFAULT): New define.
587 * config.h: Regenerate.
588 * configure: Regenerate.
589 * diagnostic.c (diagnostic_urls_init): Handle -1 for
590 DIAGNOSTICS_URLS_DEFAULT from configure-time
591 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
592 and TERM_URLS environment variable.
593 * diagnostic-url.h (diagnostic_url_format): New enum type.
594 (diagnostic_urls_enabled_p): rename to...
595 (determine_url_format): ... this, and change return type.
596 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
597 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
598 the linux console, and mingw.
599 (diagnostic_urls_enabled_p): rename to...
600 (determine_url_format): ... this, and adjust.
601 * pretty-print.h (pretty_printer::show_urls): rename to...
602 (pretty_printer::url_format): ... this, and change to enum.
603 * pretty-print.c (pretty_printer::pretty_printer,
604 pp_begin_url, pp_end_url, test_urls): Adjust.
605 * doc/install.texi (--with-diagnostics-urls): Document the new
606 configuration option.
607 (--with-diagnostics-color): Document the existing interaction
608 with GCC_COLORS better.
609 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
610 vindex reference. Update description of defaults based on the above.
611 (-fdiagnostics-color): Update description of how -fdiagnostics-color
612 interacts with GCC_COLORS.
614 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
617 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
618 conjunction with TARGET_GNU_TLS in early return.
620 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
622 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
623 the mode is not wider than UNITS_PER_WORD.
625 2020-02-14 Martin Jambor <mjambor@suse.cz>
627 PR tree-optimization/93516
628 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
629 access of the same type as the parent.
630 (propagate_subaccesses_from_lhs): Likewise.
632 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
635 * config/i386/avx512vbmi2intrin.h
636 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
637 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
638 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
639 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
640 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
641 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
642 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
643 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
644 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
645 of lacking a closing parenthesis.
646 * config/i386/avx512vbmi2vlintrin.h
647 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
648 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
649 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
650 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
651 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
652 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
653 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
654 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
655 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
656 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
657 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
658 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
659 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
660 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
661 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
662 _mm_shldi_epi32, _mm_mask_shldi_epi32,
663 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
664 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
666 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
669 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
670 the target function entry.
672 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
674 * common/config/arc/arc-common.c (arc_option_optimization_table):
675 Disable if-conversion step when optimized for size.
677 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
679 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
680 R12-R15 are always in ARCOMPACT16_REGS register class.
681 * config/arc/arc.opt (mq-class): Deprecate.
682 * config/arc/constraint.md ("q"): Remove dependency on mq-class
684 * doc/invoke.texi (mq-class): Update text.
685 * common/config/arc/arc-common.c (arc_option_optimization_table):
688 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
690 * config/arc/arc.c (arc_insn_cost): New function.
691 (TARGET_INSN_COST): Define.
692 * config/arc/arc.md (cost): New attribute.
693 (add_n): Use arc_nonmemory_operand.
694 (ashlsi3_insn): Likewise, also update constraints.
695 (ashrsi3_insn): Likewise.
697 (add_shift): Likewise.
698 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
700 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
702 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
704 (umulsidi_600): Likewise.
706 2020-02-13 Jakub Jelinek <jakub@redhat.com>
709 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
710 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
711 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
712 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
713 pass __A to the builtin followed by __W instead of __A followed by
715 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
716 _mm512_mask_popcnt_epi64): Likewise.
717 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
718 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
719 _mm256_mask_popcnt_epi64): Likewise.
721 PR tree-optimization/93582
722 * fold-const.h (shift_bytes_in_array_left,
723 shift_bytes_in_array_right): Declare.
724 * fold-const.c (shift_bytes_in_array_left,
725 shift_bytes_in_array_right): New function, moved from
726 gimple-ssa-store-merging.c, no longer static.
727 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
728 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
729 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
730 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
731 shift_bytes_in_array.
732 (verify_shift_bytes_in_array): Rename to ...
733 (verify_shift_bytes_in_array_left): ... this. Use
734 shift_bytes_in_array_left instead of shift_bytes_in_array.
735 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
736 instead of verify_shift_bytes_in_array.
737 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
738 / native_interpret_expr where the store covers all needed bits,
739 punt on PDP-endian, otherwise allow all involved offsets and sizes
740 not to be byte-aligned.
743 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
744 use const_0_to_255_operand predicate instead of immediate_operand.
745 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
746 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
747 vgf2p8affineinvqb_<mode><mask_name>,
748 vgf2p8affineqb_<mode><mask_name>): Drop mode from
749 const_0_to_255_operand predicated operands.
751 2020-02-12 Jeff Law <law@redhat.com>
753 * config/h8300/h8300.md (comparison shortening peepholes): Use
754 a mode iterator to merge the HImode and SImode peepholes.
756 2020-02-12 Jakub Jelinek <jakub@redhat.com>
759 * real.c (is_even): Make static. Function comment fix.
760 (is_halfway_below): Make static, don't assert R is not inf/nan,
761 instead return false for those. Small formatting fixes.
763 2020-02-12 Martin Sebor <msebor@redhat.com>
766 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
767 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
768 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
769 (strlen_check_and_optimize_call): Adjust callee name.
771 2020-02-12 Jeff Law <law@redhat.com>
773 * config/h8300/h8300.md (comparison shortening peepholes): Drop
774 (and (xor)) variant. Combine other two into single peephole.
776 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
778 PR rtl-optimization/93565
779 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
781 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
783 * config/aarch64/aarch64-simd.md
784 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
785 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
786 generating separate ADDV and zero_extend patterns.
787 * config/aarch64/iterators.md (VDQV_E): New iterator.
789 2020-02-12 Jeff Law <law@redhat.com>
791 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
792 expanders, splits, etc.
793 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
794 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
795 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
796 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
797 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
798 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
800 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
802 2020-02-12 Jakub Jelinek <jakub@redhat.com>
805 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
806 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
807 TARGET_AVX512DQ from condition.
808 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
809 instead of <mask_mode512bit_condition> in condition. If
810 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
812 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
815 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
818 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
820 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
822 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
823 where strlen is more legible.
824 (rs6000_builtin_vectorized_libmass): Ditto.
825 (rs6000_print_options_internal): Ditto.
827 2020-02-11 Martin Sebor <msebor@redhat.com>
829 PR tree-optimization/93683
830 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
832 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
834 * config/rs6000/predicates.md (cint34_operand): Rename the
835 -mprefixed-addr option to be -mprefixed.
836 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
837 the -mprefixed-addr option to be -mprefixed.
838 (OTHER_FUTURE_MASKS): Likewise.
839 (POWERPC_MASKS): Likewise.
840 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
841 the -mprefixed-addr option to be -mprefixed. Change error
842 messages to refer to -mprefixed.
843 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
845 (rs6000_legitimate_offset_address_p): Likewise.
846 (rs6000_mode_dependent_address): Likewise.
847 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
848 "-mprefixed" for target attributes and pragmas.
849 (address_to_insn_form): Rename the -mprefixed-addr option to be
851 (rs6000_adjust_insn_length): Likewise.
852 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
853 -mprefixed-addr option to be -mprefixed.
854 (ASM_OUTPUT_OPCODE): Likewise.
855 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
856 -mprefixed-addr option to be -mprefixed.
857 * config/rs6000/rs6000.opt (-mprefixed): Rename the
858 -mprefixed-addr option to be prefixed. Change the option from
859 being undocumented to being documented.
860 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
861 -mprefixed option. Update the -mpcrel documentation to mention
864 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
866 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
867 including FIRST_PSEUDO_REGISTER - 1.
868 * ira-color.c (print_hard_reg_set): Ditto.
870 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
872 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
873 (USTERNOP_QUALIFIERS): New define.
874 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
875 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
876 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
877 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
878 * config/arm/arm_neon.h (vusdot_s32): New.
879 (vusdot_lane_s32): New.
880 (vusdotq_lane_s32): New.
881 (vsudot_lane_s32): New.
882 (vsudotq_lane_s32): New.
883 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
884 * config/arm/iterators.md (DOTPROD_I8MM): New.
885 (sup, opsuffix): Add <us/su>.
886 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
887 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
889 2020-02-11 Richard Biener <rguenther@suse.de>
891 PR tree-optimization/93661
892 PR tree-optimization/93662
893 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
895 * tree-sra.c (get_access_for_expr): Likewise.
897 2020-02-10 Jakub Jelinek <jakub@redhat.com>
900 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
901 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
902 Change condition from TARGET_AVX2 to TARGET_AVX.
904 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
907 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
910 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
912 Try to generate zero-based comparisons.
913 * config/cris/cris.c (cris_reduce_compare): New function.
914 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
915 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
916 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
918 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
921 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
922 in Thumb state and also as a destination in Arm state. Add T16
925 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
927 * md.texi (Define Subst): Match closing paren in example.
929 2020-02-10 Jakub Jelinek <jakub@redhat.com>
933 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
934 arguments of strncmp.
936 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
939 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
940 but different source value.
941 (adjust_callers_for_value_intersection): New function.
942 (gather_edges_for_value): Adjust order of callers to let a
943 non-self-recursive caller be the first element.
944 (self_recursive_pass_through_p): Add a new parameter "simple", and
945 check generalized self-recursive pass-through jump function.
946 (self_recursive_agg_pass_through_p): Likewise.
947 (find_more_scalar_values_for_callers_subset): Compute value from
948 pass-through jump function for self-recursive.
949 (intersect_with_plats): Cleanup previous implementation code for value
950 itersection with self-recursive call edge.
951 (intersect_with_agg_replacements): Likewise.
952 (intersect_aggregates_with_edge): Deduce value from pass-through jump
953 function for self-recursive call edge. Cleanup previous implementation
954 code for value intersection with self-recursive call edge.
955 (decide_whether_version_node): Remove dead callers and adjust order
956 to let a non-self-recursive caller be the first element.
958 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
960 * recog.c: Move pass_split_before_sched2 code in front of
961 pass_split_before_regstack.
962 (pass_data_split_before_sched2): Rename pass to split3 from split4.
963 (pass_data_split_before_regstack): Rename pass to split4 from split3.
964 (rest_of_handle_split_before_sched2): Remove.
965 (pass_split_before_sched2::execute): Unconditionally call
967 (enable_split_before_sched2): New function.
968 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
969 (pass_split_before_regstack::gate): Ditto.
970 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
971 Update name check for renamed split4 pass.
972 * config/sh/sh.c (register_sh_passes): Update pass insertion
973 point for renamed split4 pass.
975 2020-02-09 Jakub Jelinek <jakub@redhat.com>
977 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
978 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
979 copying them around between host and target.
981 2020-02-08 Andrew Pinski <apinski@marvell.com>
984 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
985 STRICT_ALIGNMENT also.
987 2020-02-08 Jim Wilson <jimw@sifive.com>
990 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
992 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
993 Jakub Jelinek <jakub@redhat.com>
996 * config/i386/i386.h (CALL_USED_REGISTERS): Make
997 xmm16-xmm31 call-used even in 64-bit ms-abi.
999 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1001 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1002 (simd_ummla, simd_usmmla): Likewise.
1003 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1004 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1005 (vusmmlaq_s32): New.
1007 2020-02-07 Richard Biener <rguenther@suse.de>
1010 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1011 skipping unreachable regions.
1012 (optimize_inline_calls): Skip folding stmts when we didn't
1015 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1018 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1019 Don't return aggregates with only SFmode and DFmode in SSE
1021 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1023 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1026 * config/rs6000/rs6000-logue.c
1027 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1028 if it fails, move rs into end_addr and retry. Add
1029 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1030 the insn pattern doesn't describe well what exactly happens to
1034 * config/i386/predicates.md (avx_identity_operand): Remove.
1035 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1036 (avx_<castmode><avxsizesuffix>_<castmode>,
1037 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1038 a VEC_CONCAT of the operand and UNSPEC_CAST.
1039 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1040 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1044 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1045 recog_data.insn if distance_non_agu_define changed it.
1047 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1050 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1051 we only had X-FORM (reg+reg) addressing for vectors. Also before
1052 ISA 3.0, we only had X-FORM addressing for scalars in the
1053 traditional Altivec registers.
1055 2020-02-06 <zhongyunde@huawei.com>
1056 Vladimir Makarov <vmakarov@redhat.com>
1058 PR rtl-optimization/93561
1059 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1060 hard register range.
1062 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1064 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1067 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1069 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1070 where the low and the high 32 bits are equal to each other specially,
1071 with an rldimi instruction.
1073 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1075 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1077 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1079 * config/arm/arm-tables.opt: Regenerate.
1081 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1084 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1085 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1086 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1088 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1090 PR rtl-optimization/87763
1091 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1093 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1095 * config/aarch64/aarch64-simd-builtins.def
1096 (bfmlaq): New built-in function.
1097 (bfmlalb): New built-in function.
1098 (bfmlalt): New built-in function.
1099 (bfmlalb_lane): New built-in function.
1100 (bfmlalt_lane): New built-in function.
1101 * config/aarch64/aarch64-simd.md
1102 (aarch64_bfmmlaqv4sf): New pattern.
1103 (aarch64_bfmlal<bt>v4sf): New pattern.
1104 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1105 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1106 (vbfmlalbq_f32): New intrinsic.
1107 (vbfmlaltq_f32): New intrinsic.
1108 (vbfmlalbq_lane_f32): New intrinsic.
1109 (vbfmlaltq_lane_f32): New intrinsic.
1110 (vbfmlalbq_laneq_f32): New intrinsic.
1111 (vbfmlaltq_laneq_f32): New intrinsic.
1112 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1113 (bt): New int attribute.
1115 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1117 * config/i386/i386.md (*pushtf): Emit "#" instead of
1118 calling gcc_unreachable in insn output.
1121 (*pushsf_rex64): Ditto for alternatives other than 1.
1122 (*pushsf): Ditto for alternatives other than 1.
1124 2020-02-06 Martin Liska <mliska@suse.cz>
1126 PR gcov-profile/91971
1127 PR gcov-profile/93466
1128 * coverage.c (coverage_init): Revert mangling of
1129 path into filename. It can lead to huge filename length.
1130 Creation of subfolders seem more natural.
1132 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1135 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1136 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1137 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1139 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1142 * config/i386/predicates.md (avx_identity_operand): New predicate.
1143 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1144 define_insn_and_split.
1147 * omp-low.c (use_pointer_for_field): For nested constructs, also
1148 look for map clauses on target construct.
1149 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1150 taskreg_nesting_level.
1153 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1154 shared clause, call omp_notice_variable on outer context if any.
1156 2020-02-05 Jason Merrill <jason@redhat.com>
1159 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1160 non-zero address even if weak and not yet defined.
1162 2020-02-05 Martin Sebor <msebor@redhat.com>
1164 PR tree-optimization/92765
1165 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1166 * tree-ssa-strlen.c (compute_string_length): Remove.
1167 (determine_min_objsize): Remove.
1168 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1169 Avoid using type size as the upper bound on string length.
1170 (handle_builtin_string_cmp): Add an argument. Adjust.
1171 (strlen_check_and_optimize_call): Pass additional argument to
1172 handle_builtin_string_cmp.
1174 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1176 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1177 (*pushdi2_rex64 peephole2): Unconditionally split after
1179 (*ashl<mode>3_doubleword): Ditto.
1180 (*<shift_insn><mode>3_doubleword): Ditto.
1182 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1185 * config/rs6000/rs6000.c (get_vector_offset): Fix
1187 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1189 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1191 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1194 (Special Functions for Debugging the Analyzer): Update description
1195 of __analyzer_dump_exploded_nodes.
1197 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1200 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1201 include sets and not clobbers in the vzeroupper pattern.
1202 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1203 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1204 (*avx_vzeroupper_1): New define_insn_and_split.
1207 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1208 don't run when !optimize.
1209 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1212 2020-02-05 Richard Biener <rguenther@suse.de>
1215 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1216 checks before matching calls.
1218 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1220 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1221 function comment typo.
1224 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1225 simd_clone_create failed when i == 0, adjust clone->nargs by
1228 2020-02-05 Martin Liska <mliska@suse.cz>
1231 * doc/invoke.texi: Document that one should
1232 not combine ASLR and -fpch.
1234 2020-02-04 Richard Biener <rguenther@suse.de>
1236 PR tree-optimization/93538
1237 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1239 2020-02-04 Richard Biener <rguenther@suse.de>
1241 PR tree-optimization/91123
1242 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1243 (vn_walk_cb_data::last_vuse): New member.
1244 (vn_walk_cb_data::saved_operands): Likewsie.
1245 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1246 (vn_walk_cb_data::push_partial_def): Use finish.
1247 (vn_reference_lookup_2): Update last_vuse and use finish if
1248 we've saved operands.
1249 (vn_reference_lookup_3): Use finish and update calls to
1250 push_partial_defs everywhere. When translating through
1251 memcpy or aggregate copies save off operands and alias-set.
1252 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1253 operation for redundant store removal.
1255 2020-02-04 Richard Biener <rguenther@suse.de>
1257 PR tree-optimization/92819
1258 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1259 generating more stmts than before.
1261 2020-02-04 Martin Liska <mliska@suse.cz>
1263 * config/arm/arm.c (arm_gen_far_branch): Move the function
1264 outside of selftests.
1266 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1268 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1269 function to adjust PC-relative vector addresses.
1270 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1271 handle vectors with PC-relative addresses.
1273 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1275 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1277 (hard_reg_and_mode_to_addr_mask): Delete.
1278 (rs6000_adjust_vec_address): If the original vector address
1279 was REG+REG or REG+OFFSET and the element is not zero, do the add
1280 of the elements in the original address before adding the offset
1281 for the vector element. Use address_to_insn_form to validate the
1282 address using the register being loaded, rather than guessing
1283 whether the address is a DS-FORM or DQ-FORM address.
1285 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1287 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1288 to calculate the offset in memory from the start of a vector of a
1289 particular element. Add code to keep the element number in
1290 bounds if the element number is variable.
1291 (rs6000_adjust_vec_address): Move calculation of offset of the
1292 vector element to get_vector_offset.
1293 (rs6000_split_vec_extract_var): Do not do the initial AND of
1294 element here, move the code to get_vector_offset.
1296 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1298 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1301 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1303 * config/rs6000/constraints.md: Improve documentation.
1305 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1308 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1309 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1311 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1313 * config.gcc: Remove "carrizo" support.
1314 * config/gcn/gcn-opts.h (processor_type): Likewise.
1315 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1316 * config/gcn/gcn.opt (gpu_type): Likewise.
1317 * config/gcn/t-omp-device: Likewise.
1319 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1322 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1323 * config/arm/arm.c (arm_gen_far_branch): New function
1325 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1327 2020-02-03 Julian Brown <julian@codesourcery.com>
1328 Tobias Burnus <tobias@codesourcery.com>
1330 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1332 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1335 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1336 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1339 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1341 PR rtl-optimization/91333
1342 * ira-color.c (struct allocno_color_data): Add member
1344 (init_allocno_threads): Set the member up.
1345 (bucket_allocno_compare_func): Add compare hard reg
1348 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1350 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1352 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1353 * config.in: Regenerated.
1354 * configure: Regenerated.
1355 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1356 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1357 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1359 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1361 * configure: Regenerate.
1363 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1365 PR rtl-optimization/91333
1366 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1367 reg preferences comparison up.
1369 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1371 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1372 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1373 aarch64-sve-builtins-base.h.
1374 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1375 aarch64-sve-builtins-base.cc.
1376 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1377 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1379 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1380 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1381 (svcvtnt): New functions.
1382 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1383 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1384 (svcvtnt): New functions.
1385 (svcvt): Add a form that converts f32 to bf16.
1386 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1387 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1389 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1390 Treat B as bfloat16_t.
1391 (ternary_bfloat_lane_base): New class.
1392 (ternary_bfloat_def): Likewise.
1393 (ternary_bfloat): New shape.
1394 (ternary_bfloat_lane_def): New class.
1395 (ternary_bfloat_lane): New shape.
1396 (ternary_bfloat_lanex2_def): New class.
1397 (ternary_bfloat_lanex2): New shape.
1398 (ternary_bfloat_opt_n_def): New class.
1399 (ternary_bfloat_opt_n): New shape.
1400 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1401 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1402 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1403 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1404 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1405 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1406 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1407 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1408 the pattern off the narrow mode instead of the wider one.
1409 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1410 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1411 (sve_fp_op): Handle them.
1412 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1413 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1415 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1417 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1418 * config/aarch64/aarch64-modes.def (BF): Move definition before
1419 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1420 (SVE_MODES): Handle BF modes.
1421 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1423 (aarch64_full_sve_mode): Likewise.
1424 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1426 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1427 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1428 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1429 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1431 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1433 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1434 (TYPES_all_data): Add bf16.
1435 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1436 (register_tuple_type): Increase buffer size.
1437 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1438 (bf16): New type suffix.
1439 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1440 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1441 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1442 Change type from all_data to all_arith.
1443 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1446 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1447 Matthew Malcomson <matthew.malcomson@arm.com>
1448 Richard Sandiford <richard.sandiford@arm.com>
1450 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1451 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1452 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1453 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1454 __ARM_FEATURE_MATMUL_FP64.
1455 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1456 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1457 be disabled at the same time.
1458 (f32mm): New extension.
1459 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1460 (AARCH64_FL_F64MM): Bump to the next bit up.
1461 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1462 (TARGET_SVE_F64MM): New macros.
1463 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1464 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1465 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1466 (UNSPEC_ZIP2Q): New unspeccs.
1467 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1468 (optab, sur, perm_insn): Handle the new unspecs.
1469 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1470 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1471 TARGET_SVE_F64MM instead of separate tests.
1472 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1473 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1474 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1475 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1476 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1477 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1478 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1479 (TYPES_s_signed): New macro.
1480 (TYPES_s_integer): Use it.
1481 (TYPES_d_float): New macro.
1482 (TYPES_d_data): Use it.
1483 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1484 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1485 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1486 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1487 (svmmla): New shape.
1488 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1489 template parameters.
1490 (ternary_resize2_lane_base): Likewise.
1491 (ternary_resize2_base): New class.
1492 (ternary_qq_lane_base): Likewise.
1493 (ternary_intq_uintq_lane_def): Likewise.
1494 (ternary_intq_uintq_lane): New shape.
1495 (ternary_intq_uintq_opt_n_def): New class
1496 (ternary_intq_uintq_opt_n): New shape.
1497 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1498 (ternary_uintq_intq_def): New class.
1499 (ternary_uintq_intq): New shape.
1500 (ternary_uintq_intq_lane_def): New class.
1501 (ternary_uintq_intq_lane): New shape.
1502 (ternary_uintq_intq_opt_n_def): New class.
1503 (ternary_uintq_intq_opt_n): New shape.
1504 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1505 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1506 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1507 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1509 (svdotprod_lane_impl): ...this new class.
1510 (svmmla_impl, svusdot_impl): New classes.
1511 (svdot_lane): Update to use svdotprod_lane_impl.
1512 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1513 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1515 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1516 function, with no types defined.
1517 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1518 AARCH64_FL_I8MM functions.
1519 (svmmla): New AARCH64_FL_F32MM function.
1520 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1521 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1522 AARCH64_FL_F64MM function.
1523 (REQUIRED_EXTENSIONS):
1525 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1527 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1530 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1532 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1533 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1534 (*movti_internal): Do not check for
1535 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1536 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1537 just after check for TARGET_AVX.
1538 (*movdf_internal): Ditto.
1539 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1540 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1541 * config/i386/sse.md (mov<mode>_internal): Only check
1542 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1543 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1544 (<sse>_andnot<mode>3<mask_name>): Move check for
1545 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1546 (<code><mode>3<mask_name>): Ditto.
1547 (*andnot<mode>3): Ditto.
1548 (*andnottf3): Ditto.
1549 (*<code><mode>3): Ditto.
1550 (*<code>tf3): Ditto.
1551 (*andnot<VI:mode>3): Remove
1552 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1553 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1554 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1555 (sse4_1_blendv<ssemodesuffix>): Ditto.
1556 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1557 Explain that tune applies to 128bit instructions only.
1559 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1561 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1562 to definition of hsa_kernel_description. Parse assembly to find SGPR
1563 and VGPR count of kernel and store in hsa_kernel_description.
1565 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1567 PR rtl-optimization/91838
1568 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1569 to truncate if allowed or reject combination.
1571 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1573 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1574 (find_inv_vars_cb): Likewise.
1576 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1578 * calls.c (special_function_p): Split out the check for DECL_NAME
1579 being non-NULL and fndecl being extern at file scope into a
1580 new maybe_special_function_p and call it. Drop check for fndecl
1581 being non-NULL that was after a usage of DECL_NAME (fndecl).
1582 * tree.h (maybe_special_function_p): New inline function.
1584 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1586 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1587 (mask_gather_load<mode>): ... here, and zero-initialize the
1589 (maskload<mode>di): Zero-initialize the destination.
1592 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1595 * doc/analyzer.texi (Limitations): Note that constraints on
1596 floating-point values are currently ignored.
1598 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1601 * symtab.c (symtab_node::noninterposable_alias): If localalias
1602 already exists, but is not usable, append numbers after it until
1603 a unique name is found. Formatting fix.
1606 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1609 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1611 * config/gcn/gcn.c (print_operand): Handle LTGT.
1612 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1614 2020-01-30 Richard Biener <rguenther@suse.de>
1616 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1617 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1619 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1621 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1622 without a DECL in .data.rel.ro.local.
1624 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1627 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1631 * config/i386/sse.md
1632 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1633 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1634 any_extend code iterator instead of always zero_extend.
1635 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1636 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1637 Use any_extend code iterator instead of always zero_extend.
1638 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1639 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1640 Use any_extend code iterator instead of always zero_extend.
1641 (*sse2_pmovmskb_ext): New define_insn.
1642 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1645 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1646 (*popcountsi2_zext_falsedep): New define_insn.
1648 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1650 * config.in: Regenerated.
1651 * configure: Regenerated.
1653 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1656 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1657 LLVM's assembler changed the default in version 9.
1659 2020-01-24 Jeff Law <law@redhat.com>
1661 PR tree-optimization/89689
1662 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1664 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1668 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1670 PR rtl-optimization/87763
1671 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1672 simplification to handle subregs as well as bare regs.
1673 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1675 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1678 * ira.c (ira): Revert use of simplified LRA algorithm.
1680 2020-01-29 Martin Jambor <mjambor@suse.cz>
1682 PR tree-optimization/92706
1683 * tree-sra.c (struct access): Fields first_link, last_link,
1684 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1685 next_rhs_queued and grp_rhs_queued respectively, new fields
1686 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1687 (struct assign_link): Field next renamed to next_rhs, new field
1688 next_lhs. Updated comment.
1689 (work_queue_head): Renamed to rhs_work_queue_head.
1690 (lhs_work_queue_head): New variable.
1691 (add_link_to_lhs): New function.
1692 (relink_to_new_repr): Also relink LHS lists.
1693 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1694 (add_access_to_lhs_work_queue): New function.
1695 (pop_access_from_work_queue): Renamed to
1696 pop_access_from_rhs_work_queue.
1697 (pop_access_from_lhs_work_queue): New function.
1698 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1700 (child_would_conflict_in_lacc): Renamed to
1701 child_would_conflict_in_acc. Adjusted parameter names.
1702 (create_artificial_child_access): New parameter set_grp_read, use it.
1703 (subtree_mark_written_and_enqueue): Renamed to
1704 subtree_mark_written_and_rhs_enqueue.
1705 (propagate_subaccesses_across_link): Renamed to
1706 propagate_subaccesses_from_rhs.
1707 (propagate_subaccesses_from_lhs): New function.
1708 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1711 2020-01-29 Martin Jambor <mjambor@suse.cz>
1713 PR tree-optimization/92706
1714 * tree-sra.c (struct access): Adjust comment of
1715 grp_total_scalarization.
1716 (find_access_in_subtree): Look for single children spanning an entire
1718 (scalarizable_type_p): Allow register accesses, adjust callers.
1719 (completely_scalarize): Remove function.
1720 (scalarize_elem): Likewise.
1721 (create_total_scalarization_access): Likewise.
1722 (sort_and_splice_var_accesses): Do not track total scalarization
1724 (analyze_access_subtree): New parameter totally, adjust to new meaning
1725 of grp_total_scalarization.
1726 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1727 (can_totally_scalarize_forest_p): New function.
1728 (create_total_scalarization_access): Likewise.
1729 (create_total_access_and_reshape): Likewise.
1730 (total_should_skip_creating_access): Likewise.
1731 (totally_scalarize_subtree): Likewise.
1732 (analyze_all_variable_accesses): Perform total scalarization after
1733 subaccess propagation using the new functions above.
1734 (initialize_constant_pool_replacements): Output initializers by
1735 traversing the access tree.
1737 2020-01-29 Martin Jambor <mjambor@suse.cz>
1739 * tree-sra.c (verify_sra_access_forest): New function.
1740 (verify_all_sra_access_forests): Likewise.
1741 (create_artificial_child_access): Set parent.
1742 (analyze_all_variable_accesses): Call the verifier.
1744 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1746 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1747 if called on indirect edge.
1748 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1749 speculative call if needed.
1751 2020-01-29 Richard Biener <rguenther@suse.de>
1753 PR tree-optimization/93428
1754 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1755 permutation when the load node is created.
1756 (vect_analyze_slp_instance): Re-use it here.
1758 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1760 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1762 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1764 PR rtl-optimization/93272
1765 * ira-lives.c (process_out_of_region_eh_regs): New function.
1766 (process_bb_node_lives): Call it.
1768 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1770 * coverage.c (read_counts_file): Make error message lowercase.
1772 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1774 * profile-count.c (profile_quality_display_names): Fix ordering.
1776 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1779 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1780 hash only when edge is first within the sequence.
1781 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1782 (symbol_table::create_edge): Do not set target_prob.
1783 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1785 (cgraph_edge::make_speculative): Drop target_prob parameter.
1786 (cgraph_edge::speculative_call_info): Remove.
1787 (cgraph_edge::first_speculative_call_target): New member function.
1788 (update_call_stmt_hash_for_removing_direct_edge): New function.
1789 (cgraph_edge::resolve_speculation): Rewrite to new API.
1790 (cgraph_edge::speculative_call_for_target): New member function.
1791 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1792 multiple speculation targets.
1793 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1795 (verify_speculative_call): Verify that targets form an interval.
1796 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1797 (cgraph_edge::first_speculative_call_target): New member function.
1798 (cgraph_edge::next_speculative_call_target): New member function.
1799 (cgraph_edge::speculative_call_target_ref): New member function.
1800 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1801 (cgraph_edge): Remove target_prob.
1802 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1803 Fix handling of speculative calls.
1804 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1805 * ipa-fnsummary.c (analyze_function_body): Likewise.
1806 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1807 * ipa-profile.c (dump_histogram): Fix formating.
1808 (ipa_profile_generate_summary): Watch for overflows.
1809 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1810 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1811 (update_indirect_edges_after_inlining): Update to new API.
1812 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1814 * profile-count.h: (profile_probability::adjusted): New.
1815 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1816 updating of profile.
1817 * value-prof.c (gimple_ic_transform): Rename to ...
1818 (dump_ic_profile): ... this one; update dumping.
1819 (stream_in_histogram_value): Fix formating.
1820 (gimple_value_profile_transformations): Update.
1822 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1825 * config/i386/i386.md (*movoi_internal_avx): Remove
1826 TARGET_SSE_TYPELESS_STORES check.
1827 (*movti_internal): Prefer TARGET_AVX over
1828 TARGET_SSE_TYPELESS_STORES.
1829 (*movtf_internal): Likewise.
1830 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1831 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1832 from TARGET_SSE_TYPELESS_STORES.
1834 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1836 * diagnostic-core.h (warning_at): Rename overload to...
1837 (warning_meta): ...this.
1838 (emit_diagnostic_valist): Delete decl of overload taking
1839 diagnostic_metadata.
1840 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1841 (warning_at): Rename overload taking diagnostic_metadata to...
1842 (warning_meta): ...this.
1844 2020-01-28 Richard Biener <rguenther@suse.de>
1846 PR tree-optimization/93439
1847 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1848 * tree-cfg.c (move_sese_region_to_fn): ... here.
1849 (verify_types_in_gimple_reference): Verify used cliques are
1852 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1855 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1856 argument of a pointer to struct gcc_options and pass it to
1857 parse_mtune_ctrl_str.
1858 (ix86_function_specific_restore): Pass opts to
1859 set_ix86_tune_features.
1860 (ix86_option_override_internal): Likewise.
1861 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1862 gcc_options and use it for x_ix86_tune_ctrl_string.
1864 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1866 PR rtl-optimization/87763
1867 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1868 simplification to handle subregs as well as bare regs.
1869 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1871 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1873 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1874 for reduction chains that (now) include a call.
1876 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1878 PR tree-optimization/92822
1879 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1880 out the don't-care elements of a vector whose significant elements
1881 are duplicates, make the don't-care elements duplicates too.
1883 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1885 PR tree-optimization/93434
1886 * tree-predcom.c (split_data_refs_to_components): Record which
1887 components have had aliasing loads removed. Prevent store-store
1888 commoning for all such components.
1890 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1893 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1894 -1 or is_vshift is true, use new_vector with number of elts npatterns
1895 rather than new_unary_operation.
1897 PR tree-optimization/93454
1898 * gimple-fold.c (fold_array_ctor_reference): Perform
1899 elt_size.to_uhwi () just once, instead of calling it in every
1900 iteration. Punt if that value is above size of the temporary
1901 buffer. Decrease third native_encode_expr argument when
1902 bufoff + elt_sz is above size of buf.
1904 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1906 * config/mips/mips.c (mips_declare_object_name)
1907 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1909 2020-01-27 Martin Liska <mliska@suse.cz>
1911 PR gcov-profile/93403
1912 * tree-profile.c (gimple_init_gcov_profiler): Generate
1913 both __gcov_indirect_call_profiler_v4 and
1914 __gcov_indirect_call_profiler_v4_atomic.
1916 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1919 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1921 (@aarch64_split_simd_mov<mode>): Use it.
1922 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1923 Leave the vec_extract patterns to handle 2-element vectors.
1924 (aarch64_simd_mov_from_<mode>high): Likewise.
1925 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1926 (vec_extractv2dfv1df): Likewise.
1928 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1930 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1931 jump conditions for *compare_condjump<GPI:mode>.
1933 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1936 * digraph.cc (test_edge::test_edge): Specify template for base
1939 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1941 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1943 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1945 * config/arc/arc-protos.h (gen_mlo): Remove.
1946 (gen_mhi): Likewise.
1947 * config/arc/arc.c (AUX_MULHI): Define.
1948 (arc_must_save_reister): Special handling for r58/59.
1949 (arc_compute_frame_size): Consider mlo/mhi registers.
1950 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1952 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1953 mlo/mhi name selection.
1954 (arc_restore_callee_saves): Don't early restore blink when ISR.
1955 (arc_expand_prologue): Add mlo/mhi saving.
1956 (arc_expand_epilogue): Add mlo/mhi restoring.
1959 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1960 numbering when MUL64 option is used.
1961 (DWARF2_FRAME_REG_OUT): Define.
1962 * config/arc/arc.md (arc600_stall): New pattern.
1963 (VUNSPEC_ARC_ARC600_STALL): Define.
1964 (mulsi64): Use correct mlo/mhi registers.
1965 (mulsi_600): Clean it up.
1966 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1968 (mhi_operand): Likewise.
1970 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1971 Petro Karashchenko <petro.karashchenko@ring.com>
1973 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1974 attributes if needed.
1975 (prepare_move_operands): Generate special unspec instruction for
1977 (arc_isuncached_mem_p): Propagate uncached attribute to each
1979 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1980 (VUNSPEC_ARC_STDI): Likewise.
1981 (ALLI): New mode iterator.
1982 (mALLI): New mode attribute.
1983 (lddi): New instruction pattern.
1985 (stdidi_split): Split instruction for architectures which are not
1986 supporting ll64 option.
1987 (lddidi_split): Likewise.
1989 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1991 PR rtl-optimization/92989
1992 * lra-lives.c (process_bb_lives): Update the live-in set before
1993 processing additional clobbers.
1995 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1997 PR rtl-optimization/93170
1998 * cselib.c (cselib_invalidate_regno_val): New function, split out
2000 (cselib_invalidate_regno): ...here.
2001 (cselib_invalidated_by_call_p): New function.
2002 (cselib_process_insn): Iterate over all the hard-register entries in
2003 REG_VALUES and invalidate any that cross call-clobbered registers.
2005 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2007 * dojump.c (split_comparison): Use HONOR_NANS rather than
2008 HONOR_SNANS when splitting LTGT.
2010 2020-01-27 Martin Liska <mliska@suse.cz>
2013 * opts.c (print_filtered_help): Exclude language-specific
2014 options from --help=common unless enabled in all FEs.
2016 2020-01-27 Martin Liska <mliska@suse.cz>
2018 * opts.c (print_help): Exclude params from
2019 all except --help=param.
2021 2020-01-27 Martin Liska <mliska@suse.cz>
2024 * config/i386/i386-features.c (make_resolver_func):
2025 Align the code with ppc64 target implementation.
2026 Do not generate a unique name for resolver function.
2028 2020-01-27 Richard Biener <rguenther@suse.de>
2030 PR tree-optimization/93397
2031 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2032 converted reduction chain SLP graph adjustment.
2034 2020-01-26 Marek Polacek <polacek@redhat.com>
2037 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2040 2020-01-26 Jason Merrill <jason@redhat.com>
2043 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2046 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2048 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2049 (rx_setmem): Likewise.
2051 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2054 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2055 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2056 drop <di> from constraint of last operand.
2059 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2060 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2061 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2063 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2066 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2069 2020-01-24 Jeff Law <law@redhat.com>
2071 PR tree-optimization/92788
2072 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2075 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2078 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2079 *avx_vperm_broadcast_<mode>,
2080 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2081 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2082 Move before avx2_perm<mode>/avx512f_perm<mode>.
2085 * simplify-rtx.c (simplify_const_unary_operation,
2086 simplify_const_binary_operation): Punt for mode precision above
2087 MAX_BITSIZE_MODE_ANY_INT.
2089 2020-01-24 Andrew Pinski <apinski@marvell.com>
2091 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2094 2020-01-24 Jeff Law <law@redhat.com>
2097 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2098 for REGs. Call output_operand_lossage to get more reasonable
2101 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2103 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2104 gcn_fp_compare_operator.
2105 (vec_cmpu<mode>di): Use gcn_compare_operator.
2106 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2107 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2108 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2109 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2110 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2111 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2112 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2113 gcn_fp_compare_operator.
2114 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2115 gcn_fp_compare_operator.
2116 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2117 gcn_fp_compare_operator.
2118 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2119 gcn_fp_compare_operator.
2121 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2123 * doc/install.texi (Cross-Compiler-Specific Options): Document
2124 `--with-toolexeclibdir' option.
2126 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2128 * target.def (flags_regnum): Also mention effect on delay slot filling.
2129 * doc/tm.texi: Regenerate.
2131 2020-01-23 Jeff Law <law@redhat.com>
2133 PR translation/90162
2134 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2136 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2139 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2142 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2144 PR rtl-optimization/93402
2145 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2148 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2150 * config.in: Regenerated.
2151 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2152 for TARGET_LIBC_GNUSTACK.
2153 * configure: Regenerated.
2154 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2155 found to be 2.31 or greater.
2157 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2159 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2161 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2162 (mips_asm_file_end): New function. Delegate to
2163 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2164 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2166 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2169 * config/i386/i386-modes.def (POImode): New mode.
2170 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2171 * config/i386/i386.md (DPWI): New mode attribute.
2172 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2174 (QPWI): ... this. Use POI instead of OI for TImode.
2175 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2176 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2179 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2182 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2184 (speculation_tracker_rev): New pattern.
2185 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2186 Use speculation_tracker_rev to track the inverse condition.
2188 2020-01-23 Richard Biener <rguenther@suse.de>
2190 PR tree-optimization/93381
2191 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2192 alias-set of the def as argument and record the first one.
2193 (vn_walk_cb_data::first_set): New member.
2194 (vn_reference_lookup_3): Pass the alias-set of the current def
2195 to push_partial_def. Fix alias-set used in the aggregate copy
2197 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2198 * real.c (clear_significand_below): Fix out-of-bound access.
2200 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2203 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2204 New define_insn patterns.
2206 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2208 * doc/sourcebuild.texi (check-function-bodies): Add an
2209 optional target/xfail selector.
2211 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2213 PR rtl-optimization/93124
2214 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2215 bare USE and CLOBBER insns.
2217 2020-01-22 Andrew Pinski <apinski@marvell.com>
2219 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2221 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2224 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2225 diagnostic_manager into "ana" namespace.
2226 * selftest-run-tests.c (selftest::run_tests): Update for move of
2227 selftest::run_analyzer_selftests to
2228 ana::selftest::run_analyzer_selftests.
2230 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2232 * cfgexpand.c (union_stack_vars): Update the size.
2234 2020-01-22 Richard Biener <rguenther@suse.de>
2236 PR tree-optimization/93381
2237 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2238 throughout, handle all conversions the same.
2240 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2243 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2244 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2245 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2246 Call force_reg on high_in2 unconditionally.
2248 2020-01-22 Martin Liska <mliska@suse.cz>
2250 PR tree-optimization/92924
2251 * profile.c (compute_value_histograms): Divide
2254 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2257 * output.h (assemble_name_resolve): Declare.
2258 * varasm.c (assemble_name_resolve): New function.
2259 (assemble_name): Use it.
2260 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2262 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2264 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2265 update_web_docs_git instead of update_web_docs_svn.
2267 2020-01-21 Andrew Pinski <apinski@marvell.com>
2270 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2271 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2272 (*tlsgd_small_<mode>): Likewise.
2273 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2274 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2275 register. Convert that register back to dest using convert_mode.
2277 2020-01-21 Jim Wilson <jimw@sifive.com>
2279 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2282 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2283 Uros Bizjak <ubizjak@gmail.com>
2286 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2288 (legitimize_tls_address): Do GNU2 TLS address computation in
2289 ptr_mode and zero-extend result to Pmode.
2290 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2291 :P with :PTR and Pmode with ptr_mode.
2292 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2293 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2294 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2296 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2299 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2300 the last two operands are CONST_INT_P before using them as such.
2302 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2304 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2305 to get the integer element types.
2307 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2309 * config/aarch64/aarch64-sve-builtins.h
2310 (function_expander::convert_to_pmode): Declare.
2311 * config/aarch64/aarch64-sve-builtins.cc
2312 (function_expander::convert_to_pmode): New function.
2313 (function_expander::get_contiguous_base): Use it.
2314 (function_expander::prepare_gather_address_operands): Likewise.
2315 * config/aarch64/aarch64-sve-builtins-sve2.cc
2316 (svwhilerw_svwhilewr_impl::expand): Likewise.
2318 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2321 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2322 cfun->machine->label_is_assembled.
2323 (aarch64_print_patchable_function_entry): New.
2324 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2325 * config/aarch64/aarch64.h (struct machine_function): New field,
2328 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2331 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2334 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2337 * cgraph.c (cgraph_edge::resolve_speculation,
2338 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2339 call_stmt_site_hash.
2341 2020-01-21 Martin Liska <mliska@suse.cz>
2343 * config/rs6000/rs6000.c (common_mode_defined): Remove
2346 2020-01-21 Richard Biener <rguenther@suse.de>
2348 PR tree-optimization/92328
2349 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2350 type when value-numbering same-sized store by inserting a
2352 (eliminate_dom_walker::eliminate_stmt): When eliminating
2353 a redundant store handle bit-reinterpretation of the same value.
2355 2020-01-21 Andrew Pinski <apinski@marvel.com>
2358 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2360 (prepare_block_for_update): This. Use a worklist instead of
2363 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2365 * gcc/config/arm/arm.c (clear_operation_p):
2366 Initialise last_regno, skip first iteration
2367 based on the first_set value and use ints instead
2368 of the unnecessary HOST_WIDE_INTs.
2370 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2373 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2374 compare_mode other than SFmode or DFmode.
2376 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2379 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2380 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2381 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2383 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2385 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2387 2020-01-20 Andrew Pinski <apinski@marvell.com>
2390 * targhooks.c (default_print_patchable_function_entry): Use
2391 output_asm_insn to emit the nop instruction.
2393 2020-01-20 Fangrui Song <maskray@google.com>
2396 * targhooks.c (default_print_patchable_function_entry): Align to
2399 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2402 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2403 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2404 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2405 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2406 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2407 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2408 Remove the {q} suffix from lea.
2409 (*tls_dynamic_gnu2_call_64): Renamed to ...
2410 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2411 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2412 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2413 Pass Pmode to gen_tls_dynamic_gnu2_64.
2415 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2417 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2419 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2421 * config/aarch64/aarch64-sve-builtins-base.cc
2422 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2424 2020-01-20 Richard Biener <rguenther@suse.de>
2427 * dwarf2out.c (prune_unused_types): Unconditionally mark
2428 called function DIEs.
2430 2020-01-20 Martin Liska <mliska@suse.cz>
2432 PR tree-optimization/93199
2433 * tree-eh.c (struct leh_state): Add
2434 new field outer_non_cleanup.
2435 (cleanup_is_dead_in): Pass leh_state instead
2436 of eh_region. Add a checking that state->outer_non_cleanup
2437 points to outer non-clean up region.
2438 (lower_try_finally): Record outer_non_cleanup
2440 (lower_catch): Likewise.
2441 (lower_eh_filter): Likewise.
2442 (lower_eh_must_not_throw): Likewise.
2443 (lower_cleanup): Likewise.
2445 2020-01-20 Richard Biener <rguenther@suse.de>
2447 PR tree-optimization/93094
2448 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2449 (vect_transform_loop): Likewise.
2450 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2451 loop_vectorized_call to vect_transform_loop.
2452 * tree-vect-loop.c (vect_transform_loop): Pass down
2453 loop_vectorized_call to vect_loop_versioning.
2454 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2455 the earlier discovered loop_vectorized_call.
2457 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2459 * doc/contribute.texi: Update for SVN -> Git transition.
2460 * doc/install.texi: Likewise.
2462 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2465 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2466 speculative targets.
2467 (verify_speculative_call): New function
2468 (cgraph_node::verify_node): Use it.
2469 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2472 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2475 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2476 (cgraph_edge::make_direct): Remove all indirect targets.
2477 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2478 (cgraph_node::verify_node): Verify that only one call_stmt or
2479 lto_stmt_uid is set.
2480 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2482 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2483 (lto_output_ref): Simplify streaming of stmt.
2484 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2486 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2488 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2489 Mark parameter unused.
2491 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2493 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2495 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2497 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2499 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2501 * Makefile.in: Add coroutine-passes.o.
2502 * builtin-types.def (BT_CONST_SIZE): New.
2503 (BT_FN_BOOL_PTR): New.
2504 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2505 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2506 * coroutine-builtins.def: New file.
2507 * coroutine-passes.cc: New file.
2508 * function.h (struct GTY function): Add a bit to indicate that the
2509 function is a coroutine component.
2510 * internal-fn.c (expand_CO_FRAME): New.
2511 (expand_CO_YIELD): New.
2512 (expand_CO_SUSPN): New.
2513 (expand_CO_ACTOR): New.
2514 * internal-fn.def (CO_ACTOR): New.
2518 * passes.def: Add pass_coroutine_lower_builtins,
2519 pass_coroutine_early_expand_ifns.
2520 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2521 (make_pass_coroutine_early_expand_ifns): New.
2522 * doc/invoke.texi: Document the fcoroutines command line
2525 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2527 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2530 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2531 after checking the argument is a REG. Don't use REGNO (reg)
2532 again to set last_regno, reuse regno variable instead.
2534 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2536 * doc/analyzer.texi (Limitations): Add note about NaN.
2538 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2539 Sudakshina Das <sudi.das@arm.com>
2541 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2542 and valid immediate.
2543 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2544 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2545 * config/arm/constraints.md (Pg): New.
2546 * config/arm/predicates.md (long_shift_imm): New.
2547 (arm_reg_or_long_shift_imm): Likewise.
2548 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2549 (thumb2_lsll): Likewise.
2552 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2553 Sudakshina Das <sudi.das@arm.com>
2555 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2556 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2557 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2558 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2559 * config/arm/thumb2.md (thumb2_asrl): New.
2560 (thumb2_lsll): Likewise.
2562 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2564 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2567 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2569 * gdbinit.in (help-gcc-hooks): New command.
2570 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2571 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2574 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2576 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2577 correct target macro.
2579 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2581 * config/aarch64/aarch64-protos.h
2582 (aarch64_sve_ld1ro_operand_p): New.
2583 * config/aarch64/aarch64-sve-builtins-base.cc
2584 (class load_replicate): New.
2585 (class svld1ro_impl): New.
2586 (class svld1rq_impl): Change to inherit from load_replicate.
2587 (svld1ro): New sve intrinsic function base.
2588 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2589 New DEF_SVE_FUNCTION.
2590 * config/aarch64/aarch64-sve-builtins-base.h
2591 (svld1ro): New decl.
2592 * config/aarch64/aarch64-sve-builtins.cc
2593 (function_expander::add_mem_operand): Modify assert to allow
2595 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2597 * config/aarch64/aarch64.c
2598 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2599 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2600 (aarch64_sve_ld1ro_operand_p): New.
2601 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2602 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2603 * config/aarch64/predicates.md
2604 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2606 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2608 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2609 Introduce this ACLE specified predefined macro.
2610 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2611 (fp): Disabling this disables f64mm.
2612 (simd): Disabling this disables f64mm.
2613 (fp16): Disabling this disables f64mm.
2614 (sve): Disabling this disables f64mm.
2615 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2616 (AARCH64_ISA_F64MM): New.
2617 (TARGET_F64MM): New.
2618 * doc/invoke.texi (f64mm): Document new option.
2620 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2622 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2623 (neoversen1_tunings): Likewise.
2625 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2628 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2629 Add assert to ensure prolog has been emitted.
2630 (aarch64_split_atomic_op): Likewise.
2631 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2632 Use epilogue_completed rather than reload_completed.
2633 (aarch64_atomic_exchange<mode>): Likewise.
2634 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2635 (atomic_nand<mode>): Likewise.
2636 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2637 (atomic_fetch_nand<mode>): Likewise.
2638 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2639 (atomic_nand_fetch<mode>): Likewise.
2641 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2644 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2646 (REVERSE_CONDITION): Delete.
2647 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2648 (CCFP_CCFPE): Likewise.
2649 (e): New mode attribute.
2650 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2651 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2652 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2653 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2654 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2655 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2656 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2657 name of generator from gen_ccmpdi to gen_ccmpccdi.
2658 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2659 the previous comparison but aren't able to, use the new ccmp_rev
2662 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2664 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2665 than testing directly for INTEGER_CST.
2666 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2668 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2670 PR tree-optimization/93292
2671 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2672 get_vectype_for_scalar_type returns NULL.
2674 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2676 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2677 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2679 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2681 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2683 * params.opt: (max-predicted-iterations): Set bounds.
2684 * predict.c (real_almost_one, real_br_prob_base,
2685 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2686 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2687 probabilities; do not truncate to reg_br_prob_bases.
2688 (estimate_loops_at_level): Pass max_cyclic_prob.
2689 (estimate_loops): Compute max_cyclic_prob.
2690 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2692 * profile-count.c (profile_probability::to_sreal): New.
2693 * profile-count.h (class sreal): Move up in file.
2694 (profile_probability::to_sreal): Declare.
2696 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2699 (arm_invalid_conversion): New function for target hook.
2700 (arm_invalid_unary_op): New function for target hook.
2701 (arm_invalid_binary_op): New function for target hook.
2703 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2705 * config.gcc: Add arm_bf16.h.
2706 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2707 (arm_simd_builtin_std_type): Add BFmode.
2708 (arm_init_simd_builtin_types): Define element types for vector types.
2709 (arm_init_bf16_types): New function.
2710 (arm_init_builtins): Add arm_init_bf16_types function call.
2711 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2712 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2713 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2714 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2715 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2716 (arm_mangle_type): Add __bf16.
2717 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2718 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2719 arm_bf16_ptr_type_node.
2720 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2721 define_split between ARM registers.
2722 * config/arm/arm_bf16.h: New file.
2723 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2724 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2726 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2727 * config/arm/vfp.md: Add BFmode to movhf patterns.
2729 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2730 Andre Vieira <andre.simoesdiasvieira@arm.com>
2732 * config/arm/arm-cpus.in (mve, mve_float): New features.
2733 (dsp, mve, mve.fp): New options.
2734 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2735 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2736 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2738 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2739 Thomas Preud'homme <thomas.preudhomme@arm.com>
2741 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2743 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2744 error for using -mcmse when targeting Armv8.1-M Mainline.
2746 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2747 Thomas Preud'homme <thomas.preudhomme@arm.com>
2749 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2750 address in r4 when targeting Armv8.1-M Mainline.
2751 (nonsecure_call_value_internal): Likewise.
2752 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2753 a register match_operand again. Emit BLXNS when targeting
2755 (nonsecure_call_value_reg_thumb2): Likewise.
2757 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2758 Thomas Preud'homme <thomas.preudhomme@arm.com>
2760 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2761 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2762 variable as true when floating-point ABI is not hard. Replace
2763 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2764 Generate VLSTM and VLLDM instruction respectively before and
2765 after a function call to cmse_nonsecure_call function.
2766 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2767 (VUNSPEC_VLLDM): Likewise.
2768 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2769 (lazy_load_multiple_insn): Likewise.
2771 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2772 Thomas Preud'homme <thomas.preudhomme@arm.com>
2774 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2775 (arm_emit_vfp_multi_reg_pop): Likewise.
2776 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2777 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2778 restore callee-saved VFP registers.
2780 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2781 Thomas Preud'homme <thomas.preudhomme@arm.com>
2783 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2784 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2785 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2786 callee-saved GPRs as well as clear ip register before doing a nonsecure
2787 call then restore callee-saved GPRs after it when targeting
2789 (arm_reorg): Adapt to function rename.
2791 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2792 Thomas Preud'homme <thomas.preudhomme@arm.com>
2794 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2795 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2796 clear_vfp_multiple pattern based on a new vfp parameter.
2797 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2798 targeting Armv8.1-M Mainline.
2799 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2800 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2801 whether VFP registers are available before looking call_used_regs for a
2803 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2804 of prototype of clear_operation_p.
2805 (clear_vfp_multiple_operation): New predicate.
2806 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2807 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2809 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2810 Thomas Preud'homme <thomas.preudhomme@arm.com>
2812 * config/arm/arm-protos.h (clear_operation_p): Declare.
2813 * config/arm/arm.c (clear_operation_p): New function.
2814 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2815 targeting Armv8.1-M Mainline or successor.
2816 (output_return_instruction): Only output APSR register clearing if
2817 Armv8.1-M Mainline instructions not available.
2818 (thumb_exit): Likewise.
2819 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2820 * config/arm/thumb2.md (clear_apsr): New define_insn.
2821 (clear_multiple): Likewise.
2822 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2824 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2825 Thomas Preud'homme <thomas.preudhomme@arm.com>
2827 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2828 (use_return_insn): Also return false for Armv8.1-M Mainline.
2829 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2830 Mainline instructions are available.
2831 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2832 when targeting Armv8.1-M Mainline Security Extensions.
2833 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2834 Mainline entry function.
2835 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2836 targeting Armv8.1-M Mainline or successor.
2837 (arm_expand_epilogue): Fix indentation of caller-saved register
2838 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2840 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2841 (FP_SYSREGS): Likewise.
2842 (enum vfp_sysregs_encoding): Define enum.
2843 (fp_sysreg_names): Declare.
2844 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2845 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2846 (pop_fpsysreg_insn): Likewise.
2848 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2849 Thomas Preud'homme <thomas.preudhomme@arm.com>
2851 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2852 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2853 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2854 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2855 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2856 (ARMv8_1m_main): New feature group.
2857 (armv8.1-m.main): New architecture.
2858 * config/arm/arm-tables.opt: Regenerate.
2859 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2860 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2861 (arm_options_perform_arch_sanity_checks): Error out when targeting
2862 Armv8.1-M Mainline Security Extensions.
2863 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2865 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2867 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2868 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2869 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2870 aarch64_bfdot_laneq): New.
2871 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2872 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2873 vbfdotq_laneq_f32): New.
2874 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2875 VBFMLA_W, VBF): New.
2876 (isquadop): Add V4BF, V8BF.
2878 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2880 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2881 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2882 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2883 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2884 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2885 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2886 usdot_laneq, sudot_lane,sudot_laneq): New.
2887 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2888 (aarch64_<sur>dot_lane): New.
2889 * config/aarch64/arm_neon.h (vusdot_s32): New.
2891 (vusdot_lane_s32): New.
2892 (vsudot_lane_s32): New.
2893 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2894 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2896 2020-01-16 Martin Liska <mliska@suse.cz>
2898 * value-prof.c (dump_histogram_value): Fix
2899 obvious spacing issue.
2901 2020-01-16 Andrew Pinski <apinski@marvell.com>
2903 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2904 !storage_order_barrier_p.
2906 2020-01-16 Andrew Pinski <apinski@marvell.com>
2908 * sched-int.h (_dep): Add unused bit-field field for the padding.
2909 * sched-deps.c (init_dep_1): Init unused field.
2911 2020-01-16 Andrew Pinski <apinski@marvell.com>
2913 * optabs.h (create_expand_operand): Initialize target field also.
2915 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2917 PR tree-optimization/92429
2918 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2919 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2921 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2924 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2926 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2927 aarch64_sve_int_mode to each mode.
2929 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2931 * doc/analyzer.texi (Overview): Add note about
2932 -fdump-ipa-analyzer.
2934 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2936 PR tree-optimization/93231
2937 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2938 input_type is unsigned. Use tree_to_shwi for shift constant.
2939 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2940 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2941 inputs more efficiently.
2943 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2945 * config/i386/i386.md (*movsf_internal): Do not require
2946 SSE2 ISA for alternatives 14 and 15.
2948 2020-01-15 Richard Biener <rguenther@suse.de>
2951 * tree-eh.c (sink_clobbers): If we already visited the destination
2952 block do not defer insertion.
2953 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2954 the purpose of defered insertion.
2956 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2958 * BASE-VER: Bump to 10.0.1.
2960 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2962 PR tree-optimization/93247
2963 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2964 type of the stmt that we're going to vectorize.
2966 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2968 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2969 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2972 2020-01-15 Martin Liska <mliska@suse.cz>
2974 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2975 2 calls of streamer_read_hwi in a function call.
2977 2020-01-15 Richard Biener <rguenther@suse.de>
2979 * alias.c (record_alias_subset): Avoid redundant work when
2980 subset is already recorded.
2982 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2984 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2985 the analyzer options provide CWE identifiers.
2987 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2989 * tree-diagnostic-path.cc (path_summary::event_range::print):
2990 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2991 using get_pure_location.
2993 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2995 PR tree-optimization/93262
2996 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2997 perform head trimming only if the last argument is constant,
2998 either all ones, or larger or equal to head trim, in the latter
2999 case decrease the last argument by head_trim.
3001 PR tree-optimization/93249
3002 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3003 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3004 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3005 perform head trim unless we can prove there are no '\0' chars
3006 from the source among the first head_trim chars.
3008 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3010 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3012 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3015 * config/i386/sse.md
3016 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3017 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3018 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3019 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3020 just a single alternative instead of two, make operands 1 and 2
3023 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3026 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3029 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3031 * Makefile.in (lang_opt_files): Add analyzer.opt.
3032 (ANALYZER_OBJS): New.
3033 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3034 tristate.o and ANALYZER_OBJS.
3035 (TEXI_GCCINT_FILES): Add analyzer.texi.
3036 * common.opt (-fanalyzer): New driver option.
3037 * config.in: Regenerate.
3038 * configure: Regenerate.
3039 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3040 (gccdepdir): Also create depdir for "analyzer" subdir.
3041 * digraph.cc: New file.
3042 * digraph.h: New file.
3043 * doc/analyzer.texi: New file.
3044 * doc/gccint.texi ("Static Analyzer") New menu item.
3045 (analyzer.texi): Include it.
3046 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3047 ("Warning Options"): Add static analysis warnings to the list.
3048 (-Wno-analyzer-double-fclose): New option.
3049 (-Wno-analyzer-double-free): New option.
3050 (-Wno-analyzer-exposure-through-output-file): New option.
3051 (-Wno-analyzer-file-leak): New option.
3052 (-Wno-analyzer-free-of-non-heap): New option.
3053 (-Wno-analyzer-malloc-leak): New option.
3054 (-Wno-analyzer-possible-null-argument): New option.
3055 (-Wno-analyzer-possible-null-dereference): New option.
3056 (-Wno-analyzer-null-argument): New option.
3057 (-Wno-analyzer-null-dereference): New option.
3058 (-Wno-analyzer-stale-setjmp-buffer): New option.
3059 (-Wno-analyzer-tainted-array-index): New option.
3060 (-Wno-analyzer-use-after-free): New option.
3061 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3062 (-Wno-analyzer-use-of-uninitialized-value): New option.
3063 (-Wanalyzer-too-complex): New option.
3064 (-fanalyzer-call-summaries): New warning.
3065 (-fanalyzer-checker=): New warning.
3066 (-fanalyzer-fine-grained): New warning.
3067 (-fno-analyzer-state-merge): New warning.
3068 (-fno-analyzer-state-purge): New warning.
3069 (-fanalyzer-transitivity): New warning.
3070 (-fanalyzer-verbose-edges): New warning.
3071 (-fanalyzer-verbose-state-changes): New warning.
3072 (-fanalyzer-verbosity=): New warning.
3073 (-fdump-analyzer): New warning.
3074 (-fdump-analyzer-callgraph): New warning.
3075 (-fdump-analyzer-exploded-graph): New warning.
3076 (-fdump-analyzer-exploded-nodes): New warning.
3077 (-fdump-analyzer-exploded-nodes-2): New warning.
3078 (-fdump-analyzer-exploded-nodes-3): New warning.
3079 (-fdump-analyzer-supergraph): New warning.
3080 * doc/sourcebuild.texi (dg-require-dot): New.
3081 (dg-check-dot): New.
3082 * gdbinit.in (break-on-saved-diagnostic): New command.
3083 * graphviz.cc: New file.
3084 * graphviz.h: New file.
3085 * ordered-hash-map-tests.cc: New file.
3086 * ordered-hash-map.h: New file.
3087 * passes.def (pass_analyzer): Add before
3088 pass_ipa_whole_program_visibility.
3089 * selftest-run-tests.c (selftest::run_tests): Call
3090 selftest::ordered_hash_map_tests_cc_tests.
3091 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3093 * shortest-paths.h: New file.
3094 * timevar.def (TV_ANALYZER): New timevar.
3095 (TV_ANALYZER_SUPERGRAPH): Likewise.
3096 (TV_ANALYZER_STATE_PURGE): Likewise.
3097 (TV_ANALYZER_PLAN): Likewise.
3098 (TV_ANALYZER_SCC): Likewise.
3099 (TV_ANALYZER_WORKLIST): Likewise.
3100 (TV_ANALYZER_DUMP): Likewise.
3101 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3102 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3103 * tree-pass.h (make_pass_analyzer): New decl.
3104 * tristate.cc: New file.
3105 * tristate.h: New file.
3107 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3110 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3111 alternatives 9 and 10.
3113 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3115 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3116 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3117 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3118 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3119 (selftest::hash_map_tests_c_tests): Call it.
3120 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3121 New static constant, using the value of = H::empty_zero_p.
3122 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3123 from default_hash_traits <Value>.
3124 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3126 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3127 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3128 calls to mark_empty with !Descriptor::empty_zero_p.
3129 (hash_table::empty_slow): Conditionalize the memset call with a
3130 check that Descriptor::empty_zero_p; otherwise, loop through the
3131 entries calling mark_empty on them.
3132 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3133 (pointer_hash::empty_zero_p): Likewise.
3134 (pair_hash::empty_zero_p): Likewise.
3135 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3137 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3138 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3139 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3140 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3141 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3142 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3143 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3144 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3146 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3149 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3151 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3152 fix typo on return value.
3154 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3157 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3159 (cgraph_edge::make_speculative): Add param for setting speculative_id
3161 (cgraph_edge::speculative_call_info): Update comments and find reference
3162 by speculative_id for multiple indirect targets.
3163 (cgraph_edge::resolve_speculation): Decrease the speculations
3164 for indirect edge, drop it's speculative if not direct target
3165 left. Update comments.
3166 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3167 (cgraph_node::dump): Print num_speculative_call_targets.
3168 (cgraph_node::verify_node): Don't report error if speculative
3169 edge not include statement.
3170 (cgraph_edge::num_speculative_call_targets_p): New function.
3171 * cgraph.h (int common_target_id): Remove.
3172 (int common_target_probability): Remove.
3173 (num_speculative_call_targets): New variable.
3174 (make_speculative): Add param for setting speculative_id.
3175 (cgraph_edge::num_speculative_call_targets_p): New declare.
3176 (target_prob): New variable.
3177 (speculative_id): New variable.
3178 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3179 call summaries for multiple speculative call targets.
3180 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3181 * ipa-profile.c (struct speculative_call_target): New struct.
3182 (class speculative_call_summary): New class.
3183 (class speculative_call_summaries): New class.
3184 (call_sums): New variable.
3185 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3186 (ipa_profile_write_edge_summary): New function.
3187 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3188 (ipa_profile_dump_all_summaries): New function.
3189 (ipa_profile_read_edge_summary): New function.
3190 (ipa_profile_read_summary_section): New function.
3191 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3192 (ipa_profile): Generate num_speculative_call_targets from
3194 * ipa-ref.h (speculative_id): New variable.
3195 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3196 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3197 common_target_probability. Stream out speculative_id and
3198 num_speculative_call_targets.
3199 (input_edge): Likewise.
3200 * predict.c (dump_prediction): Remove edges count assert to be
3202 * symtab.c (symtab_node::create_reference): Init speculative_id.
3203 (symtab_node::clone_references): Clone speculative_id.
3204 (symtab_node::clone_referring): Clone speculative_id.
3205 (symtab_node::clone_reference): Clone speculative_id.
3206 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3207 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3208 if indirect call contains multiple speculative targets.
3209 * value-prof.h (check_ic_target): Remove.
3210 * value-prof.c (gimple_value_profile_transformations):
3211 Use void function gimple_ic_transform.
3212 * value-prof.c (gimple_ic_transform): Handle topn case.
3213 Fix comment typos. Change it to a void function.
3215 2020-01-13 Andrew Pinski <apinski@marvell.com>
3217 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3218 (octeontx2t98): New define.
3219 (octeontx2t96): New define.
3220 (octeontx2t93): New define.
3221 (octeontx2f95): New define.
3222 (octeontx2f95n): New define.
3223 (octeontx2f95mm): New define.
3224 * config/aarch64/aarch64-tune.md: Regenerate.
3225 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3227 2020-01-13 Jason Merrill <jason@redhat.com>
3229 PR c++/33799 - destroy return value if local cleanup throws.
3230 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3232 2020-01-13 Martin Liska <mliska@suse.cz>
3234 * ipa-cp.c (get_max_overall_size): Use newly
3235 renamed param param_ipa_cp_unit_growth.
3236 * params.opt: Remove legacy param name.
3238 2020-01-13 Martin Sebor <msebor@redhat.com>
3240 PR tree-optimization/93213
3241 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3242 stores to be eliminated.
3244 2020-01-13 Martin Liska <mliska@suse.cz>
3246 * opts.c (print_help): Do not print CL_PARAM
3247 and CL_WARNING for CL_OPTIMIZATION.
3249 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3252 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3253 depending on optimization settings.
3255 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3257 PR tree-optimization/90838
3258 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3259 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3260 argument rather than to initialize temporary for targets that
3261 don't use the mode argument at all. Initialize ctzval to avoid
3264 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3266 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3267 * tree-core.h: Document it.
3268 * gimplify.c (gimplify_omp_workshare): Set it.
3269 * omp-low.c (lower_omp_target): Use it.
3270 * tree-pretty-print.c (dump_omp_clause): Print it.
3272 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3273 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3275 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3277 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3278 * common.opt (fdiagnostics-path-format=): New option.
3279 (diagnostic_path_format): New enum.
3280 (fdiagnostics-show-path-depths): New option.
3281 * coretypes.h (diagnostic_event_id_t): New forward decl.
3282 * diagnostic-color.c (color_dict): Add "path".
3283 * diagnostic-event-id.h: New file.
3284 * diagnostic-format-json.cc (json_from_expanded_location): Make
3286 (json_end_diagnostic): Call context->make_json_for_path if it
3287 exists and the diagnostic has a path.
3288 (diagnostic_output_format_init): Clear context->print_path.
3289 * diagnostic-path.h: New file.
3290 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3291 when printing a run of events in a diagnostic_path so that they
3292 all get the same color.
3293 (layout::m_diagnostic_path_p): New field.
3294 (layout::layout): Initialize it.
3295 (layout::print_any_labels): Don't colorize the label text for an
3296 event in a diagnostic_path.
3297 (gcc_rich_location::add_location_if_nearby): Add
3298 "restrict_to_current_line_spans" and "label" params. Pass the
3299 former to layout.maybe_add_location_range; pass the latter
3300 when calling add_range.
3301 * diagnostic.c: Include "diagnostic-path.h".
3302 (diagnostic_initialize): Initialize context->path_format and
3303 context->show_path_depths.
3304 (diagnostic_show_any_path): New function.
3305 (diagnostic_path::interprocedural_p): New function.
3306 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3307 (simple_diagnostic_path::num_events): New function.
3308 (simple_diagnostic_path::get_event): New function.
3309 (simple_diagnostic_path::add_event): New function.
3310 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3311 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3312 (debug): New overload taking a diagnostic_path *.
3313 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3314 * diagnostic.h (enum diagnostic_path_format): New enum.
3315 (json::value): New forward decl.
3316 (diagnostic_context::path_format): New field.
3317 (diagnostic_context::show_path_depths): New field.
3318 (diagnostic_context::print_path): New callback field.
3319 (diagnostic_context::make_json_for_path): New callback field.
3320 (diagnostic_show_any_path): New decl.
3321 (json_from_expanded_location): New decl.
3322 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3323 (-fdiagnostics-show-path-depths): New option.
3324 (-fdiagnostics-color): Add "path" to description of default
3325 GCC_COLORS; describe it.
3326 (-fdiagnostics-format=json): Document how diagnostic paths are
3327 represented in the JSON output format.
3328 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3329 Add optional params "restrict_to_current_line_spans" and "label".
3330 * opts.c (common_handle_option): Handle
3331 OPT_fdiagnostics_path_format_ and
3332 OPT_fdiagnostics_show_path_depths.
3333 * pretty-print.c: Include "diagnostic-event-id.h".
3334 (pp_format): Implement "%@" format code for printing
3335 diagnostic_event_id_t *.
3336 (selftest::test_pp_format): Add tests for "%@".
3337 * selftest-run-tests.c (selftest::run_tests): Call
3338 selftest::tree_diagnostic_path_cc_tests.
3339 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3340 * toplev.c (general_init): Initialize global_dc->path_format and
3341 global_dc->show_path_depths.
3342 * tree-diagnostic-path.cc: New file.
3343 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3344 non-static. Drop "diagnostic" param in favor of storing the
3345 original value of "where" and re-using it.
3346 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3347 maybe_unwind_expanded_macro_loc.
3348 (tree_diagnostics_defaults): Initialize context->print_path and
3349 context->make_json_for_path.
3350 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3352 (default_tree_make_json_for_path): New decl.
3353 (maybe_unwind_expanded_macro_loc): New decl.
3355 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3357 PR tree-optimization/93210
3358 * fold-const.h (native_encode_initializer,
3359 can_native_interpret_type_p): Declare.
3360 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3362 (native_encode_initializer): New function, moved from dwarf2out.c.
3363 Adjust to native_encode_expr compatible arguments, including dry-run
3364 and partial extraction modes. Don't handle STRING_CST.
3365 (can_native_interpret_type_p): No longer static.
3366 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3367 offset / BITS_PER_UNIT fits into int and don't call it if
3368 can_native_interpret_type_p fails. If suboff is NULL and for
3369 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3370 native_encode_initializer.
3371 (fold_const_aggregate_ref_1): Formatting fix.
3372 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3373 (tree_add_const_value_attribute): Adjust caller.
3375 PR tree-optimization/90838
3376 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3377 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3378 CTZ_DEFINED_VALUE_AT_ZERO.
3380 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3383 * lra-constraints.c (match_reload): Permit input operands have the
3384 same mode as output while other input operands have a different
3387 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3389 PR tree-optimization/90838
3390 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3391 (check_ctz_string): Likewise.
3392 (optimize_count_trailing_zeroes): Likewise.
3393 (simplify_count_trailing_zeroes): Likewise.
3394 (pass_forwprop::execute): Try ctz simplification.
3395 * match.pd: Add matching for ctz idioms.
3397 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3399 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3401 (aarch64_invalid_unary_op): New function for target hook.
3402 (aarch64_invalid_binary_op): New function for target hook.
3404 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3406 * config.gcc: Add arm_bf16.h.
3407 * config/aarch64/aarch64-builtins.c
3408 (aarch64_simd_builtin_std_type): Add BFmode.
3409 (aarch64_init_simd_builtin_types): Define element types for vector
3411 (aarch64_init_bf16_types): New function.
3412 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3413 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3415 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3416 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3418 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3419 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3420 * config/aarch64/aarch64.c
3421 (aarch64_classify_vector_mode): Add support for BF types.
3422 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3423 (aarch64_vq_mode): Add support for BF types.
3424 (aarch64_simd_container_mode): Add support for BF types.
3425 (aarch64_mangle_type): Add support for BF scalar type.
3426 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3427 * config/aarch64/arm_bf16.h: New file.
3428 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3429 * config/aarch64/iterators.md: Add BF types to mode attributes.
3430 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3432 2020-01-10 Jason Merrill <jason@redhat.com>
3434 PR c++/93173 - incorrect tree sharing.
3435 * gimplify.c (copy_if_shared): No longer static.
3436 * gimplify.h: Declare it.
3438 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3440 * doc/invoke.texi (-msve-vector-bits=): Document that
3441 -msve-vector-bits=128 now generates VL-specific code for
3442 little-endian targets.
3443 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3444 build_vector_type_for_mode to construct the data vector types.
3445 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3446 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3447 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3448 for 128-bit vectors.
3450 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3452 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3455 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3457 * config/aarch64/aarch64-builtins.c
3458 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3459 rather than checking the number of elements and the element mode.
3461 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3463 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3464 get_related_vectype_for_scalar_type rather than build_vector_type
3465 to create the index type for a conditional reduction.
3467 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3469 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3470 for any type of gather or scatter, including strided accesses.
3472 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3474 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3477 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3479 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3481 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3482 parameter and its use to reset DR_OFFSET's.
3483 (vect_transform_loop): Remove orig_drs_init argument.
3484 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3485 member of dr_vec_info rather than the offset of the associated
3486 data_reference's innermost_loop_behavior.
3487 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3488 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3489 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3490 get_dr_vinfo_offset.
3491 (vectorizable_store): Likewise.
3492 (vectorizable_load): Likewise.
3494 2020-01-10 Richard Biener <rguenther@suse.de>
3496 * gimple-ssa-store-merging
3497 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3499 2020-01-10 Martin Liska <mliska@suse.cz>
3502 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3503 encapsulation that was there before r280040.
3505 2020-01-10 Richard Biener <rguenther@suse.de>
3508 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3509 sequences to avoid walking them again for secondary opportunities.
3510 (pass_lower_eh_dispatch::execute): Instead actually insert
3513 2020-01-10 Richard Biener <rguenther@suse.de>
3516 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3517 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3518 avoid quadraticness.
3520 2020-01-10 Martin Jambor <mjambor@suse.cz>
3522 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3523 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3524 to get param_ipa_sra_max_replacements.
3525 (param_splitting_across_edge): Pass the caller to
3526 pull_accesses_from_callee.
3528 2020-01-10 Martin Jambor <mjambor@suse.cz>
3530 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3531 * ipa-cp.c (max_new_size): Removed.
3532 (orig_overall_size): New variable.
3533 (get_max_overall_size): New function.
3534 (estimate_local_effects): Use it. Adjust dump.
3535 (decide_about_value): Likewise.
3536 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3537 orig_overall_size. Adjust dump.
3538 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3540 2020-01-10 Martin Jambor <mjambor@suse.cz>
3542 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3543 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3544 instead of param_ipa_max_agg_items.
3545 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3546 optimization info for the callee.
3548 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3550 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3551 markers if debug_inline_points is false.
3553 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3555 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3557 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3558 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3559 aarch64-sve-builtins-sve2.h.
3560 (aarch64-sve-builtins-sve2.o): New rule.
3561 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3562 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3563 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3564 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3565 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3566 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3568 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3569 instructions that are handled here.
3570 (@cond_asrd<mode>): Generalize to...
3571 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3572 (*cond_asrd<mode>_2): Generalize to...
3573 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3574 (*cond_asrd<mode>_z): Generalize to...
3575 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3576 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3577 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3578 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3579 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3581 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3582 (@aarch64_scatter_stnt<mode>): Likewise.
3583 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3584 (@aarch64_mul_lane_<mode>): Likewise.
3585 (@aarch64_sve_suqadd<mode>_const): Likewise.
3586 (*<sur>h<addsub><mode>): Generalize to...
3587 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3589 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3590 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3591 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3592 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3593 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3594 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3595 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3596 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3597 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3598 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3599 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3600 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3601 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3602 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3603 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3604 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3605 (@aarch64_sve2_xar<mode>): Likewise.
3606 (@aarch64_sve2_bcax<mode>): Likewise.
3607 (*aarch64_sve2_eor3<mode>): Rename to...
3608 (@aarch64_sve2_eor3<mode>): ...this.
3609 (@aarch64_sve2_bsl<mode>): New expander.
3610 (@aarch64_sve2_nbsl<mode>): Likewise.
3611 (@aarch64_sve2_bsl1n<mode>): Likewise.
3612 (@aarch64_sve2_bsl2n<mode>): Likewise.
3613 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3614 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3615 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3616 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3617 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3618 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3619 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3620 (<su>mull<bt><Vwide>): Generalize to...
3621 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3623 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3624 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3625 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3626 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3627 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3628 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3629 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3630 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3631 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3632 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3633 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3634 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3635 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3636 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3637 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3638 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3639 (<SHRNB:r>shrnb<mode>): Generalize to...
3640 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3642 (<SHRNT:r>shrnt<mode>): Generalize to...
3643 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3645 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3646 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3647 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3648 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3649 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3650 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3651 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3652 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3653 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3654 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3655 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3656 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3657 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3658 (@aarch64_sve2_cvtnt<mode>): Likewise.
3659 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3660 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3661 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3662 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3663 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3664 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3665 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3666 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3667 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3668 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3669 (@aarch64_sve2_pmul<mode>): Likewise.
3670 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3671 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3672 (@aarch64_sve2_tbl2<mode>): Likewise.
3673 (@aarch64_sve2_tbx<mode>): Likewise.
3674 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3675 (@aarch64_sve2_histcnt<mode>): Likewise.
3676 (@aarch64_sve2_histseg<mode>): Likewise.
3677 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3678 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3679 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3680 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3681 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3682 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3683 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3684 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3685 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3686 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3687 (SVE2_PMULL_PAIR_I): New mode iterators.
3688 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3689 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3690 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3691 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3692 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3693 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3694 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3695 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3696 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3697 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3698 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3699 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3700 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3701 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3702 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3703 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3704 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3705 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3706 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3707 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3708 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3709 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3710 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3711 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3712 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3713 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3714 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3715 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3716 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3717 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3718 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3720 (VNARROW, Ventype): New mode attributes.
3721 (Vewtype): Handle VNx2DI. Fix typo in comment.
3722 (VDOUBLE): New mode attribute.
3723 (sve_lane_con): Handle VNx8HI.
3724 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3725 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3726 (sve_int_op, sve_int_op_rev): Handle the above codes.
3727 (sve_pred_int_rhs2_operand): Likewise.
3728 (MULLBT, SHRNB, SHRNT): Delete.
3729 (SVE_INT_SHIFT_IMM): New int iterator.
3730 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3731 and UNSPEC_WHILEHS for TARGET_SVE2.
3732 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3733 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3734 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3735 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3736 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3737 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3738 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3739 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3740 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3741 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3742 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3743 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3744 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3745 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3746 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3747 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3748 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3749 (optab): Handle the new unspecs.
3750 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3752 (lr): Handle the new unspecs.
3754 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3755 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3756 (sve_int_qsub_op): New int attributes.
3757 (sve_fp_op, rot): Handle the new unspecs.
3758 * config/aarch64/aarch64-sve-builtins.h
3759 (function_resolver::require_matching_pointer_type): Declare.
3760 (function_resolver::resolve_unary): Add an optional boolean argument.
3761 (function_resolver::finish_opt_n_resolution): Add an optional
3762 type_suffix_index argument.
3763 (gimple_folder::redirect_call): Declare.
3764 (gimple_expander::prepare_gather_address_operands): Add an optional
3766 * config/aarch64/aarch64-sve-builtins.cc: Include
3767 aarch64-sve-builtins-sve2.h.
3768 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3769 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3770 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3771 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3772 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3773 (TYPES_s_unsigned): Likewise.
3774 (TYPES_s_integer): Use TYPES_s_unsigned.
3775 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3776 (TYPES_sd_integer): Use them.
3777 (TYPES_d_unsigned): New macro.
3778 (TYPES_d_integer): Use it.
3779 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3780 (TYPES_cvt_narrow): Likewise.
3781 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3782 (preds_mx): New variable.
3783 (function_builder::add_overloaded_function): Allow the new feature
3784 set to be more restrictive than the original one.
3785 (function_resolver::infer_pointer_type): Remove qualifiers from
3786 the pointer type before printing it.
3787 (function_resolver::require_matching_pointer_type): New function.
3788 (function_resolver::resolve_sv_displacement): Handle functions
3789 that don't support 32-bit vector indices or svint32_t vector offsets.
3790 (function_resolver::finish_opt_n_resolution): Take the inferred type
3791 as a separate argument.
3792 (function_resolver::resolve_unary): Optionally treat all forms in
3793 the same way as normal merging functions.
3794 (gimple_folder::redirect_call): New function.
3795 (function_expander::prepare_gather_address_operands): Add an argument
3796 that says whether scaled forms are available. If they aren't,
3797 handle scaling of vector indices and don't add the extension and
3799 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3800 fall back to using cond_* instead.
3801 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3802 Split out the member variables into...
3803 (rtx_code_function_base): ...this new base class.
3804 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3805 (unspec_based_function): Split out the member variables into...
3806 (unspec_based_function_base): ...this new base class.
3807 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3808 (unspec_based_function_exact_insn): New class.
3809 (unspec_based_add_function, unspec_based_add_lane_function)
3810 (unspec_based_lane_function, unspec_based_pred_function)
3811 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3812 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3813 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3815 (unspec_based_fused_function): New class.
3816 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3817 (unspec_based_fused_lane_function): New class.
3818 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3820 (CODE_FOR_MODE1): New macro.
3821 (fixed_insn_function): New class.
3822 (while_comparison): Likewise.
3823 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3824 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3825 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3826 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3827 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3828 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3829 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3830 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3831 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3832 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3833 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3834 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3835 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3836 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3837 Also add an initial argument for unary_convert_narrowt, regardless
3838 of the predication type.
3839 (build_32_64): Allow loads and stores to specify MODE_none.
3840 (build_sv_index64, build_sv_uint_offset): New functions.
3841 (long_type_suffix): New function.
3842 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3843 (binary_imm_long_base, load_gather_sv_base): Likewise.
3844 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3845 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3846 (unary_narrowb_base, unary_narrowt_base): Likewise.
3847 (binary_long_lane_def, binary_long_lane): New shape.
3848 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3849 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3850 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3851 (binary_to_uint_def, binary_to_uint): Likewise.
3852 (binary_wide_def, binary_wide): Likewise.
3853 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3854 (compare_def, compare): Likewise.
3855 (compare_ptr_def, compare_ptr): Likewise.
3856 (load_ext_gather_index_restricted_def,
3857 load_ext_gather_index_restricted): Likewise.
3858 (load_ext_gather_offset_restricted_def,
3859 load_ext_gather_offset_restricted): Likewise.
3860 (load_gather_sv_def): Inherit from load_gather_sv_base.
3861 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3862 (shift_left_imm_def, shift_left_imm): Likewise.
3863 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3864 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3865 (store_scatter_index_restricted_def,
3866 store_scatter_index_restricted): Likewise.
3867 (store_scatter_offset_restricted_def,
3868 store_scatter_offset_restricted): Likewise.
3869 (tbl_tuple_def, tbl_tuple): Likewise.
3870 (ternary_long_lane_def, ternary_long_lane): Likewise.
3871 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3872 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3873 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3874 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3875 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3876 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3877 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3878 (ternary_uint_def, ternary_uint): Likewise.
3879 (unary_convert): Fix typo in comment.
3880 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3881 (unary_long_def, unary_long): Likewise.
3882 (unary_narrowb_def, unary_narrowb): Likewise.
3883 (unary_narrowt_def, unary_narrowt): Likewise.
3884 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3885 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3886 (unary_to_int_def, unary_to_int): Likewise.
3887 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3888 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3889 (svasrd_impl): Delete.
3890 (svcadd_impl::expand): Handle integer operations too.
3891 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3892 new functions to derive the unspec numbers.
3893 (svmla_svmls_lane_impl): Replace with...
3894 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3895 integer operations too.
3896 (svwhile_impl): Rename to...
3897 (svwhilelx_impl): ...this and inherit from while_comparison.
3898 (svasrd): Use unspec_based_function.
3899 (svmla_lane): Use svmla_lane_impl.
3900 (svmls_lane): Use svmls_lane_impl.
3901 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3902 (svwhilele, svwhilelt): Use svwhilelx_impl.
3903 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3904 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3905 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3906 * config/aarch64/aarch64-sve-builtins.def: Include
3907 aarch64-sve-builtins-sve2.def.
3909 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3911 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3912 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3913 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3914 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3915 immediates as well as vector ones.
3916 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3917 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3918 (aarch64_sve_qsub_immediate): Update calls accordingly.
3920 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3922 * config/aarch64/aarch64-sve2.md: Add banner comments.
3923 (<su>mulh<r>s<mode>3): Move further up file.
3924 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3925 (*aarch64_sve2_sra<mode>): Move further down file.
3926 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3928 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3930 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3932 (while_optab_cmp): Handle them.
3933 * config/aarch64/aarch64-sve.md
3934 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3935 and add a "@" marker.
3936 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3937 instead of gen_aarch64_sve2_while_ptest.
3938 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3940 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3942 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3943 (UNSPEC_WHILELE): ...this.
3944 (UNSPEC_WHILE_LO): Rename to...
3945 (UNSPEC_WHILELO): ...this.
3946 (UNSPEC_WHILE_LS): Rename to...
3947 (UNSPEC_WHILELS): ...this.
3948 (UNSPEC_WHILE_LT): Rename to...
3949 (UNSPEC_WHILELT): ...this.
3950 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3951 (cmp_op, while_optab_cmp): Likewise.
3952 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3953 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3954 (svwhilelt): Likewise.
3956 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3958 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3959 (unary_to_uint): Define.
3960 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3961 (unary_count): Rename to...
3962 (unary_to_uint_def, unary_to_uint): ...this.
3963 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3965 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3967 * config/aarch64/aarch64-sve-builtins-functions.h
3968 (code_for_mode_function): New class.
3969 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3970 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3971 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3972 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3973 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3975 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3977 * config/aarch64/iterators.md (addsub): New code attribute.
3978 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3980 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3981 in the asm string and attributes. Fix indentation.
3982 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3984 (@aarch64_sve_<optab><mode>): ...this.
3985 * config/aarch64/aarch64-sve-builtins.h
3986 (function_expander::expand_signed_unpred_op): Delete.
3987 * config/aarch64/aarch64-sve-builtins.cc
3988 (function_expander::expand_signed_unpred_op): Likewise.
3989 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3990 try using code_for_aarch64_sve instead.
3991 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3992 (svqsub_impl): Likewise.
3993 (svqadd, svqsub): Use rtx_code_function instead.
3995 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3997 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3998 (HADDSUB, sur, addsub): Remove them.
4000 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4002 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4003 functions the same way, rather than singling out those that
4004 aren't mapped directly to optabs.
4006 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4008 * target.def (compatible_vector_types_p): New target hook.
4009 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4010 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4011 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4012 * doc/tm.texi: Regenerate.
4013 * gimple-expr.c: Include target.h.
4014 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4015 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4017 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4018 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4019 Use the original predicate if it already has a suitable type.
4021 2020-01-09 Martin Jambor <mjambor@suse.cz>
4023 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4024 resolve_speculation and redirect_call_stmt_to_callee static. Change
4025 return type of set_call_stmt to cgraph_edge *.
4026 * auto-profile.c (afdo_indirect_call): Adjust call to
4027 redirect_call_stmt_to_callee.
4028 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4029 make the this pointer explicit, adjust self-recursive calls and the
4030 call top make_direct. Return the resulting edge.
4031 (cgraph_edge::remove): Make this pointer explicit.
4032 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4033 (cgraph_edge::make_direct): Likewise, adjust call to
4034 resolve_speculation.
4035 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4036 call to set_call_stmt.
4037 (cgraph_update_edges_for_call_stmt_node): Update call to
4038 set_call_stmt and remove.
4039 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4040 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4041 (cgraph_node::create_edge_including_clones): Moved "first" definition
4042 of edge to the block where it was used. Adjusted calls to
4044 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4045 cgraph_edge::remove.
4046 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4047 make_direct and redirect_call_stmt_to_callee.
4048 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4049 resolve_speculation and make_direct.
4050 * ipa-inline-transform.c (inline_transform): Adjust call to
4051 redirect_call_stmt_to_callee.
4052 (check_speculations_1):: Adjust call to resolve_speculation.
4053 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4054 resolve-speculation.
4055 (inline_small_functions): Adjust call to resolve_speculation.
4056 (ipa_inline): Likewise.
4057 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4059 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4060 safe with regards to edge removal, adjust calls to
4061 redirect_call_stmt_to_callee.
4062 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4063 and redirect_call_stmt_to_callee.
4064 * multiple_target.c (create_dispatcher_calls): Adjust call to
4065 redirect_call_stmt_to_callee
4066 (redirect_to_specific_clone): Likewise.
4067 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4068 Adjust calls to cgraph_edge::remove.
4069 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4070 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4071 (expand_call_inline): Adjust call to cgraph_edge::remove.
4073 2020-01-09 Martin Liska <mliska@suse.cz>
4075 * params.opt: Set Optimization for
4076 param_max_speculative_devirt_maydefs.
4078 2020-01-09 Martin Sebor <msebor@redhat.com>
4082 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4084 2020-01-09 Martin Liska <mliska@suse.cz>
4086 * auto-profile.c (auto_profile): Use opt_for_fn
4088 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4089 (propagate_vals_across_arith_jfunc): Likewise.
4090 (hint_time_bonus): Likewise.
4091 (incorporate_penalties): Likewise.
4092 (good_cloning_opportunity_p): Likewise.
4093 (perform_estimation_of_a_value): Likewise.
4094 (estimate_local_effects): Likewise.
4095 (ipcp_propagate_stage): Likewise.
4096 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4097 (set_switch_stmt_execution_predicate): Likewise.
4098 (analyze_function_body): Likewise.
4099 * ipa-inline-analysis.c (offline_size): Likewise.
4100 * ipa-inline.c (early_inliner): Likewise.
4101 * ipa-prop.c (ipa_analyze_node): Likewise.
4102 (ipcp_transform_function): Likewise.
4103 * ipa-sra.c (process_scan_results): Likewise.
4104 (ipa_sra_summarize_function): Likewise.
4105 * params.opt: Rename ipcp-unit-growth to
4106 ipa-cp-unit-growth. Add Optimization for various
4107 IPA-related parameters.
4109 2020-01-09 Richard Biener <rguenther@suse.de>
4112 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4114 2020-01-09 Richard Biener <rguenther@suse.de>
4116 PR tree-optimization/93040
4117 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4119 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4121 * common/config/avr/avr-common.c (avr_option_optimization_table)
4122 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4124 2020-01-09 Martin Liska <mliska@suse.cz>
4126 * cgraphclones.c (symbol_table::materialize_all_clones):
4127 Use cgraph_node::dump_name.
4129 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4132 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4133 output_operand_lossage instead of gcc_unreachable.
4134 * doc/md.texi (riscv f constraint): Fix typo.
4137 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4138 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4139 CONST_SCALAR_INT_P instead of CONST_INT_P.
4140 (*subv<mode>4_1): Rename to ...
4141 (subv<mode>4_1): ... this.
4142 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4143 define_insn_and_split patterns.
4144 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4147 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4149 * vec.c (class selftest::count_dtor): New class.
4150 (selftest::test_auto_delete_vec): New test.
4151 (selftest::vec_c_tests): Call it.
4152 * vec.h (class auto_delete_vec): New class template.
4153 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4155 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4157 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4159 2020-01-08 Jim Wilson <jimw@sifive.com>
4161 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4162 use of TLS_MODEL_LOCAL_EXEC when not pic.
4164 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4166 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4169 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4172 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4173 *stack_protect_set_3 peephole2): Also check that the second
4174 insns source is general_operand.
4177 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4178 predicate for output operand instead of register_operand.
4179 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4180 memory destination and non-memory operands[2].
4182 2020-01-08 Martin Liska <mliska@suse.cz>
4184 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4185 ::dump_asm_name instead of (::name or ::asm_name).
4186 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4187 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4188 (analyze_functions): Likewise.
4189 (expand_all_functions): Likewise.
4190 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4191 (propagate_bits_across_jump_function): Likewise.
4192 (dump_profile_updates): Likewise.
4193 (ipcp_store_bits_results): Likewise.
4194 (ipcp_store_vr_results): Likewise.
4195 * ipa-devirt.c (dump_targets): Likewise.
4196 * ipa-fnsummary.c (analyze_function_body): Likewise.
4197 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4198 (process_hsa_functions): Likewise.
4199 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4200 (set_alias_uids): Likewise.
4201 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4202 * ipa-inline.c (recursive_inlining): Likewise.
4203 (inline_to_all_callers_1): Likewise.
4204 (ipa_inline): Likewise.
4205 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4206 (ipa_propagate_frequency): Likewise.
4207 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4208 (remove_described_reference): Likewise.
4209 * ipa-pure-const.c (worse_state): Likewise.
4210 (check_retval_uses): Likewise.
4211 (analyze_function): Likewise.
4212 (propagate_pure_const): Likewise.
4213 (propagate_nothrow): Likewise.
4214 (dump_malloc_lattice): Likewise.
4215 (propagate_malloc): Likewise.
4216 (pass_local_pure_const::execute): Likewise.
4217 * ipa-visibility.c (optimize_weakref): Likewise.
4218 (function_and_variable_visibility): Likewise.
4219 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4220 (ipa_discover_variable_flags): Likewise.
4221 * lto-streamer-out.c (output_function): Likewise.
4222 (output_constructor): Likewise.
4223 * tree-inline.c (copy_bb): Likewise.
4224 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4225 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4227 2020-01-08 Richard Biener <rguenther@suse.de>
4230 * tree-eh.c (sink_clobbers): Update virtual operands for
4231 the first and last stmt only. Add a dry-run capability.
4232 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4233 after CFG manipulations and in RPO order to catch all
4234 secondary opportunities reliably.
4236 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4239 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4241 2019-01-08 Richard Biener <rguenther@suse.de>
4244 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4245 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4246 virtual operand, also updating SSA use.
4247 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4248 Update stmt after resetting virtual operand.
4249 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4250 * gimple-iterator.c (gsi_remove): When not removing the stmt
4251 permanently do not delink immediate uses or mark the stmt modified.
4253 2020-01-08 Martin Liska <mliska@suse.cz>
4255 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4256 (ipa_call_context::estimate_size_and_time): Likewise.
4257 (inline_analyze_function): Likewise.
4259 2020-01-08 Martin Liska <mliska@suse.cz>
4261 * cgraph.c (cgraph_node::dump): Use systematically
4264 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4266 Add -nodevicespecs option for avr.
4269 * config/avr/avr.opt (-nodevicespecs): New driver option.
4270 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4271 "-specs=device-specs/..." if that option is not set.
4272 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4274 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4276 Implement 64-bit double functions for avr.
4279 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4280 --with-double-comparison.
4281 * doc/install.texi: Document them.
4282 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4283 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4284 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4285 * doc/invoke.texi (AVR Built-in Macros): Document them.
4286 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4287 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4288 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4290 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4293 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4294 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4295 when only building rm-profile multilibs.
4297 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4300 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4301 lattice for a value to check.
4302 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4303 finite propagation in self-recursive scc.
4305 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4307 * ipa-inline.c (caller_growth_limits): Restore the AND.
4309 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4311 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4312 (VEC_ALLREG_ALT): New iterator.
4313 (VEC_ALLREG_INT_MODE): New iterator.
4314 (VCMP_MODE): New iterator.
4315 (VCMP_MODE_INT): New iterator.
4316 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4317 (vec_cmp<u>v64qidi): New define_expand.
4318 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4319 (vec_cmpu<mode>di_exec): New define_expand.
4320 (vec_cmp<u>v64qidi_exec): New define_expand.
4321 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4322 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4323 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4324 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4325 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4326 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4327 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4328 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4329 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4330 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4332 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4333 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4335 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4337 * config/gcn/constraints.md (DA): Update description and match.
4339 (Db): New constraint.
4340 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4342 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4343 Implement 'Db' mixed immediate type.
4344 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4345 (addcv64si3_dup<exec_vcc>): Delete.
4346 (subcv64si3<exec_vcc>): Rework constraints.
4347 (addv64di3): Rework constraints.
4348 (addv64di3_exec): Rework constraints.
4349 (subv64di3): Rework constraints.
4350 (addv64di3_dup): Delete.
4351 (addv64di3_dup_exec): Delete.
4352 (addv64di3_zext): Rework constraints.
4353 (addv64di3_zext_exec): Rework constraints.
4354 (addv64di3_zext_dup): Rework constraints.
4355 (addv64di3_zext_dup_exec): Rework constraints.
4356 (addv64di3_zext_dup2): Rework constraints.
4357 (addv64di3_zext_dup2_exec): Rework constraints.
4358 (addv64di3_sext_dup2): Rework constraints.
4359 (addv64di3_sext_dup2_exec): Rework constraints.
4361 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4363 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4364 existing target checks.
4366 2020-01-07 Richard Biener <rguenther@suse.de>
4368 * doc/install.texi: Bump minimal supported MPC version.
4370 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4372 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4373 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4374 * langhooks.c: Include stor-layout.h.
4375 (lhd_simulate_enum_decl): New function.
4376 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4377 handle_arm_sve_h for the LTO frontend.
4378 (register_vector_type): Cope with null returns from pushdecl.
4380 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4382 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4383 (aarch64_sve::nvectors_if_data_type): Replace with...
4384 (aarch64_sve::builtin_type_p): ...this.
4385 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4386 (find_vector_type): Delete.
4387 (add_sve_type_attribute): New function.
4388 (lookup_sve_type_attribute): Likewise.
4389 (register_builtin_types): Add an "SVE type" attribute to each type.
4390 (register_tuple_type): Likewise.
4391 (svbool_type_p, nvectors_if_data_type): Delete.
4392 (mangle_builtin_type): Use lookup_sve_type_attribute.
4393 (builtin_type_p): Likewise. Add an overload that returns the
4394 number of constituent vector and predicate registers.
4395 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4396 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4397 instead of aarch64_sve_argument_p.
4398 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4399 (aarch64_pass_by_reference): Likewise.
4400 (aarch64_function_value_1): Likewise.
4401 (aarch64_return_in_memory): Likewise.
4402 (aarch64_layout_arg): Likewise.
4404 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4406 PR tree-optimization/93156
4407 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4408 least significant bit is always clear.
4410 PR tree-optimization/93118
4411 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4412 simplifier with two intermediate conversions.
4414 2020-01-07 Martin Liska <mliska@suse.cz>
4416 * params.opt: Add Optimization for various parameters.
4418 2020-01-07 Martin Liska <mliska@suse.cz>
4421 * doc/extend.texi: Explain cloning for target_clone
4424 2020-01-07 Martin Liska <mliska@suse.cz>
4426 PR tree-optimization/92860
4427 * common.opt: Make in Optimization option
4428 as it is affected by -O0, which is an Optimization
4430 * tree-inline.c (tree_inlinable_function_p):
4431 Use opt_for_fn for warn_inline.
4432 (expand_call_inline): Likewise.
4434 2020-01-07 Martin Liska <mliska@suse.cz>
4436 PR tree-optimization/92860
4437 * common.opt: Make flag_ree as optimization
4440 2020-01-07 Martin Liska <mliska@suse.cz>
4442 PR optimization/92860
4443 * params.opt: Mark param_min_crossjump_insns with Optimization
4446 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4448 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4449 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4451 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4453 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4454 helper function to return the valid addressing formats for a given
4455 hard register and mode.
4456 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4458 * config/rs6000/constraints.md (Q constraint): Update
4460 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4463 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4464 Use 'Q' for doing vector extract from memory.
4465 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4467 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4468 doing vector extract from memory.
4469 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4470 extract from memory.
4472 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4473 for the offset being 34-bits when -mcpu=future is used.
4475 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4477 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4478 instead of cmpib_comparison_operator in cmpib patterns.
4479 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4480 of cmpib_comparison_operator. Revise comment.
4482 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4484 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4485 in an IFN_DIV_POW2 node to be equal.
4487 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4489 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4490 (vect_check_scalar_mask): ...this.
4491 (vectorizable_store, vectorizable_load): Update call accordingly.
4492 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4493 argument in calls to conditional internal functions.
4495 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4497 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4498 '0' matching inputs.
4499 (subv64di3_exec): Likewise.
4501 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4503 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4504 * doc/md.texi (movstr): Likewise.
4506 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4508 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4511 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4513 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4515 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4516 to a temporary file and use move-if-change to update the real
4517 file where necessary.
4519 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4521 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4522 rather than Upa for CPY /M.
4524 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4526 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4529 2020-01-06 Martin Liska <mliska@suse.cz>
4531 PR tree-optimization/92860
4532 * params.opt: Mark param_max_combine_insns with Optimization
4535 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4538 * config/i386/i386.md (SWIDWI): New mode iterator.
4539 (DWI, dwi): Add TImode variants.
4540 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4541 <general_hilo_operand> instead of <general_operand>. Use
4542 CONST_SCALAR_INT_P instead of CONST_INT_P.
4543 (*addv<mode>4_1): Rename to ...
4544 (addv<mode>4_1): ... this.
4545 (QWI): New mode attribute.
4546 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4547 define_insn_and_split patterns.
4548 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4550 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4551 <general_hilo_operand> instead of <general_operand>.
4552 (*addcarry<mode>_1): New define_insn.
4553 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4555 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4557 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4558 Use "call" instead of "set".
4560 2020-01-03 Martin Jambor <mjambor@suse.cz>
4563 * ipa-cp.c (print_all_lattices): Skip functions without info.
4565 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4568 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4569 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4570 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4571 for 'e' simd clones.
4574 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4576 (mprefer-vector-width=): Add Save.
4577 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4578 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4579 (ix86_debug_options, ix86_function_specific_print): Adjust
4580 ix86_target_string callers.
4581 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4582 (ix86_valid_target_attribute_tree): Likewise.
4583 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4584 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4585 ix86_target_string caller.
4588 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4589 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4590 instead of gen_int_shift_amount + convert_modes.
4592 PR rtl-optimization/93088
4593 * loop-iv.c (find_single_def_src): Punt after looking through
4594 128 reg copies for regs with single definitions. Move definitions
4597 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4599 * config/arm/arm-c.c (arm_cpu_builtins): Define
4600 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4601 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4602 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4603 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4604 * config/arm/arm-tables.opt: Regenerated.
4605 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4606 arm_arch_i8mm and arm_arch_bf16 when enabled.
4607 * config/arm/arm.h (TARGET_I8MM): New macro.
4608 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4609 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4610 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4611 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4612 (v8_6_a_simd_variants): New.
4613 (v8_*_a_simd_variants): Add i8mm and bf16.
4614 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4616 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4619 * predict.c (compute_function_frequency): Don't call
4620 warn_function_cold on functions that already have cold attribute.
4622 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4625 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4626 COMDAT group function labels in .data.rel.ro.local section.
4627 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4630 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4631 comparison_operator in B and S integer comparisons. Likewise, use
4632 ordered_comparison_operator instead of cmpib_comparison_operator in
4634 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4636 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4638 Update copyright years.
4640 * gcc.c (process_command): Update copyright notice dates.
4641 * gcov-dump.c (print_version): Ditto.
4642 * gcov.c (print_version): Ditto.
4643 * gcov-tool.c (print_version): Ditto.
4644 * gengtype.c (create_file): Ditto.
4645 * doc/cpp.texi: Bump @copying's copyright year.
4646 * doc/cppinternals.texi: Ditto.
4647 * doc/gcc.texi: Ditto.
4648 * doc/gccint.texi: Ditto.
4649 * doc/gcov.texi: Ditto.
4650 * doc/install.texi: Ditto.
4651 * doc/invoke.texi: Ditto.
4653 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4655 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4658 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4660 PR tree-optimization/93098
4661 * match.pd (popcount): For shift amounts, use integer_onep
4662 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4663 tests. Make sure that precision is power of two larger than or equal
4664 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4665 instead of ULL suffixed constants. Formatting fixes.
4667 Copyright (C) 2020 Free Software Foundation, Inc.
4669 Copying and distribution of this file, with or without modification,
4670 are permitted in any medium without royalty provided the copyright
4671 notice and this notice are preserved.