1 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
4 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
5 to destination operand for shufps alternative.
6 (*vec_extractv2si_1): Ditto.
8 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
11 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
14 2020-02-20 Martin Liska <mliska@suse.cz>
17 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
19 2020-02-20 Martin Liska <mliska@suse.cz>
22 * common/config/avr/avr-common.c: Remote trailing "|".
24 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
26 * collect2.c (maybe_run_lto_and_relink): Fix typo in
29 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
31 PR tree-optimization/93767
32 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
33 access-size bias from the offset calculations for negative strides.
35 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
37 * collect2.c (c_file, o_file): Make const again.
38 (ldout,lderrout, dump_ld_file): Remove.
39 (tool_cleanup): Avoid calling not signal-safe functions.
40 (maybe_run_lto_and_relink): Avoid possible signal handler
41 access to unintialzed memory (lto_o_files).
42 (main): Avoid leaking temp files in $TMPDIR.
43 Initialize c_file/o_file with concat, which avoids exposing
44 uninitialized memory to signal handler, which calls unlink(!).
45 Avoid calling maybe_unlink when the main function returns,
46 since the atexit handler is already doing this.
47 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
49 2020-02-19 Martin Jambor <mjambor@suse.cz>
51 PR tree-optimization/93776
52 * tree-sra.c (create_access): Do not create zero size accesses.
53 (get_access_for_expr): Do not search for zero sized accesses.
55 2020-02-19 Martin Jambor <mjambor@suse.cz>
57 PR tree-optimization/93667
58 * tree-sra.c (scalarizable_type_p): Return false if record fields
59 do not follow wach other.
61 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
63 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
64 rather than fmv.x.s/fmv.s.x.
66 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
68 * config/aarch64/aarch64-simd-builtins.def
69 (intrinsic_vec_smult_lo_): New.
70 (intrinsic_vec_umult_lo_): Likewise.
71 (vec_widen_smult_hi_): Likewise.
72 (vec_widen_umult_hi_): Likewise.
73 * config/aarch64/aarch64-simd.md
74 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
75 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
76 (vmull_high_s16): Likewise.
77 (vmull_high_s32): Likewise.
78 (vmull_high_u8): Likewise.
79 (vmull_high_u16): Likewise.
80 (vmull_high_u32): Likewise.
82 (vmull_s16): Likewise.
83 (vmull_s32): Likewise.
85 (vmull_u16): Likewise.
86 (vmull_u32): Likewise.
88 2020-02-18 Martin Liska <mliska@suse.cz>
90 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
91 bootstrap by missing removal of invalid sanity check.
93 2020-02-18 Martin Liska <mliska@suse.cz>
96 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
97 Always compare LHS of gimple_assign.
99 2020-02-18 Martin Liska <mliska@suse.cz>
102 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
103 and return type of functions.
104 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
105 Drop MALLOC attribute for void functions.
106 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
107 malloc_state for a new VOID clone.
109 2020-02-18 Martin Liska <mliska@suse.cz>
112 * common.opt: Add -fprofile-reproducibility.
113 * doc/invoke.texi: Document it.
114 * value-prof.c (dump_histogram_value):
115 Document and support behavior for counters[0]
116 being a negative value.
117 (get_nth_most_common_value): Handle negative
118 counters[0] in respect to flag_profile_reproducible.
120 2020-02-18 Jakub Jelinek <jakub@redhat.com>
123 * cgraph.c (verify_speculative_call): Use speculative_id instead of
124 speculative_uid in messages. Remove trailing whitespace from error
125 message. Use num_speculative_call_targets instead of
126 num_speculative_targets in a message.
127 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
128 edge messages and stmt instead of cal_stmt in reference message.
130 PR tree-optimization/93780
131 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
132 before calling build_vector_type.
133 (execute_update_addresses_taken): Likewise.
136 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
137 typo, functoin -> function.
138 * tree.c (free_lang_data_in_decl): Fix comment typo,
139 functoin -> function.
140 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
142 2020-02-17 David Malcolm <dmalcolm@redhat.com>
144 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
146 (print_option_information): Don't call get_option_url if URLs
149 2020-02-17 Alexandre Oliva <oliva@adacore.com>
151 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
152 handling of register_common-less targets.
154 2020-02-17 Martin Liska <mliska@suse.cz>
157 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
159 2020-02-17 Martin Liska <mliska@suse.cz>
162 * config/rs6000/rs6000.c (rs6000_option_override_internal):
165 2020-02-17 Martin Liska <mliska@suse.cz>
168 * config/rx/elf.opt: Fix typo.
170 2020-02-17 Richard Biener <rguenther@suse.de>
173 * opts-global.c (print_ignored_options): Use inform and
176 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
179 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
181 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
184 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
185 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
187 2020-02-15 Jason Merrill <jason@redhat.com>
189 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
191 2020-02-15 Jakub Jelinek <jakub@redhat.com>
193 PR tree-optimization/93744
194 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
195 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
196 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
197 sure @2 in the first and @1 in the other patterns has no side-effects.
199 2020-02-15 David Malcolm <dmalcolm@redhat.com>
200 Bernd Edlinger <bernd.edlinger@hotmail.de>
204 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
205 * configure.ac (--with-diagnostics-urls): New configuration
206 option, based on --with-diagnostics-color.
207 (DIAGNOSTICS_URLS_DEFAULT): New define.
208 * config.h: Regenerate.
209 * configure: Regenerate.
210 * diagnostic.c (diagnostic_urls_init): Handle -1 for
211 DIAGNOSTICS_URLS_DEFAULT from configure-time
212 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
213 and TERM_URLS environment variable.
214 * diagnostic-url.h (diagnostic_url_format): New enum type.
215 (diagnostic_urls_enabled_p): rename to...
216 (determine_url_format): ... this, and change return type.
217 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
218 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
219 the linux console, and mingw.
220 (diagnostic_urls_enabled_p): rename to...
221 (determine_url_format): ... this, and adjust.
222 * pretty-print.h (pretty_printer::show_urls): rename to...
223 (pretty_printer::url_format): ... this, and change to enum.
224 * pretty-print.c (pretty_printer::pretty_printer,
225 pp_begin_url, pp_end_url, test_urls): Adjust.
226 * doc/install.texi (--with-diagnostics-urls): Document the new
227 configuration option.
228 (--with-diagnostics-color): Document the existing interaction
229 with GCC_COLORS better.
230 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
231 vindex reference. Update description of defaults based on the above.
232 (-fdiagnostics-color): Update description of how -fdiagnostics-color
233 interacts with GCC_COLORS.
235 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
238 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
239 conjunction with TARGET_GNU_TLS in early return.
241 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
243 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
244 the mode is not wider than UNITS_PER_WORD.
246 2020-02-14 Martin Jambor <mjambor@suse.cz>
248 PR tree-optimization/93516
249 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
250 access of the same type as the parent.
251 (propagate_subaccesses_from_lhs): Likewise.
253 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
256 * config/i386/avx512vbmi2intrin.h
257 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
258 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
259 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
260 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
261 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
262 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
263 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
264 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
265 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
266 of lacking a closing parenthesis.
267 * config/i386/avx512vbmi2vlintrin.h
268 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
269 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
270 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
271 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
272 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
273 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
274 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
275 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
276 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
277 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
278 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
279 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
280 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
281 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
282 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
283 _mm_shldi_epi32, _mm_mask_shldi_epi32,
284 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
285 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
287 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
290 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
291 the target function entry.
293 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
295 * common/config/arc/arc-common.c (arc_option_optimization_table):
296 Disable if-conversion step when optimized for size.
298 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
300 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
301 R12-R15 are always in ARCOMPACT16_REGS register class.
302 * config/arc/arc.opt (mq-class): Deprecate.
303 * config/arc/constraint.md ("q"): Remove dependency on mq-class
305 * doc/invoke.texi (mq-class): Update text.
306 * common/config/arc/arc-common.c (arc_option_optimization_table):
309 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
311 * config/arc/arc.c (arc_insn_cost): New function.
312 (TARGET_INSN_COST): Define.
313 * config/arc/arc.md (cost): New attribute.
314 (add_n): Use arc_nonmemory_operand.
315 (ashlsi3_insn): Likewise, also update constraints.
316 (ashrsi3_insn): Likewise.
318 (add_shift): Likewise.
319 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
321 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
323 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
325 (umulsidi_600): Likewise.
327 2020-02-13 Jakub Jelinek <jakub@redhat.com>
330 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
331 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
332 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
333 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
334 pass __A to the builtin followed by __W instead of __A followed by
336 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
337 _mm512_mask_popcnt_epi64): Likewise.
338 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
339 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
340 _mm256_mask_popcnt_epi64): Likewise.
342 PR tree-optimization/93582
343 * fold-const.h (shift_bytes_in_array_left,
344 shift_bytes_in_array_right): Declare.
345 * fold-const.c (shift_bytes_in_array_left,
346 shift_bytes_in_array_right): New function, moved from
347 gimple-ssa-store-merging.c, no longer static.
348 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
349 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
350 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
351 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
352 shift_bytes_in_array.
353 (verify_shift_bytes_in_array): Rename to ...
354 (verify_shift_bytes_in_array_left): ... this. Use
355 shift_bytes_in_array_left instead of shift_bytes_in_array.
356 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
357 instead of verify_shift_bytes_in_array.
358 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
359 / native_interpret_expr where the store covers all needed bits,
360 punt on PDP-endian, otherwise allow all involved offsets and sizes
361 not to be byte-aligned.
364 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
365 use const_0_to_255_operand predicate instead of immediate_operand.
366 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
367 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
368 vgf2p8affineinvqb_<mode><mask_name>,
369 vgf2p8affineqb_<mode><mask_name>): Drop mode from
370 const_0_to_255_operand predicated operands.
372 2020-02-12 Jeff Law <law@redhat.com>
374 * config/h8300/h8300.md (comparison shortening peepholes): Use
375 a mode iterator to merge the HImode and SImode peepholes.
377 2020-02-12 Jakub Jelinek <jakub@redhat.com>
380 * real.c (is_even): Make static. Function comment fix.
381 (is_halfway_below): Make static, don't assert R is not inf/nan,
382 instead return false for those. Small formatting fixes.
384 2020-02-12 Martin Sebor <msebor@redhat.com>
387 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
388 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
389 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
390 (strlen_check_and_optimize_call): Adjust callee name.
392 2020-02-12 Jeff Law <law@redhat.com>
394 * config/h8300/h8300.md (comparison shortening peepholes): Drop
395 (and (xor)) variant. Combine other two into single peephole.
397 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
399 PR rtl-optimization/93565
400 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
402 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
404 * config/aarch64/aarch64-simd.md
405 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
406 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
407 generating separate ADDV and zero_extend patterns.
408 * config/aarch64/iterators.md (VDQV_E): New iterator.
410 2020-02-12 Jeff Law <law@redhat.com>
412 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
413 expanders, splits, etc.
414 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
415 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
416 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
417 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
418 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
419 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
421 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
423 2020-02-12 Jakub Jelinek <jakub@redhat.com>
426 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
427 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
428 TARGET_AVX512DQ from condition.
429 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
430 instead of <mask_mode512bit_condition> in condition. If
431 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
433 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
436 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
439 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
441 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
443 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
444 where strlen is more legible.
445 (rs6000_builtin_vectorized_libmass): Ditto.
446 (rs6000_print_options_internal): Ditto.
448 2020-02-11 Martin Sebor <msebor@redhat.com>
450 PR tree-optimization/93683
451 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
453 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
455 * config/rs6000/predicates.md (cint34_operand): Rename the
456 -mprefixed-addr option to be -mprefixed.
457 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
458 the -mprefixed-addr option to be -mprefixed.
459 (OTHER_FUTURE_MASKS): Likewise.
460 (POWERPC_MASKS): Likewise.
461 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
462 the -mprefixed-addr option to be -mprefixed. Change error
463 messages to refer to -mprefixed.
464 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
466 (rs6000_legitimate_offset_address_p): Likewise.
467 (rs6000_mode_dependent_address): Likewise.
468 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
469 "-mprefixed" for target attributes and pragmas.
470 (address_to_insn_form): Rename the -mprefixed-addr option to be
472 (rs6000_adjust_insn_length): Likewise.
473 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
474 -mprefixed-addr option to be -mprefixed.
475 (ASM_OUTPUT_OPCODE): Likewise.
476 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
477 -mprefixed-addr option to be -mprefixed.
478 * config/rs6000/rs6000.opt (-mprefixed): Rename the
479 -mprefixed-addr option to be prefixed. Change the option from
480 being undocumented to being documented.
481 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
482 -mprefixed option. Update the -mpcrel documentation to mention
485 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
487 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
488 including FIRST_PSEUDO_REGISTER - 1.
489 * ira-color.c (print_hard_reg_set): Ditto.
491 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
493 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
494 (USTERNOP_QUALIFIERS): New define.
495 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
496 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
497 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
498 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
499 * config/arm/arm_neon.h (vusdot_s32): New.
500 (vusdot_lane_s32): New.
501 (vusdotq_lane_s32): New.
502 (vsudot_lane_s32): New.
503 (vsudotq_lane_s32): New.
504 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
505 * config/arm/iterators.md (DOTPROD_I8MM): New.
506 (sup, opsuffix): Add <us/su>.
507 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
508 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
510 2020-02-11 Richard Biener <rguenther@suse.de>
512 PR tree-optimization/93661
513 PR tree-optimization/93662
514 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
516 * tree-sra.c (get_access_for_expr): Likewise.
518 2020-02-10 Jakub Jelinek <jakub@redhat.com>
521 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
522 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
523 Change condition from TARGET_AVX2 to TARGET_AVX.
525 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
528 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
531 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
533 Try to generate zero-based comparisons.
534 * config/cris/cris.c (cris_reduce_compare): New function.
535 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
536 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
537 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
539 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
542 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
543 in Thumb state and also as a destination in Arm state. Add T16
546 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
548 * md.texi (Define Subst): Match closing paren in example.
550 2020-02-10 Jakub Jelinek <jakub@redhat.com>
554 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
555 arguments of strncmp.
557 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
560 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
561 but different source value.
562 (adjust_callers_for_value_intersection): New function.
563 (gather_edges_for_value): Adjust order of callers to let a
564 non-self-recursive caller be the first element.
565 (self_recursive_pass_through_p): Add a new parameter "simple", and
566 check generalized self-recursive pass-through jump function.
567 (self_recursive_agg_pass_through_p): Likewise.
568 (find_more_scalar_values_for_callers_subset): Compute value from
569 pass-through jump function for self-recursive.
570 (intersect_with_plats): Cleanup previous implementation code for value
571 itersection with self-recursive call edge.
572 (intersect_with_agg_replacements): Likewise.
573 (intersect_aggregates_with_edge): Deduce value from pass-through jump
574 function for self-recursive call edge. Cleanup previous implementation
575 code for value intersection with self-recursive call edge.
576 (decide_whether_version_node): Remove dead callers and adjust order
577 to let a non-self-recursive caller be the first element.
579 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
581 * recog.c: Move pass_split_before_sched2 code in front of
582 pass_split_before_regstack.
583 (pass_data_split_before_sched2): Rename pass to split3 from split4.
584 (pass_data_split_before_regstack): Rename pass to split4 from split3.
585 (rest_of_handle_split_before_sched2): Remove.
586 (pass_split_before_sched2::execute): Unconditionally call
588 (enable_split_before_sched2): New function.
589 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
590 (pass_split_before_regstack::gate): Ditto.
591 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
592 Update name check for renamed split4 pass.
593 * config/sh/sh.c (register_sh_passes): Update pass insertion
594 point for renamed split4 pass.
596 2020-02-09 Jakub Jelinek <jakub@redhat.com>
598 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
599 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
600 copying them around between host and target.
602 2020-02-08 Andrew Pinski <apinski@marvell.com>
605 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
606 STRICT_ALIGNMENT also.
608 2020-02-08 Jim Wilson <jimw@sifive.com>
611 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
613 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
614 Jakub Jelinek <jakub@redhat.com>
617 * config/i386/i386.h (CALL_USED_REGISTERS): Make
618 xmm16-xmm31 call-used even in 64-bit ms-abi.
620 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
622 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
623 (simd_ummla, simd_usmmla): Likewise.
624 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
625 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
628 2020-02-07 Richard Biener <rguenther@suse.de>
631 * tree-inline.c (fold_marked_statements): Do a PRE walk,
632 skipping unreachable regions.
633 (optimize_inline_calls): Skip folding stmts when we didn't
636 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
639 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
640 Don't return aggregates with only SFmode and DFmode in SSE
642 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
644 2020-02-07 Jakub Jelinek <jakub@redhat.com>
647 * config/rs6000/rs6000-logue.c
648 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
649 if it fails, move rs into end_addr and retry. Add
650 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
651 the insn pattern doesn't describe well what exactly happens to
655 * config/i386/predicates.md (avx_identity_operand): Remove.
656 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
657 (avx_<castmode><avxsizesuffix>_<castmode>,
658 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
659 a VEC_CONCAT of the operand and UNSPEC_CAST.
660 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
661 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
665 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
666 recog_data.insn if distance_non_agu_define changed it.
668 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
671 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
672 we only had X-FORM (reg+reg) addressing for vectors. Also before
673 ISA 3.0, we only had X-FORM addressing for scalars in the
674 traditional Altivec registers.
676 2020-02-06 <zhongyunde@huawei.com>
677 Vladimir Makarov <vmakarov@redhat.com>
679 PR rtl-optimization/93561
680 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
683 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
685 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
688 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
690 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
691 where the low and the high 32 bits are equal to each other specially,
692 with an rldimi instruction.
694 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
696 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
698 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
700 * config/arm/arm-tables.opt: Regenerate.
702 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
705 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
706 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
707 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
709 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
711 PR rtl-optimization/87763
712 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
714 2020-02-06 Delia Burduv <delia.burduv@arm.com>
716 * config/aarch64/aarch64-simd-builtins.def
717 (bfmlaq): New built-in function.
718 (bfmlalb): New built-in function.
719 (bfmlalt): New built-in function.
720 (bfmlalb_lane): New built-in function.
721 (bfmlalt_lane): New built-in function.
722 * config/aarch64/aarch64-simd.md
723 (aarch64_bfmmlaqv4sf): New pattern.
724 (aarch64_bfmlal<bt>v4sf): New pattern.
725 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
726 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
727 (vbfmlalbq_f32): New intrinsic.
728 (vbfmlaltq_f32): New intrinsic.
729 (vbfmlalbq_lane_f32): New intrinsic.
730 (vbfmlaltq_lane_f32): New intrinsic.
731 (vbfmlalbq_laneq_f32): New intrinsic.
732 (vbfmlaltq_laneq_f32): New intrinsic.
733 * config/aarch64/iterators.md (BF_MLA): New int iterator.
734 (bt): New int attribute.
736 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
738 * config/i386/i386.md (*pushtf): Emit "#" instead of
739 calling gcc_unreachable in insn output.
742 (*pushsf_rex64): Ditto for alternatives other than 1.
743 (*pushsf): Ditto for alternatives other than 1.
745 2020-02-06 Martin Liska <mliska@suse.cz>
747 PR gcov-profile/91971
748 PR gcov-profile/93466
749 * coverage.c (coverage_init): Revert mangling of
750 path into filename. It can lead to huge filename length.
751 Creation of subfolders seem more natural.
753 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
756 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
757 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
758 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
760 2020-02-06 Jakub Jelinek <jakub@redhat.com>
763 * config/i386/predicates.md (avx_identity_operand): New predicate.
764 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
765 define_insn_and_split.
768 * omp-low.c (use_pointer_for_field): For nested constructs, also
769 look for map clauses on target construct.
770 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
771 taskreg_nesting_level.
774 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
775 shared clause, call omp_notice_variable on outer context if any.
777 2020-02-05 Jason Merrill <jason@redhat.com>
780 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
781 non-zero address even if weak and not yet defined.
783 2020-02-05 Martin Sebor <msebor@redhat.com>
785 PR tree-optimization/92765
786 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
787 * tree-ssa-strlen.c (compute_string_length): Remove.
788 (determine_min_objsize): Remove.
789 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
790 Avoid using type size as the upper bound on string length.
791 (handle_builtin_string_cmp): Add an argument. Adjust.
792 (strlen_check_and_optimize_call): Pass additional argument to
793 handle_builtin_string_cmp.
795 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
797 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
798 (*pushdi2_rex64 peephole2): Unconditionally split after
800 (*ashl<mode>3_doubleword): Ditto.
801 (*<shift_insn><mode>3_doubleword): Ditto.
803 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
806 * config/rs6000/rs6000.c (get_vector_offset): Fix
808 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
810 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
812 2020-02-05 David Malcolm <dmalcolm@redhat.com>
815 (Special Functions for Debugging the Analyzer): Update description
816 of __analyzer_dump_exploded_nodes.
818 2020-02-05 Jakub Jelinek <jakub@redhat.com>
821 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
822 include sets and not clobbers in the vzeroupper pattern.
823 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
824 the parallel has 17 (64-bit) or 9 (32-bit) elts.
825 (*avx_vzeroupper_1): New define_insn_and_split.
828 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
829 don't run when !optimize.
830 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
833 2020-02-05 Richard Biener <rguenther@suse.de>
836 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
837 checks before matching calls.
839 2020-02-05 Jakub Jelinek <jakub@redhat.com>
841 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
842 function comment typo.
845 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
846 simd_clone_create failed when i == 0, adjust clone->nargs by
849 2020-02-05 Martin Liska <mliska@suse.cz>
852 * doc/invoke.texi: Document that one should
853 not combine ASLR and -fpch.
855 2020-02-04 Richard Biener <rguenther@suse.de>
857 PR tree-optimization/93538
858 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
860 2020-02-04 Richard Biener <rguenther@suse.de>
862 PR tree-optimization/91123
863 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
864 (vn_walk_cb_data::last_vuse): New member.
865 (vn_walk_cb_data::saved_operands): Likewsie.
866 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
867 (vn_walk_cb_data::push_partial_def): Use finish.
868 (vn_reference_lookup_2): Update last_vuse and use finish if
869 we've saved operands.
870 (vn_reference_lookup_3): Use finish and update calls to
871 push_partial_defs everywhere. When translating through
872 memcpy or aggregate copies save off operands and alias-set.
873 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
874 operation for redundant store removal.
876 2020-02-04 Richard Biener <rguenther@suse.de>
878 PR tree-optimization/92819
879 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
880 generating more stmts than before.
882 2020-02-04 Martin Liska <mliska@suse.cz>
884 * config/arm/arm.c (arm_gen_far_branch): Move the function
885 outside of selftests.
887 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
889 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
890 function to adjust PC-relative vector addresses.
891 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
892 handle vectors with PC-relative addresses.
894 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
896 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
898 (hard_reg_and_mode_to_addr_mask): Delete.
899 (rs6000_adjust_vec_address): If the original vector address
900 was REG+REG or REG+OFFSET and the element is not zero, do the add
901 of the elements in the original address before adding the offset
902 for the vector element. Use address_to_insn_form to validate the
903 address using the register being loaded, rather than guessing
904 whether the address is a DS-FORM or DQ-FORM address.
906 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
908 * config/rs6000/rs6000.c (get_vector_offset): New helper function
909 to calculate the offset in memory from the start of a vector of a
910 particular element. Add code to keep the element number in
911 bounds if the element number is variable.
912 (rs6000_adjust_vec_address): Move calculation of offset of the
913 vector element to get_vector_offset.
914 (rs6000_split_vec_extract_var): Do not do the initial AND of
915 element here, move the code to get_vector_offset.
917 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
919 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
922 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
924 * config/rs6000/constraints.md: Improve documentation.
926 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
929 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
930 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
932 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
934 * config.gcc: Remove "carrizo" support.
935 * config/gcn/gcn-opts.h (processor_type): Likewise.
936 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
937 * config/gcn/gcn.opt (gpu_type): Likewise.
938 * config/gcn/t-omp-device: Likewise.
940 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
943 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
944 * config/arm/arm.c (arm_gen_far_branch): New function
946 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
948 2020-02-03 Julian Brown <julian@codesourcery.com>
949 Tobias Burnus <tobias@codesourcery.com>
951 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
953 2020-02-03 Jakub Jelinek <jakub@redhat.com>
956 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
957 valid RTL to sum up the lowest and second lowest bytes of the popcnt
960 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
962 PR rtl-optimization/91333
963 * ira-color.c (struct allocno_color_data): Add member
965 (init_allocno_threads): Set the member up.
966 (bucket_allocno_compare_func): Add compare hard reg
969 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
971 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
973 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
974 * config.in: Regenerated.
975 * configure: Regenerated.
976 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
977 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
978 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
980 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
982 * configure: Regenerate.
984 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
986 PR rtl-optimization/91333
987 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
988 reg preferences comparison up.
990 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
992 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
993 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
994 aarch64-sve-builtins-base.h.
995 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
996 aarch64-sve-builtins-base.cc.
997 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
998 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1000 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1001 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1002 (svcvtnt): New functions.
1003 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1004 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1005 (svcvtnt): New functions.
1006 (svcvt): Add a form that converts f32 to bf16.
1007 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1008 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1010 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1011 Treat B as bfloat16_t.
1012 (ternary_bfloat_lane_base): New class.
1013 (ternary_bfloat_def): Likewise.
1014 (ternary_bfloat): New shape.
1015 (ternary_bfloat_lane_def): New class.
1016 (ternary_bfloat_lane): New shape.
1017 (ternary_bfloat_lanex2_def): New class.
1018 (ternary_bfloat_lanex2): New shape.
1019 (ternary_bfloat_opt_n_def): New class.
1020 (ternary_bfloat_opt_n): New shape.
1021 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1022 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1023 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1024 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1025 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1026 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1027 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1028 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1029 the pattern off the narrow mode instead of the wider one.
1030 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1031 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1032 (sve_fp_op): Handle them.
1033 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1034 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1036 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1038 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1039 * config/aarch64/aarch64-modes.def (BF): Move definition before
1040 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1041 (SVE_MODES): Handle BF modes.
1042 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1044 (aarch64_full_sve_mode): Likewise.
1045 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1047 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1048 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1049 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1050 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1052 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1054 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1055 (TYPES_all_data): Add bf16.
1056 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1057 (register_tuple_type): Increase buffer size.
1058 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1059 (bf16): New type suffix.
1060 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1061 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1062 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1063 Change type from all_data to all_arith.
1064 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1067 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1068 Matthew Malcomson <matthew.malcomson@arm.com>
1069 Richard Sandiford <richard.sandiford@arm.com>
1071 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1072 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1073 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1074 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1075 __ARM_FEATURE_MATMUL_FP64.
1076 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1077 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1078 be disabled at the same time.
1079 (f32mm): New extension.
1080 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1081 (AARCH64_FL_F64MM): Bump to the next bit up.
1082 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1083 (TARGET_SVE_F64MM): New macros.
1084 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1085 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1086 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1087 (UNSPEC_ZIP2Q): New unspeccs.
1088 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1089 (optab, sur, perm_insn): Handle the new unspecs.
1090 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1091 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1092 TARGET_SVE_F64MM instead of separate tests.
1093 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1094 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1095 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1096 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1097 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1098 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1099 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1100 (TYPES_s_signed): New macro.
1101 (TYPES_s_integer): Use it.
1102 (TYPES_d_float): New macro.
1103 (TYPES_d_data): Use it.
1104 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1105 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1106 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1107 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1108 (svmmla): New shape.
1109 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1110 template parameters.
1111 (ternary_resize2_lane_base): Likewise.
1112 (ternary_resize2_base): New class.
1113 (ternary_qq_lane_base): Likewise.
1114 (ternary_intq_uintq_lane_def): Likewise.
1115 (ternary_intq_uintq_lane): New shape.
1116 (ternary_intq_uintq_opt_n_def): New class
1117 (ternary_intq_uintq_opt_n): New shape.
1118 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1119 (ternary_uintq_intq_def): New class.
1120 (ternary_uintq_intq): New shape.
1121 (ternary_uintq_intq_lane_def): New class.
1122 (ternary_uintq_intq_lane): New shape.
1123 (ternary_uintq_intq_opt_n_def): New class.
1124 (ternary_uintq_intq_opt_n): New shape.
1125 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1126 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1127 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1128 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1130 (svdotprod_lane_impl): ...this new class.
1131 (svmmla_impl, svusdot_impl): New classes.
1132 (svdot_lane): Update to use svdotprod_lane_impl.
1133 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1134 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1136 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1137 function, with no types defined.
1138 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1139 AARCH64_FL_I8MM functions.
1140 (svmmla): New AARCH64_FL_F32MM function.
1141 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1142 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1143 AARCH64_FL_F64MM function.
1144 (REQUIRED_EXTENSIONS):
1146 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1148 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1151 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1153 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1154 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1155 (*movti_internal): Do not check for
1156 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1157 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1158 just after check for TARGET_AVX.
1159 (*movdf_internal): Ditto.
1160 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1161 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1162 * config/i386/sse.md (mov<mode>_internal): Only check
1163 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1164 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1165 (<sse>_andnot<mode>3<mask_name>): Move check for
1166 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1167 (<code><mode>3<mask_name>): Ditto.
1168 (*andnot<mode>3): Ditto.
1169 (*andnottf3): Ditto.
1170 (*<code><mode>3): Ditto.
1171 (*<code>tf3): Ditto.
1172 (*andnot<VI:mode>3): Remove
1173 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1174 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1175 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1176 (sse4_1_blendv<ssemodesuffix>): Ditto.
1177 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1178 Explain that tune applies to 128bit instructions only.
1180 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1182 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1183 to definition of hsa_kernel_description. Parse assembly to find SGPR
1184 and VGPR count of kernel and store in hsa_kernel_description.
1186 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1188 PR rtl-optimization/91838
1189 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1190 to truncate if allowed or reject combination.
1192 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1194 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1195 (find_inv_vars_cb): Likewise.
1197 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1199 * calls.c (special_function_p): Split out the check for DECL_NAME
1200 being non-NULL and fndecl being extern at file scope into a
1201 new maybe_special_function_p and call it. Drop check for fndecl
1202 being non-NULL that was after a usage of DECL_NAME (fndecl).
1203 * tree.h (maybe_special_function_p): New inline function.
1205 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1207 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1208 (mask_gather_load<mode>): ... here, and zero-initialize the
1210 (maskload<mode>di): Zero-initialize the destination.
1213 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1216 * doc/analyzer.texi (Limitations): Note that constraints on
1217 floating-point values are currently ignored.
1219 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1222 * symtab.c (symtab_node::noninterposable_alias): If localalias
1223 already exists, but is not usable, append numbers after it until
1224 a unique name is found. Formatting fix.
1227 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1230 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1232 * config/gcn/gcn.c (print_operand): Handle LTGT.
1233 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1235 2020-01-30 Richard Biener <rguenther@suse.de>
1237 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1238 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1240 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1242 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1243 without a DECL in .data.rel.ro.local.
1245 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1248 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1252 * config/i386/sse.md
1253 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1254 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1255 any_extend code iterator instead of always zero_extend.
1256 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1257 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1258 Use any_extend code iterator instead of always zero_extend.
1259 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1260 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1261 Use any_extend code iterator instead of always zero_extend.
1262 (*sse2_pmovmskb_ext): New define_insn.
1263 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1266 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1267 (*popcountsi2_zext_falsedep): New define_insn.
1269 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1271 * config.in: Regenerated.
1272 * configure: Regenerated.
1274 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1277 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1278 LLVM's assembler changed the default in version 9.
1280 2020-01-24 Jeff Law <law@redhat.com>
1282 PR tree-optimization/89689
1283 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1285 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1289 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1291 PR rtl-optimization/87763
1292 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1293 simplification to handle subregs as well as bare regs.
1294 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1296 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1299 * ira.c (ira): Revert use of simplified LRA algorithm.
1301 2020-01-29 Martin Jambor <mjambor@suse.cz>
1303 PR tree-optimization/92706
1304 * tree-sra.c (struct access): Fields first_link, last_link,
1305 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1306 next_rhs_queued and grp_rhs_queued respectively, new fields
1307 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1308 (struct assign_link): Field next renamed to next_rhs, new field
1309 next_lhs. Updated comment.
1310 (work_queue_head): Renamed to rhs_work_queue_head.
1311 (lhs_work_queue_head): New variable.
1312 (add_link_to_lhs): New function.
1313 (relink_to_new_repr): Also relink LHS lists.
1314 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1315 (add_access_to_lhs_work_queue): New function.
1316 (pop_access_from_work_queue): Renamed to
1317 pop_access_from_rhs_work_queue.
1318 (pop_access_from_lhs_work_queue): New function.
1319 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1321 (child_would_conflict_in_lacc): Renamed to
1322 child_would_conflict_in_acc. Adjusted parameter names.
1323 (create_artificial_child_access): New parameter set_grp_read, use it.
1324 (subtree_mark_written_and_enqueue): Renamed to
1325 subtree_mark_written_and_rhs_enqueue.
1326 (propagate_subaccesses_across_link): Renamed to
1327 propagate_subaccesses_from_rhs.
1328 (propagate_subaccesses_from_lhs): New function.
1329 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1332 2020-01-29 Martin Jambor <mjambor@suse.cz>
1334 PR tree-optimization/92706
1335 * tree-sra.c (struct access): Adjust comment of
1336 grp_total_scalarization.
1337 (find_access_in_subtree): Look for single children spanning an entire
1339 (scalarizable_type_p): Allow register accesses, adjust callers.
1340 (completely_scalarize): Remove function.
1341 (scalarize_elem): Likewise.
1342 (create_total_scalarization_access): Likewise.
1343 (sort_and_splice_var_accesses): Do not track total scalarization
1345 (analyze_access_subtree): New parameter totally, adjust to new meaning
1346 of grp_total_scalarization.
1347 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1348 (can_totally_scalarize_forest_p): New function.
1349 (create_total_scalarization_access): Likewise.
1350 (create_total_access_and_reshape): Likewise.
1351 (total_should_skip_creating_access): Likewise.
1352 (totally_scalarize_subtree): Likewise.
1353 (analyze_all_variable_accesses): Perform total scalarization after
1354 subaccess propagation using the new functions above.
1355 (initialize_constant_pool_replacements): Output initializers by
1356 traversing the access tree.
1358 2020-01-29 Martin Jambor <mjambor@suse.cz>
1360 * tree-sra.c (verify_sra_access_forest): New function.
1361 (verify_all_sra_access_forests): Likewise.
1362 (create_artificial_child_access): Set parent.
1363 (analyze_all_variable_accesses): Call the verifier.
1365 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1367 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1368 if called on indirect edge.
1369 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1370 speculative call if needed.
1372 2020-01-29 Richard Biener <rguenther@suse.de>
1374 PR tree-optimization/93428
1375 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1376 permutation when the load node is created.
1377 (vect_analyze_slp_instance): Re-use it here.
1379 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1381 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1383 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1385 PR rtl-optimization/93272
1386 * ira-lives.c (process_out_of_region_eh_regs): New function.
1387 (process_bb_node_lives): Call it.
1389 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1391 * coverage.c (read_counts_file): Make error message lowercase.
1393 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1395 * profile-count.c (profile_quality_display_names): Fix ordering.
1397 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1400 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1401 hash only when edge is first within the sequence.
1402 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1403 (symbol_table::create_edge): Do not set target_prob.
1404 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1406 (cgraph_edge::make_speculative): Drop target_prob parameter.
1407 (cgraph_edge::speculative_call_info): Remove.
1408 (cgraph_edge::first_speculative_call_target): New member function.
1409 (update_call_stmt_hash_for_removing_direct_edge): New function.
1410 (cgraph_edge::resolve_speculation): Rewrite to new API.
1411 (cgraph_edge::speculative_call_for_target): New member function.
1412 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1413 multiple speculation targets.
1414 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1416 (verify_speculative_call): Verify that targets form an interval.
1417 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1418 (cgraph_edge::first_speculative_call_target): New member function.
1419 (cgraph_edge::next_speculative_call_target): New member function.
1420 (cgraph_edge::speculative_call_target_ref): New member function.
1421 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1422 (cgraph_edge): Remove target_prob.
1423 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1424 Fix handling of speculative calls.
1425 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1426 * ipa-fnsummary.c (analyze_function_body): Likewise.
1427 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1428 * ipa-profile.c (dump_histogram): Fix formating.
1429 (ipa_profile_generate_summary): Watch for overflows.
1430 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1431 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1432 (update_indirect_edges_after_inlining): Update to new API.
1433 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1435 * profile-count.h: (profile_probability::adjusted): New.
1436 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1437 updating of profile.
1438 * value-prof.c (gimple_ic_transform): Rename to ...
1439 (dump_ic_profile): ... this one; update dumping.
1440 (stream_in_histogram_value): Fix formating.
1441 (gimple_value_profile_transformations): Update.
1443 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1446 * config/i386/i386.md (*movoi_internal_avx): Remove
1447 TARGET_SSE_TYPELESS_STORES check.
1448 (*movti_internal): Prefer TARGET_AVX over
1449 TARGET_SSE_TYPELESS_STORES.
1450 (*movtf_internal): Likewise.
1451 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1452 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1453 from TARGET_SSE_TYPELESS_STORES.
1455 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1457 * diagnostic-core.h (warning_at): Rename overload to...
1458 (warning_meta): ...this.
1459 (emit_diagnostic_valist): Delete decl of overload taking
1460 diagnostic_metadata.
1461 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1462 (warning_at): Rename overload taking diagnostic_metadata to...
1463 (warning_meta): ...this.
1465 2020-01-28 Richard Biener <rguenther@suse.de>
1467 PR tree-optimization/93439
1468 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1469 * tree-cfg.c (move_sese_region_to_fn): ... here.
1470 (verify_types_in_gimple_reference): Verify used cliques are
1473 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1476 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1477 argument of a pointer to struct gcc_options and pass it to
1478 parse_mtune_ctrl_str.
1479 (ix86_function_specific_restore): Pass opts to
1480 set_ix86_tune_features.
1481 (ix86_option_override_internal): Likewise.
1482 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1483 gcc_options and use it for x_ix86_tune_ctrl_string.
1485 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1487 PR rtl-optimization/87763
1488 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1489 simplification to handle subregs as well as bare regs.
1490 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1492 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1494 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1495 for reduction chains that (now) include a call.
1497 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1499 PR tree-optimization/92822
1500 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1501 out the don't-care elements of a vector whose significant elements
1502 are duplicates, make the don't-care elements duplicates too.
1504 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1506 PR tree-optimization/93434
1507 * tree-predcom.c (split_data_refs_to_components): Record which
1508 components have had aliasing loads removed. Prevent store-store
1509 commoning for all such components.
1511 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1514 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1515 -1 or is_vshift is true, use new_vector with number of elts npatterns
1516 rather than new_unary_operation.
1518 PR tree-optimization/93454
1519 * gimple-fold.c (fold_array_ctor_reference): Perform
1520 elt_size.to_uhwi () just once, instead of calling it in every
1521 iteration. Punt if that value is above size of the temporary
1522 buffer. Decrease third native_encode_expr argument when
1523 bufoff + elt_sz is above size of buf.
1525 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1527 * config/mips/mips.c (mips_declare_object_name)
1528 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1530 2020-01-27 Martin Liska <mliska@suse.cz>
1532 PR gcov-profile/93403
1533 * tree-profile.c (gimple_init_gcov_profiler): Generate
1534 both __gcov_indirect_call_profiler_v4 and
1535 __gcov_indirect_call_profiler_v4_atomic.
1537 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1540 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1542 (@aarch64_split_simd_mov<mode>): Use it.
1543 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1544 Leave the vec_extract patterns to handle 2-element vectors.
1545 (aarch64_simd_mov_from_<mode>high): Likewise.
1546 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1547 (vec_extractv2dfv1df): Likewise.
1549 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1551 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1552 jump conditions for *compare_condjump<GPI:mode>.
1554 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1557 * digraph.cc (test_edge::test_edge): Specify template for base
1560 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1562 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1564 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1566 * config/arc/arc-protos.h (gen_mlo): Remove.
1567 (gen_mhi): Likewise.
1568 * config/arc/arc.c (AUX_MULHI): Define.
1569 (arc_must_save_reister): Special handling for r58/59.
1570 (arc_compute_frame_size): Consider mlo/mhi registers.
1571 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1573 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1574 mlo/mhi name selection.
1575 (arc_restore_callee_saves): Don't early restore blink when ISR.
1576 (arc_expand_prologue): Add mlo/mhi saving.
1577 (arc_expand_epilogue): Add mlo/mhi restoring.
1580 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1581 numbering when MUL64 option is used.
1582 (DWARF2_FRAME_REG_OUT): Define.
1583 * config/arc/arc.md (arc600_stall): New pattern.
1584 (VUNSPEC_ARC_ARC600_STALL): Define.
1585 (mulsi64): Use correct mlo/mhi registers.
1586 (mulsi_600): Clean it up.
1587 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1589 (mhi_operand): Likewise.
1591 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1592 Petro Karashchenko <petro.karashchenko@ring.com>
1594 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1595 attributes if needed.
1596 (prepare_move_operands): Generate special unspec instruction for
1598 (arc_isuncached_mem_p): Propagate uncached attribute to each
1600 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1601 (VUNSPEC_ARC_STDI): Likewise.
1602 (ALLI): New mode iterator.
1603 (mALLI): New mode attribute.
1604 (lddi): New instruction pattern.
1606 (stdidi_split): Split instruction for architectures which are not
1607 supporting ll64 option.
1608 (lddidi_split): Likewise.
1610 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1612 PR rtl-optimization/92989
1613 * lra-lives.c (process_bb_lives): Update the live-in set before
1614 processing additional clobbers.
1616 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1618 PR rtl-optimization/93170
1619 * cselib.c (cselib_invalidate_regno_val): New function, split out
1621 (cselib_invalidate_regno): ...here.
1622 (cselib_invalidated_by_call_p): New function.
1623 (cselib_process_insn): Iterate over all the hard-register entries in
1624 REG_VALUES and invalidate any that cross call-clobbered registers.
1626 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1628 * dojump.c (split_comparison): Use HONOR_NANS rather than
1629 HONOR_SNANS when splitting LTGT.
1631 2020-01-27 Martin Liska <mliska@suse.cz>
1634 * opts.c (print_filtered_help): Exclude language-specific
1635 options from --help=common unless enabled in all FEs.
1637 2020-01-27 Martin Liska <mliska@suse.cz>
1639 * opts.c (print_help): Exclude params from
1640 all except --help=param.
1642 2020-01-27 Martin Liska <mliska@suse.cz>
1645 * config/i386/i386-features.c (make_resolver_func):
1646 Align the code with ppc64 target implementation.
1647 Do not generate a unique name for resolver function.
1649 2020-01-27 Richard Biener <rguenther@suse.de>
1651 PR tree-optimization/93397
1652 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1653 converted reduction chain SLP graph adjustment.
1655 2020-01-26 Marek Polacek <polacek@redhat.com>
1658 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1661 2020-01-26 Jason Merrill <jason@redhat.com>
1664 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1667 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1669 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1670 (rx_setmem): Likewise.
1672 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1675 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1676 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1677 drop <di> from constraint of last operand.
1680 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1681 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1682 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1684 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1687 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1690 2020-01-24 Jeff Law <law@redhat.com>
1692 PR tree-optimization/92788
1693 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1696 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1699 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1700 *avx_vperm_broadcast_<mode>,
1701 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1702 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1703 Move before avx2_perm<mode>/avx512f_perm<mode>.
1706 * simplify-rtx.c (simplify_const_unary_operation,
1707 simplify_const_binary_operation): Punt for mode precision above
1708 MAX_BITSIZE_MODE_ANY_INT.
1710 2020-01-24 Andrew Pinski <apinski@marvell.com>
1712 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1715 2020-01-24 Jeff Law <law@redhat.com>
1718 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1719 for REGs. Call output_operand_lossage to get more reasonable
1722 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1724 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1725 gcn_fp_compare_operator.
1726 (vec_cmpu<mode>di): Use gcn_compare_operator.
1727 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1728 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1729 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1730 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1731 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1732 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1733 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1734 gcn_fp_compare_operator.
1735 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1736 gcn_fp_compare_operator.
1737 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1738 gcn_fp_compare_operator.
1739 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1740 gcn_fp_compare_operator.
1742 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1744 * doc/install.texi (Cross-Compiler-Specific Options): Document
1745 `--with-toolexeclibdir' option.
1747 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1749 * target.def (flags_regnum): Also mention effect on delay slot filling.
1750 * doc/tm.texi: Regenerate.
1752 2020-01-23 Jeff Law <law@redhat.com>
1754 PR translation/90162
1755 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1757 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1760 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1763 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1765 PR rtl-optimization/93402
1766 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1769 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1771 * config.in: Regenerated.
1772 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1773 for TARGET_LIBC_GNUSTACK.
1774 * configure: Regenerated.
1775 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1776 found to be 2.31 or greater.
1778 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1780 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1782 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1783 (mips_asm_file_end): New function. Delegate to
1784 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1785 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1787 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1790 * config/i386/i386-modes.def (POImode): New mode.
1791 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1792 * config/i386/i386.md (DPWI): New mode attribute.
1793 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1795 (QPWI): ... this. Use POI instead of OI for TImode.
1796 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1797 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1800 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1803 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1805 (speculation_tracker_rev): New pattern.
1806 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1807 Use speculation_tracker_rev to track the inverse condition.
1809 2020-01-23 Richard Biener <rguenther@suse.de>
1811 PR tree-optimization/93381
1812 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1813 alias-set of the def as argument and record the first one.
1814 (vn_walk_cb_data::first_set): New member.
1815 (vn_reference_lookup_3): Pass the alias-set of the current def
1816 to push_partial_def. Fix alias-set used in the aggregate copy
1818 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1819 * real.c (clear_significand_below): Fix out-of-bound access.
1821 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1824 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1825 New define_insn patterns.
1827 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1829 * doc/sourcebuild.texi (check-function-bodies): Add an
1830 optional target/xfail selector.
1832 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1834 PR rtl-optimization/93124
1835 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1836 bare USE and CLOBBER insns.
1838 2020-01-22 Andrew Pinski <apinski@marvell.com>
1840 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1842 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1845 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1846 diagnostic_manager into "ana" namespace.
1847 * selftest-run-tests.c (selftest::run_tests): Update for move of
1848 selftest::run_analyzer_selftests to
1849 ana::selftest::run_analyzer_selftests.
1851 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1853 * cfgexpand.c (union_stack_vars): Update the size.
1855 2020-01-22 Richard Biener <rguenther@suse.de>
1857 PR tree-optimization/93381
1858 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1859 throughout, handle all conversions the same.
1861 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1864 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1865 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1866 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1867 Call force_reg on high_in2 unconditionally.
1869 2020-01-22 Martin Liska <mliska@suse.cz>
1871 PR tree-optimization/92924
1872 * profile.c (compute_value_histograms): Divide
1875 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1878 * output.h (assemble_name_resolve): Declare.
1879 * varasm.c (assemble_name_resolve): New function.
1880 (assemble_name): Use it.
1881 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1883 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1885 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1886 update_web_docs_git instead of update_web_docs_svn.
1888 2020-01-21 Andrew Pinski <apinski@marvell.com>
1891 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1892 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1893 (*tlsgd_small_<mode>): Likewise.
1894 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1895 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1896 register. Convert that register back to dest using convert_mode.
1898 2020-01-21 Jim Wilson <jimw@sifive.com>
1900 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1903 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1904 Uros Bizjak <ubizjak@gmail.com>
1907 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1909 (legitimize_tls_address): Do GNU2 TLS address computation in
1910 ptr_mode and zero-extend result to Pmode.
1911 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1912 :P with :PTR and Pmode with ptr_mode.
1913 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1914 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1915 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1917 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1920 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1921 the last two operands are CONST_INT_P before using them as such.
1923 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1925 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1926 to get the integer element types.
1928 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1930 * config/aarch64/aarch64-sve-builtins.h
1931 (function_expander::convert_to_pmode): Declare.
1932 * config/aarch64/aarch64-sve-builtins.cc
1933 (function_expander::convert_to_pmode): New function.
1934 (function_expander::get_contiguous_base): Use it.
1935 (function_expander::prepare_gather_address_operands): Likewise.
1936 * config/aarch64/aarch64-sve-builtins-sve2.cc
1937 (svwhilerw_svwhilewr_impl::expand): Likewise.
1939 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1942 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1943 cfun->machine->label_is_assembled.
1944 (aarch64_print_patchable_function_entry): New.
1945 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1946 * config/aarch64/aarch64.h (struct machine_function): New field,
1949 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1952 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1955 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1958 * cgraph.c (cgraph_edge::resolve_speculation,
1959 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1960 call_stmt_site_hash.
1962 2020-01-21 Martin Liska <mliska@suse.cz>
1964 * config/rs6000/rs6000.c (common_mode_defined): Remove
1967 2020-01-21 Richard Biener <rguenther@suse.de>
1969 PR tree-optimization/92328
1970 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1971 type when value-numbering same-sized store by inserting a
1973 (eliminate_dom_walker::eliminate_stmt): When eliminating
1974 a redundant store handle bit-reinterpretation of the same value.
1976 2020-01-21 Andrew Pinski <apinski@marvel.com>
1979 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1981 (prepare_block_for_update): This. Use a worklist instead of
1984 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1986 * gcc/config/arm/arm.c (clear_operation_p):
1987 Initialise last_regno, skip first iteration
1988 based on the first_set value and use ints instead
1989 of the unnecessary HOST_WIDE_INTs.
1991 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1994 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1995 compare_mode other than SFmode or DFmode.
1997 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2000 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2001 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2002 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2004 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2006 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2008 2020-01-20 Andrew Pinski <apinski@marvell.com>
2011 * targhooks.c (default_print_patchable_function_entry): Use
2012 output_asm_insn to emit the nop instruction.
2014 2020-01-20 Fangrui Song <maskray@google.com>
2017 * targhooks.c (default_print_patchable_function_entry): Align to
2020 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2023 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2024 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2025 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2026 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2027 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2028 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2029 Remove the {q} suffix from lea.
2030 (*tls_dynamic_gnu2_call_64): Renamed to ...
2031 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2032 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2033 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2034 Pass Pmode to gen_tls_dynamic_gnu2_64.
2036 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2038 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2040 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2042 * config/aarch64/aarch64-sve-builtins-base.cc
2043 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2045 2020-01-20 Richard Biener <rguenther@suse.de>
2048 * dwarf2out.c (prune_unused_types): Unconditionally mark
2049 called function DIEs.
2051 2020-01-20 Martin Liska <mliska@suse.cz>
2053 PR tree-optimization/93199
2054 * tree-eh.c (struct leh_state): Add
2055 new field outer_non_cleanup.
2056 (cleanup_is_dead_in): Pass leh_state instead
2057 of eh_region. Add a checking that state->outer_non_cleanup
2058 points to outer non-clean up region.
2059 (lower_try_finally): Record outer_non_cleanup
2061 (lower_catch): Likewise.
2062 (lower_eh_filter): Likewise.
2063 (lower_eh_must_not_throw): Likewise.
2064 (lower_cleanup): Likewise.
2066 2020-01-20 Richard Biener <rguenther@suse.de>
2068 PR tree-optimization/93094
2069 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2070 (vect_transform_loop): Likewise.
2071 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2072 loop_vectorized_call to vect_transform_loop.
2073 * tree-vect-loop.c (vect_transform_loop): Pass down
2074 loop_vectorized_call to vect_loop_versioning.
2075 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2076 the earlier discovered loop_vectorized_call.
2078 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2080 * doc/contribute.texi: Update for SVN -> Git transition.
2081 * doc/install.texi: Likewise.
2083 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2086 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2087 speculative targets.
2088 (verify_speculative_call): New function
2089 (cgraph_node::verify_node): Use it.
2090 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2093 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2096 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2097 (cgraph_edge::make_direct): Remove all indirect targets.
2098 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2099 (cgraph_node::verify_node): Verify that only one call_stmt or
2100 lto_stmt_uid is set.
2101 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2103 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2104 (lto_output_ref): Simplify streaming of stmt.
2105 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2107 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2109 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2110 Mark parameter unused.
2112 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2114 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2116 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2118 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2120 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2122 * Makefile.in: Add coroutine-passes.o.
2123 * builtin-types.def (BT_CONST_SIZE): New.
2124 (BT_FN_BOOL_PTR): New.
2125 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2126 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2127 * coroutine-builtins.def: New file.
2128 * coroutine-passes.cc: New file.
2129 * function.h (struct GTY function): Add a bit to indicate that the
2130 function is a coroutine component.
2131 * internal-fn.c (expand_CO_FRAME): New.
2132 (expand_CO_YIELD): New.
2133 (expand_CO_SUSPN): New.
2134 (expand_CO_ACTOR): New.
2135 * internal-fn.def (CO_ACTOR): New.
2139 * passes.def: Add pass_coroutine_lower_builtins,
2140 pass_coroutine_early_expand_ifns.
2141 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2142 (make_pass_coroutine_early_expand_ifns): New.
2143 * doc/invoke.texi: Document the fcoroutines command line
2146 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2148 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2151 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2152 after checking the argument is a REG. Don't use REGNO (reg)
2153 again to set last_regno, reuse regno variable instead.
2155 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2157 * doc/analyzer.texi (Limitations): Add note about NaN.
2159 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2160 Sudakshina Das <sudi.das@arm.com>
2162 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2163 and valid immediate.
2164 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2165 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2166 * config/arm/constraints.md (Pg): New.
2167 * config/arm/predicates.md (long_shift_imm): New.
2168 (arm_reg_or_long_shift_imm): Likewise.
2169 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2170 (thumb2_lsll): Likewise.
2173 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2174 Sudakshina Das <sudi.das@arm.com>
2176 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2177 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2178 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2179 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2180 * config/arm/thumb2.md (thumb2_asrl): New.
2181 (thumb2_lsll): Likewise.
2183 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2185 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2188 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2190 * gdbinit.in (help-gcc-hooks): New command.
2191 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2192 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2195 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2197 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2198 correct target macro.
2200 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2202 * config/aarch64/aarch64-protos.h
2203 (aarch64_sve_ld1ro_operand_p): New.
2204 * config/aarch64/aarch64-sve-builtins-base.cc
2205 (class load_replicate): New.
2206 (class svld1ro_impl): New.
2207 (class svld1rq_impl): Change to inherit from load_replicate.
2208 (svld1ro): New sve intrinsic function base.
2209 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2210 New DEF_SVE_FUNCTION.
2211 * config/aarch64/aarch64-sve-builtins-base.h
2212 (svld1ro): New decl.
2213 * config/aarch64/aarch64-sve-builtins.cc
2214 (function_expander::add_mem_operand): Modify assert to allow
2216 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2218 * config/aarch64/aarch64.c
2219 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2220 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2221 (aarch64_sve_ld1ro_operand_p): New.
2222 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2223 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2224 * config/aarch64/predicates.md
2225 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2227 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2229 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2230 Introduce this ACLE specified predefined macro.
2231 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2232 (fp): Disabling this disables f64mm.
2233 (simd): Disabling this disables f64mm.
2234 (fp16): Disabling this disables f64mm.
2235 (sve): Disabling this disables f64mm.
2236 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2237 (AARCH64_ISA_F64MM): New.
2238 (TARGET_F64MM): New.
2239 * doc/invoke.texi (f64mm): Document new option.
2241 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2243 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2244 (neoversen1_tunings): Likewise.
2246 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2249 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2250 Add assert to ensure prolog has been emitted.
2251 (aarch64_split_atomic_op): Likewise.
2252 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2253 Use epilogue_completed rather than reload_completed.
2254 (aarch64_atomic_exchange<mode>): Likewise.
2255 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2256 (atomic_nand<mode>): Likewise.
2257 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2258 (atomic_fetch_nand<mode>): Likewise.
2259 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2260 (atomic_nand_fetch<mode>): Likewise.
2262 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2265 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2267 (REVERSE_CONDITION): Delete.
2268 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2269 (CCFP_CCFPE): Likewise.
2270 (e): New mode attribute.
2271 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2272 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2273 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2274 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2275 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2276 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2277 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2278 name of generator from gen_ccmpdi to gen_ccmpccdi.
2279 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2280 the previous comparison but aren't able to, use the new ccmp_rev
2283 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2285 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2286 than testing directly for INTEGER_CST.
2287 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2289 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2291 PR tree-optimization/93292
2292 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2293 get_vectype_for_scalar_type returns NULL.
2295 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2297 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2298 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2300 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2302 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2304 * params.opt: (max-predicted-iterations): Set bounds.
2305 * predict.c (real_almost_one, real_br_prob_base,
2306 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2307 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2308 probabilities; do not truncate to reg_br_prob_bases.
2309 (estimate_loops_at_level): Pass max_cyclic_prob.
2310 (estimate_loops): Compute max_cyclic_prob.
2311 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2313 * profile-count.c (profile_probability::to_sreal): New.
2314 * profile-count.h (class sreal): Move up in file.
2315 (profile_probability::to_sreal): Declare.
2317 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2320 (arm_invalid_conversion): New function for target hook.
2321 (arm_invalid_unary_op): New function for target hook.
2322 (arm_invalid_binary_op): New function for target hook.
2324 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2326 * config.gcc: Add arm_bf16.h.
2327 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2328 (arm_simd_builtin_std_type): Add BFmode.
2329 (arm_init_simd_builtin_types): Define element types for vector types.
2330 (arm_init_bf16_types): New function.
2331 (arm_init_builtins): Add arm_init_bf16_types function call.
2332 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2333 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2334 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2335 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2336 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2337 (arm_mangle_type): Add __bf16.
2338 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2339 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2340 arm_bf16_ptr_type_node.
2341 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2342 define_split between ARM registers.
2343 * config/arm/arm_bf16.h: New file.
2344 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2345 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2347 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2348 * config/arm/vfp.md: Add BFmode to movhf patterns.
2350 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2351 Andre Vieira <andre.simoesdiasvieira@arm.com>
2353 * config/arm/arm-cpus.in (mve, mve_float): New features.
2354 (dsp, mve, mve.fp): New options.
2355 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2356 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2357 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2359 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2360 Thomas Preud'homme <thomas.preudhomme@arm.com>
2362 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2364 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2365 error for using -mcmse when targeting Armv8.1-M Mainline.
2367 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2368 Thomas Preud'homme <thomas.preudhomme@arm.com>
2370 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2371 address in r4 when targeting Armv8.1-M Mainline.
2372 (nonsecure_call_value_internal): Likewise.
2373 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2374 a register match_operand again. Emit BLXNS when targeting
2376 (nonsecure_call_value_reg_thumb2): Likewise.
2378 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2379 Thomas Preud'homme <thomas.preudhomme@arm.com>
2381 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2382 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2383 variable as true when floating-point ABI is not hard. Replace
2384 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2385 Generate VLSTM and VLLDM instruction respectively before and
2386 after a function call to cmse_nonsecure_call function.
2387 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2388 (VUNSPEC_VLLDM): Likewise.
2389 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2390 (lazy_load_multiple_insn): Likewise.
2392 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2393 Thomas Preud'homme <thomas.preudhomme@arm.com>
2395 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2396 (arm_emit_vfp_multi_reg_pop): Likewise.
2397 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2398 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2399 restore callee-saved VFP registers.
2401 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2402 Thomas Preud'homme <thomas.preudhomme@arm.com>
2404 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2405 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2406 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2407 callee-saved GPRs as well as clear ip register before doing a nonsecure
2408 call then restore callee-saved GPRs after it when targeting
2410 (arm_reorg): Adapt to function rename.
2412 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2413 Thomas Preud'homme <thomas.preudhomme@arm.com>
2415 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2416 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2417 clear_vfp_multiple pattern based on a new vfp parameter.
2418 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2419 targeting Armv8.1-M Mainline.
2420 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2421 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2422 whether VFP registers are available before looking call_used_regs for a
2424 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2425 of prototype of clear_operation_p.
2426 (clear_vfp_multiple_operation): New predicate.
2427 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2428 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2430 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2431 Thomas Preud'homme <thomas.preudhomme@arm.com>
2433 * config/arm/arm-protos.h (clear_operation_p): Declare.
2434 * config/arm/arm.c (clear_operation_p): New function.
2435 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2436 targeting Armv8.1-M Mainline or successor.
2437 (output_return_instruction): Only output APSR register clearing if
2438 Armv8.1-M Mainline instructions not available.
2439 (thumb_exit): Likewise.
2440 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2441 * config/arm/thumb2.md (clear_apsr): New define_insn.
2442 (clear_multiple): Likewise.
2443 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2445 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2446 Thomas Preud'homme <thomas.preudhomme@arm.com>
2448 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2449 (use_return_insn): Also return false for Armv8.1-M Mainline.
2450 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2451 Mainline instructions are available.
2452 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2453 when targeting Armv8.1-M Mainline Security Extensions.
2454 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2455 Mainline entry function.
2456 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2457 targeting Armv8.1-M Mainline or successor.
2458 (arm_expand_epilogue): Fix indentation of caller-saved register
2459 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2461 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2462 (FP_SYSREGS): Likewise.
2463 (enum vfp_sysregs_encoding): Define enum.
2464 (fp_sysreg_names): Declare.
2465 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2466 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2467 (pop_fpsysreg_insn): Likewise.
2469 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2470 Thomas Preud'homme <thomas.preudhomme@arm.com>
2472 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2473 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2474 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2475 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2476 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2477 (ARMv8_1m_main): New feature group.
2478 (armv8.1-m.main): New architecture.
2479 * config/arm/arm-tables.opt: Regenerate.
2480 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2481 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2482 (arm_options_perform_arch_sanity_checks): Error out when targeting
2483 Armv8.1-M Mainline Security Extensions.
2484 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2486 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2488 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2489 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2490 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2491 aarch64_bfdot_laneq): New.
2492 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2493 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2494 vbfdotq_laneq_f32): New.
2495 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2496 VBFMLA_W, VBF): New.
2497 (isquadop): Add V4BF, V8BF.
2499 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2501 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2502 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2503 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2504 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2505 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2506 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2507 usdot_laneq, sudot_lane,sudot_laneq): New.
2508 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2509 (aarch64_<sur>dot_lane): New.
2510 * config/aarch64/arm_neon.h (vusdot_s32): New.
2512 (vusdot_lane_s32): New.
2513 (vsudot_lane_s32): New.
2514 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2515 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2517 2020-01-16 Martin Liska <mliska@suse.cz>
2519 * value-prof.c (dump_histogram_value): Fix
2520 obvious spacing issue.
2522 2020-01-16 Andrew Pinski <apinski@marvell.com>
2524 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2525 !storage_order_barrier_p.
2527 2020-01-16 Andrew Pinski <apinski@marvell.com>
2529 * sched-int.h (_dep): Add unused bit-field field for the padding.
2530 * sched-deps.c (init_dep_1): Init unused field.
2532 2020-01-16 Andrew Pinski <apinski@marvell.com>
2534 * optabs.h (create_expand_operand): Initialize target field also.
2536 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2538 PR tree-optimization/92429
2539 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2540 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2542 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2545 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2547 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2548 aarch64_sve_int_mode to each mode.
2550 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2552 * doc/analyzer.texi (Overview): Add note about
2553 -fdump-ipa-analyzer.
2555 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2557 PR tree-optimization/93231
2558 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2559 input_type is unsigned. Use tree_to_shwi for shift constant.
2560 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2561 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2562 inputs more efficiently.
2564 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2566 * config/i386/i386.md (*movsf_internal): Do not require
2567 SSE2 ISA for alternatives 14 and 15.
2569 2020-01-15 Richard Biener <rguenther@suse.de>
2572 * tree-eh.c (sink_clobbers): If we already visited the destination
2573 block do not defer insertion.
2574 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2575 the purpose of defered insertion.
2577 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2579 * BASE-VER: Bump to 10.0.1.
2581 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2583 PR tree-optimization/93247
2584 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2585 type of the stmt that we're going to vectorize.
2587 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2589 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2590 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2593 2020-01-15 Martin Liska <mliska@suse.cz>
2595 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2596 2 calls of streamer_read_hwi in a function call.
2598 2020-01-15 Richard Biener <rguenther@suse.de>
2600 * alias.c (record_alias_subset): Avoid redundant work when
2601 subset is already recorded.
2603 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2605 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2606 the analyzer options provide CWE identifiers.
2608 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2610 * tree-diagnostic-path.cc (path_summary::event_range::print):
2611 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2612 using get_pure_location.
2614 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2616 PR tree-optimization/93262
2617 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2618 perform head trimming only if the last argument is constant,
2619 either all ones, or larger or equal to head trim, in the latter
2620 case decrease the last argument by head_trim.
2622 PR tree-optimization/93249
2623 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2624 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2625 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2626 perform head trim unless we can prove there are no '\0' chars
2627 from the source among the first head_trim chars.
2629 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2631 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2633 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2636 * config/i386/sse.md
2637 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2638 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2639 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2640 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2641 just a single alternative instead of two, make operands 1 and 2
2644 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2647 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2650 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2652 * Makefile.in (lang_opt_files): Add analyzer.opt.
2653 (ANALYZER_OBJS): New.
2654 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2655 tristate.o and ANALYZER_OBJS.
2656 (TEXI_GCCINT_FILES): Add analyzer.texi.
2657 * common.opt (-fanalyzer): New driver option.
2658 * config.in: Regenerate.
2659 * configure: Regenerate.
2660 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2661 (gccdepdir): Also create depdir for "analyzer" subdir.
2662 * digraph.cc: New file.
2663 * digraph.h: New file.
2664 * doc/analyzer.texi: New file.
2665 * doc/gccint.texi ("Static Analyzer") New menu item.
2666 (analyzer.texi): Include it.
2667 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2668 ("Warning Options"): Add static analysis warnings to the list.
2669 (-Wno-analyzer-double-fclose): New option.
2670 (-Wno-analyzer-double-free): New option.
2671 (-Wno-analyzer-exposure-through-output-file): New option.
2672 (-Wno-analyzer-file-leak): New option.
2673 (-Wno-analyzer-free-of-non-heap): New option.
2674 (-Wno-analyzer-malloc-leak): New option.
2675 (-Wno-analyzer-possible-null-argument): New option.
2676 (-Wno-analyzer-possible-null-dereference): New option.
2677 (-Wno-analyzer-null-argument): New option.
2678 (-Wno-analyzer-null-dereference): New option.
2679 (-Wno-analyzer-stale-setjmp-buffer): New option.
2680 (-Wno-analyzer-tainted-array-index): New option.
2681 (-Wno-analyzer-use-after-free): New option.
2682 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2683 (-Wno-analyzer-use-of-uninitialized-value): New option.
2684 (-Wanalyzer-too-complex): New option.
2685 (-fanalyzer-call-summaries): New warning.
2686 (-fanalyzer-checker=): New warning.
2687 (-fanalyzer-fine-grained): New warning.
2688 (-fno-analyzer-state-merge): New warning.
2689 (-fno-analyzer-state-purge): New warning.
2690 (-fanalyzer-transitivity): New warning.
2691 (-fanalyzer-verbose-edges): New warning.
2692 (-fanalyzer-verbose-state-changes): New warning.
2693 (-fanalyzer-verbosity=): New warning.
2694 (-fdump-analyzer): New warning.
2695 (-fdump-analyzer-callgraph): New warning.
2696 (-fdump-analyzer-exploded-graph): New warning.
2697 (-fdump-analyzer-exploded-nodes): New warning.
2698 (-fdump-analyzer-exploded-nodes-2): New warning.
2699 (-fdump-analyzer-exploded-nodes-3): New warning.
2700 (-fdump-analyzer-supergraph): New warning.
2701 * doc/sourcebuild.texi (dg-require-dot): New.
2702 (dg-check-dot): New.
2703 * gdbinit.in (break-on-saved-diagnostic): New command.
2704 * graphviz.cc: New file.
2705 * graphviz.h: New file.
2706 * ordered-hash-map-tests.cc: New file.
2707 * ordered-hash-map.h: New file.
2708 * passes.def (pass_analyzer): Add before
2709 pass_ipa_whole_program_visibility.
2710 * selftest-run-tests.c (selftest::run_tests): Call
2711 selftest::ordered_hash_map_tests_cc_tests.
2712 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2714 * shortest-paths.h: New file.
2715 * timevar.def (TV_ANALYZER): New timevar.
2716 (TV_ANALYZER_SUPERGRAPH): Likewise.
2717 (TV_ANALYZER_STATE_PURGE): Likewise.
2718 (TV_ANALYZER_PLAN): Likewise.
2719 (TV_ANALYZER_SCC): Likewise.
2720 (TV_ANALYZER_WORKLIST): Likewise.
2721 (TV_ANALYZER_DUMP): Likewise.
2722 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2723 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2724 * tree-pass.h (make_pass_analyzer): New decl.
2725 * tristate.cc: New file.
2726 * tristate.h: New file.
2728 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2731 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2732 alternatives 9 and 10.
2734 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2736 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2737 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2738 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2739 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2740 (selftest::hash_map_tests_c_tests): Call it.
2741 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2742 New static constant, using the value of = H::empty_zero_p.
2743 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2744 from default_hash_traits <Value>.
2745 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2747 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2748 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2749 calls to mark_empty with !Descriptor::empty_zero_p.
2750 (hash_table::empty_slow): Conditionalize the memset call with a
2751 check that Descriptor::empty_zero_p; otherwise, loop through the
2752 entries calling mark_empty on them.
2753 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2754 (pointer_hash::empty_zero_p): Likewise.
2755 (pair_hash::empty_zero_p): Likewise.
2756 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2758 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2759 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2760 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2761 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2762 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2763 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2764 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2765 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2767 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2770 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2772 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2773 fix typo on return value.
2775 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2778 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2780 (cgraph_edge::make_speculative): Add param for setting speculative_id
2782 (cgraph_edge::speculative_call_info): Update comments and find reference
2783 by speculative_id for multiple indirect targets.
2784 (cgraph_edge::resolve_speculation): Decrease the speculations
2785 for indirect edge, drop it's speculative if not direct target
2786 left. Update comments.
2787 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2788 (cgraph_node::dump): Print num_speculative_call_targets.
2789 (cgraph_node::verify_node): Don't report error if speculative
2790 edge not include statement.
2791 (cgraph_edge::num_speculative_call_targets_p): New function.
2792 * cgraph.h (int common_target_id): Remove.
2793 (int common_target_probability): Remove.
2794 (num_speculative_call_targets): New variable.
2795 (make_speculative): Add param for setting speculative_id.
2796 (cgraph_edge::num_speculative_call_targets_p): New declare.
2797 (target_prob): New variable.
2798 (speculative_id): New variable.
2799 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2800 call summaries for multiple speculative call targets.
2801 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2802 * ipa-profile.c (struct speculative_call_target): New struct.
2803 (class speculative_call_summary): New class.
2804 (class speculative_call_summaries): New class.
2805 (call_sums): New variable.
2806 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2807 (ipa_profile_write_edge_summary): New function.
2808 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2809 (ipa_profile_dump_all_summaries): New function.
2810 (ipa_profile_read_edge_summary): New function.
2811 (ipa_profile_read_summary_section): New function.
2812 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2813 (ipa_profile): Generate num_speculative_call_targets from
2815 * ipa-ref.h (speculative_id): New variable.
2816 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2817 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2818 common_target_probability. Stream out speculative_id and
2819 num_speculative_call_targets.
2820 (input_edge): Likewise.
2821 * predict.c (dump_prediction): Remove edges count assert to be
2823 * symtab.c (symtab_node::create_reference): Init speculative_id.
2824 (symtab_node::clone_references): Clone speculative_id.
2825 (symtab_node::clone_referring): Clone speculative_id.
2826 (symtab_node::clone_reference): Clone speculative_id.
2827 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2828 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2829 if indirect call contains multiple speculative targets.
2830 * value-prof.h (check_ic_target): Remove.
2831 * value-prof.c (gimple_value_profile_transformations):
2832 Use void function gimple_ic_transform.
2833 * value-prof.c (gimple_ic_transform): Handle topn case.
2834 Fix comment typos. Change it to a void function.
2836 2020-01-13 Andrew Pinski <apinski@marvell.com>
2838 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2839 (octeontx2t98): New define.
2840 (octeontx2t96): New define.
2841 (octeontx2t93): New define.
2842 (octeontx2f95): New define.
2843 (octeontx2f95n): New define.
2844 (octeontx2f95mm): New define.
2845 * config/aarch64/aarch64-tune.md: Regenerate.
2846 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2848 2020-01-13 Jason Merrill <jason@redhat.com>
2850 PR c++/33799 - destroy return value if local cleanup throws.
2851 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2853 2020-01-13 Martin Liska <mliska@suse.cz>
2855 * ipa-cp.c (get_max_overall_size): Use newly
2856 renamed param param_ipa_cp_unit_growth.
2857 * params.opt: Remove legacy param name.
2859 2020-01-13 Martin Sebor <msebor@redhat.com>
2861 PR tree-optimization/93213
2862 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2863 stores to be eliminated.
2865 2020-01-13 Martin Liska <mliska@suse.cz>
2867 * opts.c (print_help): Do not print CL_PARAM
2868 and CL_WARNING for CL_OPTIMIZATION.
2870 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2873 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2874 depending on optimization settings.
2876 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2878 PR tree-optimization/90838
2879 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2880 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2881 argument rather than to initialize temporary for targets that
2882 don't use the mode argument at all. Initialize ctzval to avoid
2885 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2887 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2888 * tree-core.h: Document it.
2889 * gimplify.c (gimplify_omp_workshare): Set it.
2890 * omp-low.c (lower_omp_target): Use it.
2891 * tree-pretty-print.c (dump_omp_clause): Print it.
2893 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2894 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2896 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2898 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2899 * common.opt (fdiagnostics-path-format=): New option.
2900 (diagnostic_path_format): New enum.
2901 (fdiagnostics-show-path-depths): New option.
2902 * coretypes.h (diagnostic_event_id_t): New forward decl.
2903 * diagnostic-color.c (color_dict): Add "path".
2904 * diagnostic-event-id.h: New file.
2905 * diagnostic-format-json.cc (json_from_expanded_location): Make
2907 (json_end_diagnostic): Call context->make_json_for_path if it
2908 exists and the diagnostic has a path.
2909 (diagnostic_output_format_init): Clear context->print_path.
2910 * diagnostic-path.h: New file.
2911 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2912 when printing a run of events in a diagnostic_path so that they
2913 all get the same color.
2914 (layout::m_diagnostic_path_p): New field.
2915 (layout::layout): Initialize it.
2916 (layout::print_any_labels): Don't colorize the label text for an
2917 event in a diagnostic_path.
2918 (gcc_rich_location::add_location_if_nearby): Add
2919 "restrict_to_current_line_spans" and "label" params. Pass the
2920 former to layout.maybe_add_location_range; pass the latter
2921 when calling add_range.
2922 * diagnostic.c: Include "diagnostic-path.h".
2923 (diagnostic_initialize): Initialize context->path_format and
2924 context->show_path_depths.
2925 (diagnostic_show_any_path): New function.
2926 (diagnostic_path::interprocedural_p): New function.
2927 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2928 (simple_diagnostic_path::num_events): New function.
2929 (simple_diagnostic_path::get_event): New function.
2930 (simple_diagnostic_path::add_event): New function.
2931 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2932 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2933 (debug): New overload taking a diagnostic_path *.
2934 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2935 * diagnostic.h (enum diagnostic_path_format): New enum.
2936 (json::value): New forward decl.
2937 (diagnostic_context::path_format): New field.
2938 (diagnostic_context::show_path_depths): New field.
2939 (diagnostic_context::print_path): New callback field.
2940 (diagnostic_context::make_json_for_path): New callback field.
2941 (diagnostic_show_any_path): New decl.
2942 (json_from_expanded_location): New decl.
2943 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2944 (-fdiagnostics-show-path-depths): New option.
2945 (-fdiagnostics-color): Add "path" to description of default
2946 GCC_COLORS; describe it.
2947 (-fdiagnostics-format=json): Document how diagnostic paths are
2948 represented in the JSON output format.
2949 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2950 Add optional params "restrict_to_current_line_spans" and "label".
2951 * opts.c (common_handle_option): Handle
2952 OPT_fdiagnostics_path_format_ and
2953 OPT_fdiagnostics_show_path_depths.
2954 * pretty-print.c: Include "diagnostic-event-id.h".
2955 (pp_format): Implement "%@" format code for printing
2956 diagnostic_event_id_t *.
2957 (selftest::test_pp_format): Add tests for "%@".
2958 * selftest-run-tests.c (selftest::run_tests): Call
2959 selftest::tree_diagnostic_path_cc_tests.
2960 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2961 * toplev.c (general_init): Initialize global_dc->path_format and
2962 global_dc->show_path_depths.
2963 * tree-diagnostic-path.cc: New file.
2964 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2965 non-static. Drop "diagnostic" param in favor of storing the
2966 original value of "where" and re-using it.
2967 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2968 maybe_unwind_expanded_macro_loc.
2969 (tree_diagnostics_defaults): Initialize context->print_path and
2970 context->make_json_for_path.
2971 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2973 (default_tree_make_json_for_path): New decl.
2974 (maybe_unwind_expanded_macro_loc): New decl.
2976 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2978 PR tree-optimization/93210
2979 * fold-const.h (native_encode_initializer,
2980 can_native_interpret_type_p): Declare.
2981 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2983 (native_encode_initializer): New function, moved from dwarf2out.c.
2984 Adjust to native_encode_expr compatible arguments, including dry-run
2985 and partial extraction modes. Don't handle STRING_CST.
2986 (can_native_interpret_type_p): No longer static.
2987 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2988 offset / BITS_PER_UNIT fits into int and don't call it if
2989 can_native_interpret_type_p fails. If suboff is NULL and for
2990 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2991 native_encode_initializer.
2992 (fold_const_aggregate_ref_1): Formatting fix.
2993 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2994 (tree_add_const_value_attribute): Adjust caller.
2996 PR tree-optimization/90838
2997 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2998 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2999 CTZ_DEFINED_VALUE_AT_ZERO.
3001 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3004 * lra-constraints.c (match_reload): Permit input operands have the
3005 same mode as output while other input operands have a different
3008 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3010 PR tree-optimization/90838
3011 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3012 (check_ctz_string): Likewise.
3013 (optimize_count_trailing_zeroes): Likewise.
3014 (simplify_count_trailing_zeroes): Likewise.
3015 (pass_forwprop::execute): Try ctz simplification.
3016 * match.pd: Add matching for ctz idioms.
3018 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3020 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3022 (aarch64_invalid_unary_op): New function for target hook.
3023 (aarch64_invalid_binary_op): New function for target hook.
3025 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3027 * config.gcc: Add arm_bf16.h.
3028 * config/aarch64/aarch64-builtins.c
3029 (aarch64_simd_builtin_std_type): Add BFmode.
3030 (aarch64_init_simd_builtin_types): Define element types for vector
3032 (aarch64_init_bf16_types): New function.
3033 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3034 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3036 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3037 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3039 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3040 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3041 * config/aarch64/aarch64.c
3042 (aarch64_classify_vector_mode): Add support for BF types.
3043 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3044 (aarch64_vq_mode): Add support for BF types.
3045 (aarch64_simd_container_mode): Add support for BF types.
3046 (aarch64_mangle_type): Add support for BF scalar type.
3047 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3048 * config/aarch64/arm_bf16.h: New file.
3049 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3050 * config/aarch64/iterators.md: Add BF types to mode attributes.
3051 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3053 2020-01-10 Jason Merrill <jason@redhat.com>
3055 PR c++/93173 - incorrect tree sharing.
3056 * gimplify.c (copy_if_shared): No longer static.
3057 * gimplify.h: Declare it.
3059 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3061 * doc/invoke.texi (-msve-vector-bits=): Document that
3062 -msve-vector-bits=128 now generates VL-specific code for
3063 little-endian targets.
3064 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3065 build_vector_type_for_mode to construct the data vector types.
3066 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3067 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3068 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3069 for 128-bit vectors.
3071 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3073 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3076 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3078 * config/aarch64/aarch64-builtins.c
3079 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3080 rather than checking the number of elements and the element mode.
3082 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3084 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3085 get_related_vectype_for_scalar_type rather than build_vector_type
3086 to create the index type for a conditional reduction.
3088 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3090 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3091 for any type of gather or scatter, including strided accesses.
3093 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3095 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3098 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3100 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3102 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3103 parameter and its use to reset DR_OFFSET's.
3104 (vect_transform_loop): Remove orig_drs_init argument.
3105 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3106 member of dr_vec_info rather than the offset of the associated
3107 data_reference's innermost_loop_behavior.
3108 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3109 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3110 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3111 get_dr_vinfo_offset.
3112 (vectorizable_store): Likewise.
3113 (vectorizable_load): Likewise.
3115 2020-01-10 Richard Biener <rguenther@suse.de>
3117 * gimple-ssa-store-merging
3118 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3120 2020-01-10 Martin Liska <mliska@suse.cz>
3123 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3124 encapsulation that was there before r280040.
3126 2020-01-10 Richard Biener <rguenther@suse.de>
3129 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3130 sequences to avoid walking them again for secondary opportunities.
3131 (pass_lower_eh_dispatch::execute): Instead actually insert
3134 2020-01-10 Richard Biener <rguenther@suse.de>
3137 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3138 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3139 avoid quadraticness.
3141 2020-01-10 Martin Jambor <mjambor@suse.cz>
3143 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3144 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3145 to get param_ipa_sra_max_replacements.
3146 (param_splitting_across_edge): Pass the caller to
3147 pull_accesses_from_callee.
3149 2020-01-10 Martin Jambor <mjambor@suse.cz>
3151 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3152 * ipa-cp.c (max_new_size): Removed.
3153 (orig_overall_size): New variable.
3154 (get_max_overall_size): New function.
3155 (estimate_local_effects): Use it. Adjust dump.
3156 (decide_about_value): Likewise.
3157 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3158 orig_overall_size. Adjust dump.
3159 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3161 2020-01-10 Martin Jambor <mjambor@suse.cz>
3163 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3164 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3165 instead of param_ipa_max_agg_items.
3166 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3167 optimization info for the callee.
3169 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3171 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3172 markers if debug_inline_points is false.
3174 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3176 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3178 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3179 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3180 aarch64-sve-builtins-sve2.h.
3181 (aarch64-sve-builtins-sve2.o): New rule.
3182 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3183 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3184 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3185 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3186 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3187 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3189 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3190 instructions that are handled here.
3191 (@cond_asrd<mode>): Generalize to...
3192 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3193 (*cond_asrd<mode>_2): Generalize to...
3194 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3195 (*cond_asrd<mode>_z): Generalize to...
3196 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3197 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3198 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3199 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3200 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3202 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3203 (@aarch64_scatter_stnt<mode>): Likewise.
3204 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3205 (@aarch64_mul_lane_<mode>): Likewise.
3206 (@aarch64_sve_suqadd<mode>_const): Likewise.
3207 (*<sur>h<addsub><mode>): Generalize to...
3208 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3210 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3211 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3212 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3213 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3214 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3215 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3216 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3217 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3218 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3219 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3220 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3221 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3222 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3223 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3224 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3225 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3226 (@aarch64_sve2_xar<mode>): Likewise.
3227 (@aarch64_sve2_bcax<mode>): Likewise.
3228 (*aarch64_sve2_eor3<mode>): Rename to...
3229 (@aarch64_sve2_eor3<mode>): ...this.
3230 (@aarch64_sve2_bsl<mode>): New expander.
3231 (@aarch64_sve2_nbsl<mode>): Likewise.
3232 (@aarch64_sve2_bsl1n<mode>): Likewise.
3233 (@aarch64_sve2_bsl2n<mode>): Likewise.
3234 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3235 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3236 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3237 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3238 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3239 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3240 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3241 (<su>mull<bt><Vwide>): Generalize to...
3242 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3244 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3245 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3246 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3247 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3248 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3249 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3250 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3251 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3252 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3253 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3254 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3255 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3256 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3257 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3258 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3259 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3260 (<SHRNB:r>shrnb<mode>): Generalize to...
3261 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3263 (<SHRNT:r>shrnt<mode>): Generalize to...
3264 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3266 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3267 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3268 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3269 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3270 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3271 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3272 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3273 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3274 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3275 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3276 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3277 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3278 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3279 (@aarch64_sve2_cvtnt<mode>): Likewise.
3280 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3281 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3282 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3283 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3284 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3285 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3286 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3287 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3288 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3289 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3290 (@aarch64_sve2_pmul<mode>): Likewise.
3291 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3292 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3293 (@aarch64_sve2_tbl2<mode>): Likewise.
3294 (@aarch64_sve2_tbx<mode>): Likewise.
3295 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3296 (@aarch64_sve2_histcnt<mode>): Likewise.
3297 (@aarch64_sve2_histseg<mode>): Likewise.
3298 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3299 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3300 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3301 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3302 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3303 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3304 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3305 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3306 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3307 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3308 (SVE2_PMULL_PAIR_I): New mode iterators.
3309 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3310 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3311 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3312 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3313 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3314 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3315 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3316 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3317 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3318 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3319 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3320 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3321 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3322 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3323 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3324 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3325 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3326 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3327 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3328 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3329 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3330 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3331 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3332 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3333 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3334 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3335 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3336 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3337 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3338 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3339 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3341 (VNARROW, Ventype): New mode attributes.
3342 (Vewtype): Handle VNx2DI. Fix typo in comment.
3343 (VDOUBLE): New mode attribute.
3344 (sve_lane_con): Handle VNx8HI.
3345 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3346 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3347 (sve_int_op, sve_int_op_rev): Handle the above codes.
3348 (sve_pred_int_rhs2_operand): Likewise.
3349 (MULLBT, SHRNB, SHRNT): Delete.
3350 (SVE_INT_SHIFT_IMM): New int iterator.
3351 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3352 and UNSPEC_WHILEHS for TARGET_SVE2.
3353 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3354 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3355 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3356 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3357 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3358 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3359 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3360 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3361 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3362 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3363 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3364 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3365 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3366 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3367 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3368 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3369 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3370 (optab): Handle the new unspecs.
3371 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3373 (lr): Handle the new unspecs.
3375 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3376 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3377 (sve_int_qsub_op): New int attributes.
3378 (sve_fp_op, rot): Handle the new unspecs.
3379 * config/aarch64/aarch64-sve-builtins.h
3380 (function_resolver::require_matching_pointer_type): Declare.
3381 (function_resolver::resolve_unary): Add an optional boolean argument.
3382 (function_resolver::finish_opt_n_resolution): Add an optional
3383 type_suffix_index argument.
3384 (gimple_folder::redirect_call): Declare.
3385 (gimple_expander::prepare_gather_address_operands): Add an optional
3387 * config/aarch64/aarch64-sve-builtins.cc: Include
3388 aarch64-sve-builtins-sve2.h.
3389 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3390 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3391 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3392 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3393 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3394 (TYPES_s_unsigned): Likewise.
3395 (TYPES_s_integer): Use TYPES_s_unsigned.
3396 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3397 (TYPES_sd_integer): Use them.
3398 (TYPES_d_unsigned): New macro.
3399 (TYPES_d_integer): Use it.
3400 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3401 (TYPES_cvt_narrow): Likewise.
3402 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3403 (preds_mx): New variable.
3404 (function_builder::add_overloaded_function): Allow the new feature
3405 set to be more restrictive than the original one.
3406 (function_resolver::infer_pointer_type): Remove qualifiers from
3407 the pointer type before printing it.
3408 (function_resolver::require_matching_pointer_type): New function.
3409 (function_resolver::resolve_sv_displacement): Handle functions
3410 that don't support 32-bit vector indices or svint32_t vector offsets.
3411 (function_resolver::finish_opt_n_resolution): Take the inferred type
3412 as a separate argument.
3413 (function_resolver::resolve_unary): Optionally treat all forms in
3414 the same way as normal merging functions.
3415 (gimple_folder::redirect_call): New function.
3416 (function_expander::prepare_gather_address_operands): Add an argument
3417 that says whether scaled forms are available. If they aren't,
3418 handle scaling of vector indices and don't add the extension and
3420 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3421 fall back to using cond_* instead.
3422 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3423 Split out the member variables into...
3424 (rtx_code_function_base): ...this new base class.
3425 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3426 (unspec_based_function): Split out the member variables into...
3427 (unspec_based_function_base): ...this new base class.
3428 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3429 (unspec_based_function_exact_insn): New class.
3430 (unspec_based_add_function, unspec_based_add_lane_function)
3431 (unspec_based_lane_function, unspec_based_pred_function)
3432 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3433 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3434 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3436 (unspec_based_fused_function): New class.
3437 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3438 (unspec_based_fused_lane_function): New class.
3439 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3441 (CODE_FOR_MODE1): New macro.
3442 (fixed_insn_function): New class.
3443 (while_comparison): Likewise.
3444 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3445 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3446 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3447 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3448 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3449 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3450 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3451 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3452 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3453 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3454 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3455 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3456 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3457 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3458 Also add an initial argument for unary_convert_narrowt, regardless
3459 of the predication type.
3460 (build_32_64): Allow loads and stores to specify MODE_none.
3461 (build_sv_index64, build_sv_uint_offset): New functions.
3462 (long_type_suffix): New function.
3463 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3464 (binary_imm_long_base, load_gather_sv_base): Likewise.
3465 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3466 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3467 (unary_narrowb_base, unary_narrowt_base): Likewise.
3468 (binary_long_lane_def, binary_long_lane): New shape.
3469 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3470 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3471 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3472 (binary_to_uint_def, binary_to_uint): Likewise.
3473 (binary_wide_def, binary_wide): Likewise.
3474 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3475 (compare_def, compare): Likewise.
3476 (compare_ptr_def, compare_ptr): Likewise.
3477 (load_ext_gather_index_restricted_def,
3478 load_ext_gather_index_restricted): Likewise.
3479 (load_ext_gather_offset_restricted_def,
3480 load_ext_gather_offset_restricted): Likewise.
3481 (load_gather_sv_def): Inherit from load_gather_sv_base.
3482 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3483 (shift_left_imm_def, shift_left_imm): Likewise.
3484 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3485 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3486 (store_scatter_index_restricted_def,
3487 store_scatter_index_restricted): Likewise.
3488 (store_scatter_offset_restricted_def,
3489 store_scatter_offset_restricted): Likewise.
3490 (tbl_tuple_def, tbl_tuple): Likewise.
3491 (ternary_long_lane_def, ternary_long_lane): Likewise.
3492 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3493 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3494 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3495 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3496 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3497 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3498 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3499 (ternary_uint_def, ternary_uint): Likewise.
3500 (unary_convert): Fix typo in comment.
3501 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3502 (unary_long_def, unary_long): Likewise.
3503 (unary_narrowb_def, unary_narrowb): Likewise.
3504 (unary_narrowt_def, unary_narrowt): Likewise.
3505 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3506 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3507 (unary_to_int_def, unary_to_int): Likewise.
3508 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3509 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3510 (svasrd_impl): Delete.
3511 (svcadd_impl::expand): Handle integer operations too.
3512 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3513 new functions to derive the unspec numbers.
3514 (svmla_svmls_lane_impl): Replace with...
3515 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3516 integer operations too.
3517 (svwhile_impl): Rename to...
3518 (svwhilelx_impl): ...this and inherit from while_comparison.
3519 (svasrd): Use unspec_based_function.
3520 (svmla_lane): Use svmla_lane_impl.
3521 (svmls_lane): Use svmls_lane_impl.
3522 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3523 (svwhilele, svwhilelt): Use svwhilelx_impl.
3524 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3525 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3526 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3527 * config/aarch64/aarch64-sve-builtins.def: Include
3528 aarch64-sve-builtins-sve2.def.
3530 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3532 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3533 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3534 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3535 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3536 immediates as well as vector ones.
3537 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3538 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3539 (aarch64_sve_qsub_immediate): Update calls accordingly.
3541 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3543 * config/aarch64/aarch64-sve2.md: Add banner comments.
3544 (<su>mulh<r>s<mode>3): Move further up file.
3545 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3546 (*aarch64_sve2_sra<mode>): Move further down file.
3547 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3549 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3551 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3553 (while_optab_cmp): Handle them.
3554 * config/aarch64/aarch64-sve.md
3555 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3556 and add a "@" marker.
3557 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3558 instead of gen_aarch64_sve2_while_ptest.
3559 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3561 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3563 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3564 (UNSPEC_WHILELE): ...this.
3565 (UNSPEC_WHILE_LO): Rename to...
3566 (UNSPEC_WHILELO): ...this.
3567 (UNSPEC_WHILE_LS): Rename to...
3568 (UNSPEC_WHILELS): ...this.
3569 (UNSPEC_WHILE_LT): Rename to...
3570 (UNSPEC_WHILELT): ...this.
3571 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3572 (cmp_op, while_optab_cmp): Likewise.
3573 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3574 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3575 (svwhilelt): Likewise.
3577 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3579 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3580 (unary_to_uint): Define.
3581 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3582 (unary_count): Rename to...
3583 (unary_to_uint_def, unary_to_uint): ...this.
3584 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3586 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3588 * config/aarch64/aarch64-sve-builtins-functions.h
3589 (code_for_mode_function): New class.
3590 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3591 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3592 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3593 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3594 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3596 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3598 * config/aarch64/iterators.md (addsub): New code attribute.
3599 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3601 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3602 in the asm string and attributes. Fix indentation.
3603 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3605 (@aarch64_sve_<optab><mode>): ...this.
3606 * config/aarch64/aarch64-sve-builtins.h
3607 (function_expander::expand_signed_unpred_op): Delete.
3608 * config/aarch64/aarch64-sve-builtins.cc
3609 (function_expander::expand_signed_unpred_op): Likewise.
3610 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3611 try using code_for_aarch64_sve instead.
3612 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3613 (svqsub_impl): Likewise.
3614 (svqadd, svqsub): Use rtx_code_function instead.
3616 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3618 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3619 (HADDSUB, sur, addsub): Remove them.
3621 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3623 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3624 functions the same way, rather than singling out those that
3625 aren't mapped directly to optabs.
3627 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3629 * target.def (compatible_vector_types_p): New target hook.
3630 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3631 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3632 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3633 * doc/tm.texi: Regenerate.
3634 * gimple-expr.c: Include target.h.
3635 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3636 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3638 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3639 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3640 Use the original predicate if it already has a suitable type.
3642 2020-01-09 Martin Jambor <mjambor@suse.cz>
3644 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3645 resolve_speculation and redirect_call_stmt_to_callee static. Change
3646 return type of set_call_stmt to cgraph_edge *.
3647 * auto-profile.c (afdo_indirect_call): Adjust call to
3648 redirect_call_stmt_to_callee.
3649 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3650 make the this pointer explicit, adjust self-recursive calls and the
3651 call top make_direct. Return the resulting edge.
3652 (cgraph_edge::remove): Make this pointer explicit.
3653 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3654 (cgraph_edge::make_direct): Likewise, adjust call to
3655 resolve_speculation.
3656 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3657 call to set_call_stmt.
3658 (cgraph_update_edges_for_call_stmt_node): Update call to
3659 set_call_stmt and remove.
3660 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3661 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3662 (cgraph_node::create_edge_including_clones): Moved "first" definition
3663 of edge to the block where it was used. Adjusted calls to
3665 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3666 cgraph_edge::remove.
3667 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3668 make_direct and redirect_call_stmt_to_callee.
3669 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3670 resolve_speculation and make_direct.
3671 * ipa-inline-transform.c (inline_transform): Adjust call to
3672 redirect_call_stmt_to_callee.
3673 (check_speculations_1):: Adjust call to resolve_speculation.
3674 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3675 resolve-speculation.
3676 (inline_small_functions): Adjust call to resolve_speculation.
3677 (ipa_inline): Likewise.
3678 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3680 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3681 safe with regards to edge removal, adjust calls to
3682 redirect_call_stmt_to_callee.
3683 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3684 and redirect_call_stmt_to_callee.
3685 * multiple_target.c (create_dispatcher_calls): Adjust call to
3686 redirect_call_stmt_to_callee
3687 (redirect_to_specific_clone): Likewise.
3688 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3689 Adjust calls to cgraph_edge::remove.
3690 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3691 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3692 (expand_call_inline): Adjust call to cgraph_edge::remove.
3694 2020-01-09 Martin Liska <mliska@suse.cz>
3696 * params.opt: Set Optimization for
3697 param_max_speculative_devirt_maydefs.
3699 2020-01-09 Martin Sebor <msebor@redhat.com>
3703 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3705 2020-01-09 Martin Liska <mliska@suse.cz>
3707 * auto-profile.c (auto_profile): Use opt_for_fn
3709 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3710 (propagate_vals_across_arith_jfunc): Likewise.
3711 (hint_time_bonus): Likewise.
3712 (incorporate_penalties): Likewise.
3713 (good_cloning_opportunity_p): Likewise.
3714 (perform_estimation_of_a_value): Likewise.
3715 (estimate_local_effects): Likewise.
3716 (ipcp_propagate_stage): Likewise.
3717 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3718 (set_switch_stmt_execution_predicate): Likewise.
3719 (analyze_function_body): Likewise.
3720 * ipa-inline-analysis.c (offline_size): Likewise.
3721 * ipa-inline.c (early_inliner): Likewise.
3722 * ipa-prop.c (ipa_analyze_node): Likewise.
3723 (ipcp_transform_function): Likewise.
3724 * ipa-sra.c (process_scan_results): Likewise.
3725 (ipa_sra_summarize_function): Likewise.
3726 * params.opt: Rename ipcp-unit-growth to
3727 ipa-cp-unit-growth. Add Optimization for various
3728 IPA-related parameters.
3730 2020-01-09 Richard Biener <rguenther@suse.de>
3733 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3735 2020-01-09 Richard Biener <rguenther@suse.de>
3737 PR tree-optimization/93040
3738 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3740 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3742 * common/config/avr/avr-common.c (avr_option_optimization_table)
3743 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3745 2020-01-09 Martin Liska <mliska@suse.cz>
3747 * cgraphclones.c (symbol_table::materialize_all_clones):
3748 Use cgraph_node::dump_name.
3750 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3753 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3754 output_operand_lossage instead of gcc_unreachable.
3755 * doc/md.texi (riscv f constraint): Fix typo.
3758 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3759 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3760 CONST_SCALAR_INT_P instead of CONST_INT_P.
3761 (*subv<mode>4_1): Rename to ...
3762 (subv<mode>4_1): ... this.
3763 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3764 define_insn_and_split patterns.
3765 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3768 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3770 * vec.c (class selftest::count_dtor): New class.
3771 (selftest::test_auto_delete_vec): New test.
3772 (selftest::vec_c_tests): Call it.
3773 * vec.h (class auto_delete_vec): New class template.
3774 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3776 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3778 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3780 2020-01-08 Jim Wilson <jimw@sifive.com>
3782 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3783 use of TLS_MODEL_LOCAL_EXEC when not pic.
3785 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3787 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3790 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3793 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3794 *stack_protect_set_3 peephole2): Also check that the second
3795 insns source is general_operand.
3798 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3799 predicate for output operand instead of register_operand.
3800 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3801 memory destination and non-memory operands[2].
3803 2020-01-08 Martin Liska <mliska@suse.cz>
3805 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3806 ::dump_asm_name instead of (::name or ::asm_name).
3807 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3808 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3809 (analyze_functions): Likewise.
3810 (expand_all_functions): Likewise.
3811 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3812 (propagate_bits_across_jump_function): Likewise.
3813 (dump_profile_updates): Likewise.
3814 (ipcp_store_bits_results): Likewise.
3815 (ipcp_store_vr_results): Likewise.
3816 * ipa-devirt.c (dump_targets): Likewise.
3817 * ipa-fnsummary.c (analyze_function_body): Likewise.
3818 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3819 (process_hsa_functions): Likewise.
3820 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3821 (set_alias_uids): Likewise.
3822 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3823 * ipa-inline.c (recursive_inlining): Likewise.
3824 (inline_to_all_callers_1): Likewise.
3825 (ipa_inline): Likewise.
3826 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3827 (ipa_propagate_frequency): Likewise.
3828 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3829 (remove_described_reference): Likewise.
3830 * ipa-pure-const.c (worse_state): Likewise.
3831 (check_retval_uses): Likewise.
3832 (analyze_function): Likewise.
3833 (propagate_pure_const): Likewise.
3834 (propagate_nothrow): Likewise.
3835 (dump_malloc_lattice): Likewise.
3836 (propagate_malloc): Likewise.
3837 (pass_local_pure_const::execute): Likewise.
3838 * ipa-visibility.c (optimize_weakref): Likewise.
3839 (function_and_variable_visibility): Likewise.
3840 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3841 (ipa_discover_variable_flags): Likewise.
3842 * lto-streamer-out.c (output_function): Likewise.
3843 (output_constructor): Likewise.
3844 * tree-inline.c (copy_bb): Likewise.
3845 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3846 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3848 2020-01-08 Richard Biener <rguenther@suse.de>
3851 * tree-eh.c (sink_clobbers): Update virtual operands for
3852 the first and last stmt only. Add a dry-run capability.
3853 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3854 after CFG manipulations and in RPO order to catch all
3855 secondary opportunities reliably.
3857 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3860 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3862 2019-01-08 Richard Biener <rguenther@suse.de>
3865 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3866 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3867 virtual operand, also updating SSA use.
3868 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3869 Update stmt after resetting virtual operand.
3870 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3871 * gimple-iterator.c (gsi_remove): When not removing the stmt
3872 permanently do not delink immediate uses or mark the stmt modified.
3874 2020-01-08 Martin Liska <mliska@suse.cz>
3876 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3877 (ipa_call_context::estimate_size_and_time): Likewise.
3878 (inline_analyze_function): Likewise.
3880 2020-01-08 Martin Liska <mliska@suse.cz>
3882 * cgraph.c (cgraph_node::dump): Use systematically
3885 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3887 Add -nodevicespecs option for avr.
3890 * config/avr/avr.opt (-nodevicespecs): New driver option.
3891 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3892 "-specs=device-specs/..." if that option is not set.
3893 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3895 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3897 Implement 64-bit double functions for avr.
3900 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3901 --with-double-comparison.
3902 * doc/install.texi: Document them.
3903 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3904 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3905 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3906 * doc/invoke.texi (AVR Built-in Macros): Document them.
3907 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3908 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3909 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3911 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3914 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3915 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3916 when only building rm-profile multilibs.
3918 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3921 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3922 lattice for a value to check.
3923 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3924 finite propagation in self-recursive scc.
3926 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3928 * ipa-inline.c (caller_growth_limits): Restore the AND.
3930 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3932 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3933 (VEC_ALLREG_ALT): New iterator.
3934 (VEC_ALLREG_INT_MODE): New iterator.
3935 (VCMP_MODE): New iterator.
3936 (VCMP_MODE_INT): New iterator.
3937 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3938 (vec_cmp<u>v64qidi): New define_expand.
3939 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3940 (vec_cmpu<mode>di_exec): New define_expand.
3941 (vec_cmp<u>v64qidi_exec): New define_expand.
3942 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3943 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3944 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3945 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3946 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3947 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3948 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3949 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3950 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3951 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3953 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3954 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3956 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3958 * config/gcn/constraints.md (DA): Update description and match.
3960 (Db): New constraint.
3961 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3963 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3964 Implement 'Db' mixed immediate type.
3965 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3966 (addcv64si3_dup<exec_vcc>): Delete.
3967 (subcv64si3<exec_vcc>): Rework constraints.
3968 (addv64di3): Rework constraints.
3969 (addv64di3_exec): Rework constraints.
3970 (subv64di3): Rework constraints.
3971 (addv64di3_dup): Delete.
3972 (addv64di3_dup_exec): Delete.
3973 (addv64di3_zext): Rework constraints.
3974 (addv64di3_zext_exec): Rework constraints.
3975 (addv64di3_zext_dup): Rework constraints.
3976 (addv64di3_zext_dup_exec): Rework constraints.
3977 (addv64di3_zext_dup2): Rework constraints.
3978 (addv64di3_zext_dup2_exec): Rework constraints.
3979 (addv64di3_sext_dup2): Rework constraints.
3980 (addv64di3_sext_dup2_exec): Rework constraints.
3982 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3984 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3985 existing target checks.
3987 2020-01-07 Richard Biener <rguenther@suse.de>
3989 * doc/install.texi: Bump minimal supported MPC version.
3991 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3993 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3994 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3995 * langhooks.c: Include stor-layout.h.
3996 (lhd_simulate_enum_decl): New function.
3997 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3998 handle_arm_sve_h for the LTO frontend.
3999 (register_vector_type): Cope with null returns from pushdecl.
4001 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4003 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4004 (aarch64_sve::nvectors_if_data_type): Replace with...
4005 (aarch64_sve::builtin_type_p): ...this.
4006 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4007 (find_vector_type): Delete.
4008 (add_sve_type_attribute): New function.
4009 (lookup_sve_type_attribute): Likewise.
4010 (register_builtin_types): Add an "SVE type" attribute to each type.
4011 (register_tuple_type): Likewise.
4012 (svbool_type_p, nvectors_if_data_type): Delete.
4013 (mangle_builtin_type): Use lookup_sve_type_attribute.
4014 (builtin_type_p): Likewise. Add an overload that returns the
4015 number of constituent vector and predicate registers.
4016 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4017 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4018 instead of aarch64_sve_argument_p.
4019 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4020 (aarch64_pass_by_reference): Likewise.
4021 (aarch64_function_value_1): Likewise.
4022 (aarch64_return_in_memory): Likewise.
4023 (aarch64_layout_arg): Likewise.
4025 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4027 PR tree-optimization/93156
4028 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4029 least significant bit is always clear.
4031 PR tree-optimization/93118
4032 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4033 simplifier with two intermediate conversions.
4035 2020-01-07 Martin Liska <mliska@suse.cz>
4037 * params.opt: Add Optimization for various parameters.
4039 2020-01-07 Martin Liska <mliska@suse.cz>
4042 * doc/extend.texi: Explain cloning for target_clone
4045 2020-01-07 Martin Liska <mliska@suse.cz>
4047 PR tree-optimization/92860
4048 * common.opt: Make in Optimization option
4049 as it is affected by -O0, which is an Optimization
4051 * tree-inline.c (tree_inlinable_function_p):
4052 Use opt_for_fn for warn_inline.
4053 (expand_call_inline): Likewise.
4055 2020-01-07 Martin Liska <mliska@suse.cz>
4057 PR tree-optimization/92860
4058 * common.opt: Make flag_ree as optimization
4061 2020-01-07 Martin Liska <mliska@suse.cz>
4063 PR optimization/92860
4064 * params.opt: Mark param_min_crossjump_insns with Optimization
4067 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4069 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4070 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4072 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4074 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4075 helper function to return the valid addressing formats for a given
4076 hard register and mode.
4077 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4079 * config/rs6000/constraints.md (Q constraint): Update
4081 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4084 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4085 Use 'Q' for doing vector extract from memory.
4086 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4088 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4089 doing vector extract from memory.
4090 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4091 extract from memory.
4093 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4094 for the offset being 34-bits when -mcpu=future is used.
4096 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4098 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4099 instead of cmpib_comparison_operator in cmpib patterns.
4100 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4101 of cmpib_comparison_operator. Revise comment.
4103 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4105 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4106 in an IFN_DIV_POW2 node to be equal.
4108 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4110 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4111 (vect_check_scalar_mask): ...this.
4112 (vectorizable_store, vectorizable_load): Update call accordingly.
4113 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4114 argument in calls to conditional internal functions.
4116 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4118 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4119 '0' matching inputs.
4120 (subv64di3_exec): Likewise.
4122 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4124 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4125 * doc/md.texi (movstr): Likewise.
4127 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4129 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4132 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4134 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4136 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4137 to a temporary file and use move-if-change to update the real
4138 file where necessary.
4140 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4142 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4143 rather than Upa for CPY /M.
4145 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4147 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4150 2020-01-06 Martin Liska <mliska@suse.cz>
4152 PR tree-optimization/92860
4153 * params.opt: Mark param_max_combine_insns with Optimization
4156 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4159 * config/i386/i386.md (SWIDWI): New mode iterator.
4160 (DWI, dwi): Add TImode variants.
4161 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4162 <general_hilo_operand> instead of <general_operand>. Use
4163 CONST_SCALAR_INT_P instead of CONST_INT_P.
4164 (*addv<mode>4_1): Rename to ...
4165 (addv<mode>4_1): ... this.
4166 (QWI): New mode attribute.
4167 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4168 define_insn_and_split patterns.
4169 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4171 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4172 <general_hilo_operand> instead of <general_operand>.
4173 (*addcarry<mode>_1): New define_insn.
4174 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4176 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4178 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4179 Use "call" instead of "set".
4181 2020-01-03 Martin Jambor <mjambor@suse.cz>
4184 * ipa-cp.c (print_all_lattices): Skip functions without info.
4186 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4189 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4190 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4191 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4192 for 'e' simd clones.
4195 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4197 (mprefer-vector-width=): Add Save.
4198 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4199 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4200 (ix86_debug_options, ix86_function_specific_print): Adjust
4201 ix86_target_string callers.
4202 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4203 (ix86_valid_target_attribute_tree): Likewise.
4204 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4205 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4206 ix86_target_string caller.
4209 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4210 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4211 instead of gen_int_shift_amount + convert_modes.
4213 PR rtl-optimization/93088
4214 * loop-iv.c (find_single_def_src): Punt after looking through
4215 128 reg copies for regs with single definitions. Move definitions
4218 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4220 * config/arm/arm-c.c (arm_cpu_builtins): Define
4221 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4222 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4223 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4224 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4225 * config/arm/arm-tables.opt: Regenerated.
4226 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4227 arm_arch_i8mm and arm_arch_bf16 when enabled.
4228 * config/arm/arm.h (TARGET_I8MM): New macro.
4229 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4230 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4231 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4232 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4233 (v8_6_a_simd_variants): New.
4234 (v8_*_a_simd_variants): Add i8mm and bf16.
4235 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4237 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4240 * predict.c (compute_function_frequency): Don't call
4241 warn_function_cold on functions that already have cold attribute.
4243 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4246 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4247 COMDAT group function labels in .data.rel.ro.local section.
4248 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4251 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4252 comparison_operator in B and S integer comparisons. Likewise, use
4253 ordered_comparison_operator instead of cmpib_comparison_operator in
4255 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4257 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4259 Update copyright years.
4261 * gcc.c (process_command): Update copyright notice dates.
4262 * gcov-dump.c (print_version): Ditto.
4263 * gcov.c (print_version): Ditto.
4264 * gcov-tool.c (print_version): Ditto.
4265 * gengtype.c (create_file): Ditto.
4266 * doc/cpp.texi: Bump @copying's copyright year.
4267 * doc/cppinternals.texi: Ditto.
4268 * doc/gcc.texi: Ditto.
4269 * doc/gccint.texi: Ditto.
4270 * doc/gcov.texi: Ditto.
4271 * doc/install.texi: Ditto.
4272 * doc/invoke.texi: Ditto.
4274 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4276 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4279 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4281 PR tree-optimization/93098
4282 * match.pd (popcount): For shift amounts, use integer_onep
4283 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4284 tests. Make sure that precision is power of two larger than or equal
4285 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4286 instead of ULL suffixed constants. Formatting fixes.
4288 Copyright (C) 2020 Free Software Foundation, Inc.
4290 Copying and distribution of this file, with or without modification,
4291 are permitted in any medium without royalty provided the copyright
4292 notice and this notice are preserved.