1 2020-02-18 Martin Liska <mliska@suse.cz>
3 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
4 bootstrap by missing removal of invalid sanity check.
6 2020-02-18 Martin Liska <mliska@suse.cz>
9 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
10 Always compare LHS of gimple_assign.
12 2020-02-18 Martin Liska <mliska@suse.cz>
15 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
16 and return type of functions.
17 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
18 Drop MALLOC attribute for void functions.
19 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
20 malloc_state for a new VOID clone.
22 2020-02-18 Martin Liska <mliska@suse.cz>
25 * common.opt: Add -fprofile-reproducibility.
26 * doc/invoke.texi: Document it.
27 * value-prof.c (dump_histogram_value):
28 Document and support behavior for counters[0]
29 being a negative value.
30 (get_nth_most_common_value): Handle negative
31 counters[0] in respect to flag_profile_reproducible.
33 2020-02-18 Jakub Jelinek <jakub@redhat.com>
36 * cgraph.c (verify_speculative_call): Use speculative_id instead of
37 speculative_uid in messages. Remove trailing whitespace from error
38 message. Use num_speculative_call_targets instead of
39 num_speculative_targets in a message.
40 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
41 edge messages and stmt instead of cal_stmt in reference message.
43 PR tree-optimization/93780
44 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
45 before calling build_vector_type.
46 (execute_update_addresses_taken): Likewise.
49 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
50 typo, functoin -> function.
51 * tree.c (free_lang_data_in_decl): Fix comment typo,
53 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
55 2020-02-17 David Malcolm <dmalcolm@redhat.com>
57 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
59 (print_option_information): Don't call get_option_url if URLs
62 2020-02-17 Alexandre Oliva <oliva@adacore.com>
64 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
65 handling of register_common-less targets.
67 2020-02-17 Martin Liska <mliska@suse.cz>
70 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
72 2020-02-17 Martin Liska <mliska@suse.cz>
75 * config/rs6000/rs6000.c (rs6000_option_override_internal):
78 2020-02-17 Martin Liska <mliska@suse.cz>
81 * config/rx/elf.opt: Fix typo.
83 2020-02-17 Richard Biener <rguenther@suse.de>
86 * opts-global.c (print_ignored_options): Use inform and
89 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
92 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
94 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
97 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
98 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
100 2020-02-15 Jason Merrill <jason@redhat.com>
102 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
104 2020-02-15 Jakub Jelinek <jakub@redhat.com>
106 PR tree-optimization/93744
107 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
108 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
109 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
110 sure @2 in the first and @1 in the other patterns has no side-effects.
112 2020-02-15 David Malcolm <dmalcolm@redhat.com>
113 Bernd Edlinger <bernd.edlinger@hotmail.de>
117 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
118 * configure.ac (--with-diagnostics-urls): New configuration
119 option, based on --with-diagnostics-color.
120 (DIAGNOSTICS_URLS_DEFAULT): New define.
121 * config.h: Regenerate.
122 * configure: Regenerate.
123 * diagnostic.c (diagnostic_urls_init): Handle -1 for
124 DIAGNOSTICS_URLS_DEFAULT from configure-time
125 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
126 and TERM_URLS environment variable.
127 * diagnostic-url.h (diagnostic_url_format): New enum type.
128 (diagnostic_urls_enabled_p): rename to...
129 (determine_url_format): ... this, and change return type.
130 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
131 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
132 the linux console, and mingw.
133 (diagnostic_urls_enabled_p): rename to...
134 (determine_url_format): ... this, and adjust.
135 * pretty-print.h (pretty_printer::show_urls): rename to...
136 (pretty_printer::url_format): ... this, and change to enum.
137 * pretty-print.c (pretty_printer::pretty_printer,
138 pp_begin_url, pp_end_url, test_urls): Adjust.
139 * doc/install.texi (--with-diagnostics-urls): Document the new
140 configuration option.
141 (--with-diagnostics-color): Document the existing interaction
142 with GCC_COLORS better.
143 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
144 vindex reference. Update description of defaults based on the above.
145 (-fdiagnostics-color): Update description of how -fdiagnostics-color
146 interacts with GCC_COLORS.
148 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
151 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
152 conjunction with TARGET_GNU_TLS in early return.
154 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
156 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
157 the mode is not wider than UNITS_PER_WORD.
159 2020-02-14 Martin Jambor <mjambor@suse.cz>
161 PR tree-optimization/93516
162 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
163 access of the same type as the parent.
164 (propagate_subaccesses_from_lhs): Likewise.
166 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
169 * config/i386/avx512vbmi2intrin.h
170 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
171 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
172 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
173 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
174 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
175 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
176 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
177 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
178 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
179 of lacking a closing parenthesis.
180 * config/i386/avx512vbmi2vlintrin.h
181 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
182 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
183 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
184 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
185 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
186 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
187 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
188 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
189 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
190 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
191 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
192 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
193 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
194 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
195 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
196 _mm_shldi_epi32, _mm_mask_shldi_epi32,
197 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
198 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
200 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
203 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
204 the target function entry.
206 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
208 * common/config/arc/arc-common.c (arc_option_optimization_table):
209 Disable if-conversion step when optimized for size.
211 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
213 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
214 R12-R15 are always in ARCOMPACT16_REGS register class.
215 * config/arc/arc.opt (mq-class): Deprecate.
216 * config/arc/constraint.md ("q"): Remove dependency on mq-class
218 * doc/invoke.texi (mq-class): Update text.
219 * common/config/arc/arc-common.c (arc_option_optimization_table):
222 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
224 * config/arc/arc.c (arc_insn_cost): New function.
225 (TARGET_INSN_COST): Define.
226 * config/arc/arc.md (cost): New attribute.
227 (add_n): Use arc_nonmemory_operand.
228 (ashlsi3_insn): Likewise, also update constraints.
229 (ashrsi3_insn): Likewise.
231 (add_shift): Likewise.
232 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
234 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
236 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
238 (umulsidi_600): Likewise.
240 2020-02-13 Jakub Jelinek <jakub@redhat.com>
243 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
244 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
245 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
246 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
247 pass __A to the builtin followed by __W instead of __A followed by
249 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
250 _mm512_mask_popcnt_epi64): Likewise.
251 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
252 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
253 _mm256_mask_popcnt_epi64): Likewise.
255 PR tree-optimization/93582
256 * fold-const.h (shift_bytes_in_array_left,
257 shift_bytes_in_array_right): Declare.
258 * fold-const.c (shift_bytes_in_array_left,
259 shift_bytes_in_array_right): New function, moved from
260 gimple-ssa-store-merging.c, no longer static.
261 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
262 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
263 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
264 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
265 shift_bytes_in_array.
266 (verify_shift_bytes_in_array): Rename to ...
267 (verify_shift_bytes_in_array_left): ... this. Use
268 shift_bytes_in_array_left instead of shift_bytes_in_array.
269 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
270 instead of verify_shift_bytes_in_array.
271 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
272 / native_interpret_expr where the store covers all needed bits,
273 punt on PDP-endian, otherwise allow all involved offsets and sizes
274 not to be byte-aligned.
277 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
278 use const_0_to_255_operand predicate instead of immediate_operand.
279 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
280 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
281 vgf2p8affineinvqb_<mode><mask_name>,
282 vgf2p8affineqb_<mode><mask_name>): Drop mode from
283 const_0_to_255_operand predicated operands.
285 2020-02-12 Jeff Law <law@redhat.com>
287 * config/h8300/h8300.md (comparison shortening peepholes): Use
288 a mode iterator to merge the HImode and SImode peepholes.
290 2020-02-12 Jakub Jelinek <jakub@redhat.com>
293 * real.c (is_even): Make static. Function comment fix.
294 (is_halfway_below): Make static, don't assert R is not inf/nan,
295 instead return false for those. Small formatting fixes.
297 2020-02-12 Martin Sebor <msebor@redhat.com>
300 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
301 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
302 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
303 (strlen_check_and_optimize_call): Adjust callee name.
305 2020-02-12 Jeff Law <law@redhat.com>
307 * config/h8300/h8300.md (comparison shortening peepholes): Drop
308 (and (xor)) variant. Combine other two into single peephole.
310 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
312 PR rtl-optimization/93565
313 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
315 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
317 * config/aarch64/aarch64-simd.md
318 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
319 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
320 generating separate ADDV and zero_extend patterns.
321 * config/aarch64/iterators.md (VDQV_E): New iterator.
323 2020-02-12 Jeff Law <law@redhat.com>
325 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
326 expanders, splits, etc.
327 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
328 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
329 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
330 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
331 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
332 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
334 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
336 2020-02-12 Jakub Jelinek <jakub@redhat.com>
339 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
340 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
341 TARGET_AVX512DQ from condition.
342 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
343 instead of <mask_mode512bit_condition> in condition. If
344 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
346 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
349 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
352 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
354 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
356 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
357 where strlen is more legible.
358 (rs6000_builtin_vectorized_libmass): Ditto.
359 (rs6000_print_options_internal): Ditto.
361 2020-02-11 Martin Sebor <msebor@redhat.com>
363 PR tree-optimization/93683
364 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
366 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
368 * config/rs6000/predicates.md (cint34_operand): Rename the
369 -mprefixed-addr option to be -mprefixed.
370 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
371 the -mprefixed-addr option to be -mprefixed.
372 (OTHER_FUTURE_MASKS): Likewise.
373 (POWERPC_MASKS): Likewise.
374 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
375 the -mprefixed-addr option to be -mprefixed. Change error
376 messages to refer to -mprefixed.
377 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
379 (rs6000_legitimate_offset_address_p): Likewise.
380 (rs6000_mode_dependent_address): Likewise.
381 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
382 "-mprefixed" for target attributes and pragmas.
383 (address_to_insn_form): Rename the -mprefixed-addr option to be
385 (rs6000_adjust_insn_length): Likewise.
386 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
387 -mprefixed-addr option to be -mprefixed.
388 (ASM_OUTPUT_OPCODE): Likewise.
389 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
390 -mprefixed-addr option to be -mprefixed.
391 * config/rs6000/rs6000.opt (-mprefixed): Rename the
392 -mprefixed-addr option to be prefixed. Change the option from
393 being undocumented to being documented.
394 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
395 -mprefixed option. Update the -mpcrel documentation to mention
398 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
400 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
401 including FIRST_PSEUDO_REGISTER - 1.
402 * ira-color.c (print_hard_reg_set): Ditto.
404 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
406 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
407 (USTERNOP_QUALIFIERS): New define.
408 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
409 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
410 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
411 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
412 * config/arm/arm_neon.h (vusdot_s32): New.
413 (vusdot_lane_s32): New.
414 (vusdotq_lane_s32): New.
415 (vsudot_lane_s32): New.
416 (vsudotq_lane_s32): New.
417 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
418 * config/arm/iterators.md (DOTPROD_I8MM): New.
419 (sup, opsuffix): Add <us/su>.
420 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
421 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
423 2020-02-11 Richard Biener <rguenther@suse.de>
425 PR tree-optimization/93661
426 PR tree-optimization/93662
427 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
429 * tree-sra.c (get_access_for_expr): Likewise.
431 2020-02-10 Jakub Jelinek <jakub@redhat.com>
434 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
435 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
436 Change condition from TARGET_AVX2 to TARGET_AVX.
438 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
441 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
444 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
446 Try to generate zero-based comparisons.
447 * config/cris/cris.c (cris_reduce_compare): New function.
448 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
449 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
450 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
452 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
455 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
456 in Thumb state and also as a destination in Arm state. Add T16
459 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
461 * md.texi (Define Subst): Match closing paren in example.
463 2020-02-10 Jakub Jelinek <jakub@redhat.com>
467 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
468 arguments of strncmp.
470 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
473 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
474 but different source value.
475 (adjust_callers_for_value_intersection): New function.
476 (gather_edges_for_value): Adjust order of callers to let a
477 non-self-recursive caller be the first element.
478 (self_recursive_pass_through_p): Add a new parameter "simple", and
479 check generalized self-recursive pass-through jump function.
480 (self_recursive_agg_pass_through_p): Likewise.
481 (find_more_scalar_values_for_callers_subset): Compute value from
482 pass-through jump function for self-recursive.
483 (intersect_with_plats): Cleanup previous implementation code for value
484 itersection with self-recursive call edge.
485 (intersect_with_agg_replacements): Likewise.
486 (intersect_aggregates_with_edge): Deduce value from pass-through jump
487 function for self-recursive call edge. Cleanup previous implementation
488 code for value intersection with self-recursive call edge.
489 (decide_whether_version_node): Remove dead callers and adjust order
490 to let a non-self-recursive caller be the first element.
492 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
494 * recog.c: Move pass_split_before_sched2 code in front of
495 pass_split_before_regstack.
496 (pass_data_split_before_sched2): Rename pass to split3 from split4.
497 (pass_data_split_before_regstack): Rename pass to split4 from split3.
498 (rest_of_handle_split_before_sched2): Remove.
499 (pass_split_before_sched2::execute): Unconditionally call
501 (enable_split_before_sched2): New function.
502 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
503 (pass_split_before_regstack::gate): Ditto.
504 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
505 Update name check for renamed split4 pass.
506 * config/sh/sh.c (register_sh_passes): Update pass insertion
507 point for renamed split4 pass.
509 2020-02-09 Jakub Jelinek <jakub@redhat.com>
511 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
512 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
513 copying them around between host and target.
515 2020-02-08 Andrew Pinski <apinski@marvell.com>
518 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
519 STRICT_ALIGNMENT also.
521 2020-02-08 Jim Wilson <jimw@sifive.com>
524 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
526 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
527 Jakub Jelinek <jakub@redhat.com>
530 * config/i386/i386.h (CALL_USED_REGISTERS): Make
531 xmm16-xmm31 call-used even in 64-bit ms-abi.
533 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
535 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
536 (simd_ummla, simd_usmmla): Likewise.
537 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
538 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
541 2020-02-07 Richard Biener <rguenther@suse.de>
544 * tree-inline.c (fold_marked_statements): Do a PRE walk,
545 skipping unreachable regions.
546 (optimize_inline_calls): Skip folding stmts when we didn't
549 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
552 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
553 Don't return aggregates with only SFmode and DFmode in SSE
555 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
557 2020-02-07 Jakub Jelinek <jakub@redhat.com>
560 * config/rs6000/rs6000-logue.c
561 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
562 if it fails, move rs into end_addr and retry. Add
563 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
564 the insn pattern doesn't describe well what exactly happens to
568 * config/i386/predicates.md (avx_identity_operand): Remove.
569 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
570 (avx_<castmode><avxsizesuffix>_<castmode>,
571 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
572 a VEC_CONCAT of the operand and UNSPEC_CAST.
573 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
574 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
578 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
579 recog_data.insn if distance_non_agu_define changed it.
581 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
584 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
585 we only had X-FORM (reg+reg) addressing for vectors. Also before
586 ISA 3.0, we only had X-FORM addressing for scalars in the
587 traditional Altivec registers.
589 2020-02-06 <zhongyunde@huawei.com>
590 Vladimir Makarov <vmakarov@redhat.com>
592 PR rtl-optimization/93561
593 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
596 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
598 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
601 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
603 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
604 where the low and the high 32 bits are equal to each other specially,
605 with an rldimi instruction.
607 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
609 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
611 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
613 * config/arm/arm-tables.opt: Regenerate.
615 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
618 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
619 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
620 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
622 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
624 PR rtl-optimization/87763
625 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
627 2020-02-06 Delia Burduv <delia.burduv@arm.com>
629 * config/aarch64/aarch64-simd-builtins.def
630 (bfmlaq): New built-in function.
631 (bfmlalb): New built-in function.
632 (bfmlalt): New built-in function.
633 (bfmlalb_lane): New built-in function.
634 (bfmlalt_lane): New built-in function.
635 * config/aarch64/aarch64-simd.md
636 (aarch64_bfmmlaqv4sf): New pattern.
637 (aarch64_bfmlal<bt>v4sf): New pattern.
638 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
639 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
640 (vbfmlalbq_f32): New intrinsic.
641 (vbfmlaltq_f32): New intrinsic.
642 (vbfmlalbq_lane_f32): New intrinsic.
643 (vbfmlaltq_lane_f32): New intrinsic.
644 (vbfmlalbq_laneq_f32): New intrinsic.
645 (vbfmlaltq_laneq_f32): New intrinsic.
646 * config/aarch64/iterators.md (BF_MLA): New int iterator.
647 (bt): New int attribute.
649 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
651 * config/i386/i386.md (*pushtf): Emit "#" instead of
652 calling gcc_unreachable in insn output.
655 (*pushsf_rex64): Ditto for alternatives other than 1.
656 (*pushsf): Ditto for alternatives other than 1.
658 2020-02-06 Martin Liska <mliska@suse.cz>
660 PR gcov-profile/91971
661 PR gcov-profile/93466
662 * coverage.c (coverage_init): Revert mangling of
663 path into filename. It can lead to huge filename length.
664 Creation of subfolders seem more natural.
666 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
669 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
670 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
671 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
673 2020-02-06 Jakub Jelinek <jakub@redhat.com>
676 * config/i386/predicates.md (avx_identity_operand): New predicate.
677 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
678 define_insn_and_split.
681 * omp-low.c (use_pointer_for_field): For nested constructs, also
682 look for map clauses on target construct.
683 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
684 taskreg_nesting_level.
687 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
688 shared clause, call omp_notice_variable on outer context if any.
690 2020-02-05 Jason Merrill <jason@redhat.com>
693 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
694 non-zero address even if weak and not yet defined.
696 2020-02-05 Martin Sebor <msebor@redhat.com>
698 PR tree-optimization/92765
699 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
700 * tree-ssa-strlen.c (compute_string_length): Remove.
701 (determine_min_objsize): Remove.
702 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
703 Avoid using type size as the upper bound on string length.
704 (handle_builtin_string_cmp): Add an argument. Adjust.
705 (strlen_check_and_optimize_call): Pass additional argument to
706 handle_builtin_string_cmp.
708 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
710 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
711 (*pushdi2_rex64 peephole2): Unconditionally split after
713 (*ashl<mode>3_doubleword): Ditto.
714 (*<shift_insn><mode>3_doubleword): Ditto.
716 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
719 * config/rs6000/rs6000.c (get_vector_offset): Fix
721 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
723 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
725 2020-02-05 David Malcolm <dmalcolm@redhat.com>
728 (Special Functions for Debugging the Analyzer): Update description
729 of __analyzer_dump_exploded_nodes.
731 2020-02-05 Jakub Jelinek <jakub@redhat.com>
734 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
735 include sets and not clobbers in the vzeroupper pattern.
736 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
737 the parallel has 17 (64-bit) or 9 (32-bit) elts.
738 (*avx_vzeroupper_1): New define_insn_and_split.
741 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
742 don't run when !optimize.
743 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
746 2020-02-05 Richard Biener <rguenther@suse.de>
749 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
750 checks before matching calls.
752 2020-02-05 Jakub Jelinek <jakub@redhat.com>
754 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
755 function comment typo.
758 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
759 simd_clone_create failed when i == 0, adjust clone->nargs by
762 2020-02-05 Martin Liska <mliska@suse.cz>
765 * doc/invoke.texi: Document that one should
766 not combine ASLR and -fpch.
768 2020-02-04 Richard Biener <rguenther@suse.de>
770 PR tree-optimization/93538
771 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
773 2020-02-04 Richard Biener <rguenther@suse.de>
775 PR tree-optimization/91123
776 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
777 (vn_walk_cb_data::last_vuse): New member.
778 (vn_walk_cb_data::saved_operands): Likewsie.
779 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
780 (vn_walk_cb_data::push_partial_def): Use finish.
781 (vn_reference_lookup_2): Update last_vuse and use finish if
782 we've saved operands.
783 (vn_reference_lookup_3): Use finish and update calls to
784 push_partial_defs everywhere. When translating through
785 memcpy or aggregate copies save off operands and alias-set.
786 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
787 operation for redundant store removal.
789 2020-02-04 Richard Biener <rguenther@suse.de>
791 PR tree-optimization/92819
792 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
793 generating more stmts than before.
795 2020-02-04 Martin Liska <mliska@suse.cz>
797 * config/arm/arm.c (arm_gen_far_branch): Move the function
798 outside of selftests.
800 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
802 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
803 function to adjust PC-relative vector addresses.
804 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
805 handle vectors with PC-relative addresses.
807 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
809 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
811 (hard_reg_and_mode_to_addr_mask): Delete.
812 (rs6000_adjust_vec_address): If the original vector address
813 was REG+REG or REG+OFFSET and the element is not zero, do the add
814 of the elements in the original address before adding the offset
815 for the vector element. Use address_to_insn_form to validate the
816 address using the register being loaded, rather than guessing
817 whether the address is a DS-FORM or DQ-FORM address.
819 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
821 * config/rs6000/rs6000.c (get_vector_offset): New helper function
822 to calculate the offset in memory from the start of a vector of a
823 particular element. Add code to keep the element number in
824 bounds if the element number is variable.
825 (rs6000_adjust_vec_address): Move calculation of offset of the
826 vector element to get_vector_offset.
827 (rs6000_split_vec_extract_var): Do not do the initial AND of
828 element here, move the code to get_vector_offset.
830 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
832 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
835 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
837 * config/rs6000/constraints.md: Improve documentation.
839 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
842 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
843 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
845 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
847 * config.gcc: Remove "carrizo" support.
848 * config/gcn/gcn-opts.h (processor_type): Likewise.
849 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
850 * config/gcn/gcn.opt (gpu_type): Likewise.
851 * config/gcn/t-omp-device: Likewise.
853 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
856 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
857 * config/arm/arm.c (arm_gen_far_branch): New function
859 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
861 2020-02-03 Julian Brown <julian@codesourcery.com>
862 Tobias Burnus <tobias@codesourcery.com>
864 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
866 2020-02-03 Jakub Jelinek <jakub@redhat.com>
869 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
870 valid RTL to sum up the lowest and second lowest bytes of the popcnt
873 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
875 PR rtl-optimization/91333
876 * ira-color.c (struct allocno_color_data): Add member
878 (init_allocno_threads): Set the member up.
879 (bucket_allocno_compare_func): Add compare hard reg
882 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
884 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
886 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
887 * config.in: Regenerated.
888 * configure: Regenerated.
889 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
890 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
891 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
893 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
895 * configure: Regenerate.
897 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
899 PR rtl-optimization/91333
900 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
901 reg preferences comparison up.
903 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
905 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
906 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
907 aarch64-sve-builtins-base.h.
908 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
909 aarch64-sve-builtins-base.cc.
910 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
911 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
913 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
914 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
915 (svcvtnt): New functions.
916 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
917 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
918 (svcvtnt): New functions.
919 (svcvt): Add a form that converts f32 to bf16.
920 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
921 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
923 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
924 Treat B as bfloat16_t.
925 (ternary_bfloat_lane_base): New class.
926 (ternary_bfloat_def): Likewise.
927 (ternary_bfloat): New shape.
928 (ternary_bfloat_lane_def): New class.
929 (ternary_bfloat_lane): New shape.
930 (ternary_bfloat_lanex2_def): New class.
931 (ternary_bfloat_lanex2): New shape.
932 (ternary_bfloat_opt_n_def): New class.
933 (ternary_bfloat_opt_n): New shape.
934 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
935 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
936 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
937 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
938 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
939 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
940 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
941 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
942 the pattern off the narrow mode instead of the wider one.
943 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
944 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
945 (sve_fp_op): Handle them.
946 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
947 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
949 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
951 * config/aarch64/arm_sve.h: Include arm_bf16.h.
952 * config/aarch64/aarch64-modes.def (BF): Move definition before
953 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
954 (SVE_MODES): Handle BF modes.
955 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
957 (aarch64_full_sve_mode): Likewise.
958 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
960 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
961 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
962 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
963 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
965 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
967 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
968 (TYPES_all_data): Add bf16.
969 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
970 (register_tuple_type): Increase buffer size.
971 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
972 (bf16): New type suffix.
973 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
974 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
975 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
976 Change type from all_data to all_arith.
977 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
980 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
981 Matthew Malcomson <matthew.malcomson@arm.com>
982 Richard Sandiford <richard.sandiford@arm.com>
984 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
985 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
986 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
987 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
988 __ARM_FEATURE_MATMUL_FP64.
989 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
990 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
991 be disabled at the same time.
992 (f32mm): New extension.
993 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
994 (AARCH64_FL_F64MM): Bump to the next bit up.
995 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
996 (TARGET_SVE_F64MM): New macros.
997 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
998 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
999 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1000 (UNSPEC_ZIP2Q): New unspeccs.
1001 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1002 (optab, sur, perm_insn): Handle the new unspecs.
1003 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1004 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1005 TARGET_SVE_F64MM instead of separate tests.
1006 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1007 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1008 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1009 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1010 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1011 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1012 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1013 (TYPES_s_signed): New macro.
1014 (TYPES_s_integer): Use it.
1015 (TYPES_d_float): New macro.
1016 (TYPES_d_data): Use it.
1017 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1018 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1019 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1020 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1021 (svmmla): New shape.
1022 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1023 template parameters.
1024 (ternary_resize2_lane_base): Likewise.
1025 (ternary_resize2_base): New class.
1026 (ternary_qq_lane_base): Likewise.
1027 (ternary_intq_uintq_lane_def): Likewise.
1028 (ternary_intq_uintq_lane): New shape.
1029 (ternary_intq_uintq_opt_n_def): New class
1030 (ternary_intq_uintq_opt_n): New shape.
1031 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1032 (ternary_uintq_intq_def): New class.
1033 (ternary_uintq_intq): New shape.
1034 (ternary_uintq_intq_lane_def): New class.
1035 (ternary_uintq_intq_lane): New shape.
1036 (ternary_uintq_intq_opt_n_def): New class.
1037 (ternary_uintq_intq_opt_n): New shape.
1038 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1039 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1040 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1041 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1043 (svdotprod_lane_impl): ...this new class.
1044 (svmmla_impl, svusdot_impl): New classes.
1045 (svdot_lane): Update to use svdotprod_lane_impl.
1046 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1047 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1049 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1050 function, with no types defined.
1051 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1052 AARCH64_FL_I8MM functions.
1053 (svmmla): New AARCH64_FL_F32MM function.
1054 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1055 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1056 AARCH64_FL_F64MM function.
1057 (REQUIRED_EXTENSIONS):
1059 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1061 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1064 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1066 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1067 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1068 (*movti_internal): Do not check for
1069 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1070 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1071 just after check for TARGET_AVX.
1072 (*movdf_internal): Ditto.
1073 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1074 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1075 * config/i386/sse.md (mov<mode>_internal): Only check
1076 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1077 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1078 (<sse>_andnot<mode>3<mask_name>): Move check for
1079 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1080 (<code><mode>3<mask_name>): Ditto.
1081 (*andnot<mode>3): Ditto.
1082 (*andnottf3): Ditto.
1083 (*<code><mode>3): Ditto.
1084 (*<code>tf3): Ditto.
1085 (*andnot<VI:mode>3): Remove
1086 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1087 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1088 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1089 (sse4_1_blendv<ssemodesuffix>): Ditto.
1090 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1091 Explain that tune applies to 128bit instructions only.
1093 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1095 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1096 to definition of hsa_kernel_description. Parse assembly to find SGPR
1097 and VGPR count of kernel and store in hsa_kernel_description.
1099 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1101 PR rtl-optimization/91838
1102 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1103 to truncate if allowed or reject combination.
1105 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1107 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1108 (find_inv_vars_cb): Likewise.
1110 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1112 * calls.c (special_function_p): Split out the check for DECL_NAME
1113 being non-NULL and fndecl being extern at file scope into a
1114 new maybe_special_function_p and call it. Drop check for fndecl
1115 being non-NULL that was after a usage of DECL_NAME (fndecl).
1116 * tree.h (maybe_special_function_p): New inline function.
1118 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1120 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1121 (mask_gather_load<mode>): ... here, and zero-initialize the
1123 (maskload<mode>di): Zero-initialize the destination.
1126 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1129 * doc/analyzer.texi (Limitations): Note that constraints on
1130 floating-point values are currently ignored.
1132 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1135 * symtab.c (symtab_node::noninterposable_alias): If localalias
1136 already exists, but is not usable, append numbers after it until
1137 a unique name is found. Formatting fix.
1140 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1143 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1145 * config/gcn/gcn.c (print_operand): Handle LTGT.
1146 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1148 2020-01-30 Richard Biener <rguenther@suse.de>
1150 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1151 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1153 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1155 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1156 without a DECL in .data.rel.ro.local.
1158 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1161 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1165 * config/i386/sse.md
1166 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1167 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1168 any_extend code iterator instead of always zero_extend.
1169 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1170 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1171 Use any_extend code iterator instead of always zero_extend.
1172 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1173 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1174 Use any_extend code iterator instead of always zero_extend.
1175 (*sse2_pmovmskb_ext): New define_insn.
1176 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1179 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1180 (*popcountsi2_zext_falsedep): New define_insn.
1182 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1184 * config.in: Regenerated.
1185 * configure: Regenerated.
1187 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1190 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1191 LLVM's assembler changed the default in version 9.
1193 2020-01-24 Jeff Law <law@redhat.com>
1195 PR tree-optimization/89689
1196 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1198 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1202 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1204 PR rtl-optimization/87763
1205 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1206 simplification to handle subregs as well as bare regs.
1207 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1209 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1212 * ira.c (ira): Revert use of simplified LRA algorithm.
1214 2020-01-29 Martin Jambor <mjambor@suse.cz>
1216 PR tree-optimization/92706
1217 * tree-sra.c (struct access): Fields first_link, last_link,
1218 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1219 next_rhs_queued and grp_rhs_queued respectively, new fields
1220 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1221 (struct assign_link): Field next renamed to next_rhs, new field
1222 next_lhs. Updated comment.
1223 (work_queue_head): Renamed to rhs_work_queue_head.
1224 (lhs_work_queue_head): New variable.
1225 (add_link_to_lhs): New function.
1226 (relink_to_new_repr): Also relink LHS lists.
1227 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1228 (add_access_to_lhs_work_queue): New function.
1229 (pop_access_from_work_queue): Renamed to
1230 pop_access_from_rhs_work_queue.
1231 (pop_access_from_lhs_work_queue): New function.
1232 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1234 (child_would_conflict_in_lacc): Renamed to
1235 child_would_conflict_in_acc. Adjusted parameter names.
1236 (create_artificial_child_access): New parameter set_grp_read, use it.
1237 (subtree_mark_written_and_enqueue): Renamed to
1238 subtree_mark_written_and_rhs_enqueue.
1239 (propagate_subaccesses_across_link): Renamed to
1240 propagate_subaccesses_from_rhs.
1241 (propagate_subaccesses_from_lhs): New function.
1242 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1245 2020-01-29 Martin Jambor <mjambor@suse.cz>
1247 PR tree-optimization/92706
1248 * tree-sra.c (struct access): Adjust comment of
1249 grp_total_scalarization.
1250 (find_access_in_subtree): Look for single children spanning an entire
1252 (scalarizable_type_p): Allow register accesses, adjust callers.
1253 (completely_scalarize): Remove function.
1254 (scalarize_elem): Likewise.
1255 (create_total_scalarization_access): Likewise.
1256 (sort_and_splice_var_accesses): Do not track total scalarization
1258 (analyze_access_subtree): New parameter totally, adjust to new meaning
1259 of grp_total_scalarization.
1260 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1261 (can_totally_scalarize_forest_p): New function.
1262 (create_total_scalarization_access): Likewise.
1263 (create_total_access_and_reshape): Likewise.
1264 (total_should_skip_creating_access): Likewise.
1265 (totally_scalarize_subtree): Likewise.
1266 (analyze_all_variable_accesses): Perform total scalarization after
1267 subaccess propagation using the new functions above.
1268 (initialize_constant_pool_replacements): Output initializers by
1269 traversing the access tree.
1271 2020-01-29 Martin Jambor <mjambor@suse.cz>
1273 * tree-sra.c (verify_sra_access_forest): New function.
1274 (verify_all_sra_access_forests): Likewise.
1275 (create_artificial_child_access): Set parent.
1276 (analyze_all_variable_accesses): Call the verifier.
1278 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1280 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1281 if called on indirect edge.
1282 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1283 speculative call if needed.
1285 2020-01-29 Richard Biener <rguenther@suse.de>
1287 PR tree-optimization/93428
1288 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1289 permutation when the load node is created.
1290 (vect_analyze_slp_instance): Re-use it here.
1292 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1294 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1296 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1298 PR rtl-optimization/93272
1299 * ira-lives.c (process_out_of_region_eh_regs): New function.
1300 (process_bb_node_lives): Call it.
1302 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1304 * coverage.c (read_counts_file): Make error message lowercase.
1306 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1308 * profile-count.c (profile_quality_display_names): Fix ordering.
1310 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1313 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1314 hash only when edge is first within the sequence.
1315 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1316 (symbol_table::create_edge): Do not set target_prob.
1317 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1319 (cgraph_edge::make_speculative): Drop target_prob parameter.
1320 (cgraph_edge::speculative_call_info): Remove.
1321 (cgraph_edge::first_speculative_call_target): New member function.
1322 (update_call_stmt_hash_for_removing_direct_edge): New function.
1323 (cgraph_edge::resolve_speculation): Rewrite to new API.
1324 (cgraph_edge::speculative_call_for_target): New member function.
1325 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1326 multiple speculation targets.
1327 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1329 (verify_speculative_call): Verify that targets form an interval.
1330 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1331 (cgraph_edge::first_speculative_call_target): New member function.
1332 (cgraph_edge::next_speculative_call_target): New member function.
1333 (cgraph_edge::speculative_call_target_ref): New member function.
1334 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1335 (cgraph_edge): Remove target_prob.
1336 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1337 Fix handling of speculative calls.
1338 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1339 * ipa-fnsummary.c (analyze_function_body): Likewise.
1340 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1341 * ipa-profile.c (dump_histogram): Fix formating.
1342 (ipa_profile_generate_summary): Watch for overflows.
1343 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1344 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1345 (update_indirect_edges_after_inlining): Update to new API.
1346 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1348 * profile-count.h: (profile_probability::adjusted): New.
1349 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1350 updating of profile.
1351 * value-prof.c (gimple_ic_transform): Rename to ...
1352 (dump_ic_profile): ... this one; update dumping.
1353 (stream_in_histogram_value): Fix formating.
1354 (gimple_value_profile_transformations): Update.
1356 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1359 * config/i386/i386.md (*movoi_internal_avx): Remove
1360 TARGET_SSE_TYPELESS_STORES check.
1361 (*movti_internal): Prefer TARGET_AVX over
1362 TARGET_SSE_TYPELESS_STORES.
1363 (*movtf_internal): Likewise.
1364 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1365 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1366 from TARGET_SSE_TYPELESS_STORES.
1368 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1370 * diagnostic-core.h (warning_at): Rename overload to...
1371 (warning_meta): ...this.
1372 (emit_diagnostic_valist): Delete decl of overload taking
1373 diagnostic_metadata.
1374 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1375 (warning_at): Rename overload taking diagnostic_metadata to...
1376 (warning_meta): ...this.
1378 2020-01-28 Richard Biener <rguenther@suse.de>
1380 PR tree-optimization/93439
1381 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1382 * tree-cfg.c (move_sese_region_to_fn): ... here.
1383 (verify_types_in_gimple_reference): Verify used cliques are
1386 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1389 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1390 argument of a pointer to struct gcc_options and pass it to
1391 parse_mtune_ctrl_str.
1392 (ix86_function_specific_restore): Pass opts to
1393 set_ix86_tune_features.
1394 (ix86_option_override_internal): Likewise.
1395 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1396 gcc_options and use it for x_ix86_tune_ctrl_string.
1398 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1400 PR rtl-optimization/87763
1401 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1402 simplification to handle subregs as well as bare regs.
1403 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1405 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1407 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1408 for reduction chains that (now) include a call.
1410 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1412 PR tree-optimization/92822
1413 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1414 out the don't-care elements of a vector whose significant elements
1415 are duplicates, make the don't-care elements duplicates too.
1417 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1419 PR tree-optimization/93434
1420 * tree-predcom.c (split_data_refs_to_components): Record which
1421 components have had aliasing loads removed. Prevent store-store
1422 commoning for all such components.
1424 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1427 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1428 -1 or is_vshift is true, use new_vector with number of elts npatterns
1429 rather than new_unary_operation.
1431 PR tree-optimization/93454
1432 * gimple-fold.c (fold_array_ctor_reference): Perform
1433 elt_size.to_uhwi () just once, instead of calling it in every
1434 iteration. Punt if that value is above size of the temporary
1435 buffer. Decrease third native_encode_expr argument when
1436 bufoff + elt_sz is above size of buf.
1438 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1440 * config/mips/mips.c (mips_declare_object_name)
1441 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1443 2020-01-27 Martin Liska <mliska@suse.cz>
1445 PR gcov-profile/93403
1446 * tree-profile.c (gimple_init_gcov_profiler): Generate
1447 both __gcov_indirect_call_profiler_v4 and
1448 __gcov_indirect_call_profiler_v4_atomic.
1450 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1453 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1455 (@aarch64_split_simd_mov<mode>): Use it.
1456 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1457 Leave the vec_extract patterns to handle 2-element vectors.
1458 (aarch64_simd_mov_from_<mode>high): Likewise.
1459 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1460 (vec_extractv2dfv1df): Likewise.
1462 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1464 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1465 jump conditions for *compare_condjump<GPI:mode>.
1467 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1470 * digraph.cc (test_edge::test_edge): Specify template for base
1473 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1475 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1477 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1479 * config/arc/arc-protos.h (gen_mlo): Remove.
1480 (gen_mhi): Likewise.
1481 * config/arc/arc.c (AUX_MULHI): Define.
1482 (arc_must_save_reister): Special handling for r58/59.
1483 (arc_compute_frame_size): Consider mlo/mhi registers.
1484 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1486 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1487 mlo/mhi name selection.
1488 (arc_restore_callee_saves): Don't early restore blink when ISR.
1489 (arc_expand_prologue): Add mlo/mhi saving.
1490 (arc_expand_epilogue): Add mlo/mhi restoring.
1493 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1494 numbering when MUL64 option is used.
1495 (DWARF2_FRAME_REG_OUT): Define.
1496 * config/arc/arc.md (arc600_stall): New pattern.
1497 (VUNSPEC_ARC_ARC600_STALL): Define.
1498 (mulsi64): Use correct mlo/mhi registers.
1499 (mulsi_600): Clean it up.
1500 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1502 (mhi_operand): Likewise.
1504 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1505 Petro Karashchenko <petro.karashchenko@ring.com>
1507 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1508 attributes if needed.
1509 (prepare_move_operands): Generate special unspec instruction for
1511 (arc_isuncached_mem_p): Propagate uncached attribute to each
1513 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1514 (VUNSPEC_ARC_STDI): Likewise.
1515 (ALLI): New mode iterator.
1516 (mALLI): New mode attribute.
1517 (lddi): New instruction pattern.
1519 (stdidi_split): Split instruction for architectures which are not
1520 supporting ll64 option.
1521 (lddidi_split): Likewise.
1523 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1525 PR rtl-optimization/92989
1526 * lra-lives.c (process_bb_lives): Update the live-in set before
1527 processing additional clobbers.
1529 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1531 PR rtl-optimization/93170
1532 * cselib.c (cselib_invalidate_regno_val): New function, split out
1534 (cselib_invalidate_regno): ...here.
1535 (cselib_invalidated_by_call_p): New function.
1536 (cselib_process_insn): Iterate over all the hard-register entries in
1537 REG_VALUES and invalidate any that cross call-clobbered registers.
1539 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1541 * dojump.c (split_comparison): Use HONOR_NANS rather than
1542 HONOR_SNANS when splitting LTGT.
1544 2020-01-27 Martin Liska <mliska@suse.cz>
1547 * opts.c (print_filtered_help): Exclude language-specific
1548 options from --help=common unless enabled in all FEs.
1550 2020-01-27 Martin Liska <mliska@suse.cz>
1552 * opts.c (print_help): Exclude params from
1553 all except --help=param.
1555 2020-01-27 Martin Liska <mliska@suse.cz>
1558 * config/i386/i386-features.c (make_resolver_func):
1559 Align the code with ppc64 target implementation.
1560 Do not generate a unique name for resolver function.
1562 2020-01-27 Richard Biener <rguenther@suse.de>
1564 PR tree-optimization/93397
1565 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1566 converted reduction chain SLP graph adjustment.
1568 2020-01-26 Marek Polacek <polacek@redhat.com>
1571 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1574 2020-01-26 Jason Merrill <jason@redhat.com>
1577 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1580 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1582 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1583 (rx_setmem): Likewise.
1585 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1588 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1589 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1590 drop <di> from constraint of last operand.
1593 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1594 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1595 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1597 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1600 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1603 2020-01-24 Jeff Law <law@redhat.com>
1605 PR tree-optimization/92788
1606 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1609 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1612 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1613 *avx_vperm_broadcast_<mode>,
1614 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1615 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1616 Move before avx2_perm<mode>/avx512f_perm<mode>.
1619 * simplify-rtx.c (simplify_const_unary_operation,
1620 simplify_const_binary_operation): Punt for mode precision above
1621 MAX_BITSIZE_MODE_ANY_INT.
1623 2020-01-24 Andrew Pinski <apinski@marvell.com>
1625 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1628 2020-01-24 Jeff Law <law@redhat.com>
1631 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1632 for REGs. Call output_operand_lossage to get more reasonable
1635 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1637 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1638 gcn_fp_compare_operator.
1639 (vec_cmpu<mode>di): Use gcn_compare_operator.
1640 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1641 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1642 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1643 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1644 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1645 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1646 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1647 gcn_fp_compare_operator.
1648 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1649 gcn_fp_compare_operator.
1650 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1651 gcn_fp_compare_operator.
1652 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1653 gcn_fp_compare_operator.
1655 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1657 * doc/install.texi (Cross-Compiler-Specific Options): Document
1658 `--with-toolexeclibdir' option.
1660 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1662 * target.def (flags_regnum): Also mention effect on delay slot filling.
1663 * doc/tm.texi: Regenerate.
1665 2020-01-23 Jeff Law <law@redhat.com>
1667 PR translation/90162
1668 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1670 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1673 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1676 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1678 PR rtl-optimization/93402
1679 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1682 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1684 * config.in: Regenerated.
1685 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1686 for TARGET_LIBC_GNUSTACK.
1687 * configure: Regenerated.
1688 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1689 found to be 2.31 or greater.
1691 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1693 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1695 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1696 (mips_asm_file_end): New function. Delegate to
1697 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1698 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1700 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1703 * config/i386/i386-modes.def (POImode): New mode.
1704 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1705 * config/i386/i386.md (DPWI): New mode attribute.
1706 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1708 (QPWI): ... this. Use POI instead of OI for TImode.
1709 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1710 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1713 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1716 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1718 (speculation_tracker_rev): New pattern.
1719 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1720 Use speculation_tracker_rev to track the inverse condition.
1722 2020-01-23 Richard Biener <rguenther@suse.de>
1724 PR tree-optimization/93381
1725 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1726 alias-set of the def as argument and record the first one.
1727 (vn_walk_cb_data::first_set): New member.
1728 (vn_reference_lookup_3): Pass the alias-set of the current def
1729 to push_partial_def. Fix alias-set used in the aggregate copy
1731 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1732 * real.c (clear_significand_below): Fix out-of-bound access.
1734 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1737 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1738 New define_insn patterns.
1740 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1742 * doc/sourcebuild.texi (check-function-bodies): Add an
1743 optional target/xfail selector.
1745 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1747 PR rtl-optimization/93124
1748 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1749 bare USE and CLOBBER insns.
1751 2020-01-22 Andrew Pinski <apinski@marvell.com>
1753 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1755 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1758 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1759 diagnostic_manager into "ana" namespace.
1760 * selftest-run-tests.c (selftest::run_tests): Update for move of
1761 selftest::run_analyzer_selftests to
1762 ana::selftest::run_analyzer_selftests.
1764 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1766 * cfgexpand.c (union_stack_vars): Update the size.
1768 2020-01-22 Richard Biener <rguenther@suse.de>
1770 PR tree-optimization/93381
1771 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1772 throughout, handle all conversions the same.
1774 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1777 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1778 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1779 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1780 Call force_reg on high_in2 unconditionally.
1782 2020-01-22 Martin Liska <mliska@suse.cz>
1784 PR tree-optimization/92924
1785 * profile.c (compute_value_histograms): Divide
1788 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1791 * output.h (assemble_name_resolve): Declare.
1792 * varasm.c (assemble_name_resolve): New function.
1793 (assemble_name): Use it.
1794 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1796 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1798 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1799 update_web_docs_git instead of update_web_docs_svn.
1801 2020-01-21 Andrew Pinski <apinski@marvell.com>
1804 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1805 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1806 (*tlsgd_small_<mode>): Likewise.
1807 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1808 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1809 register. Convert that register back to dest using convert_mode.
1811 2020-01-21 Jim Wilson <jimw@sifive.com>
1813 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1816 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1817 Uros Bizjak <ubizjak@gmail.com>
1820 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1822 (legitimize_tls_address): Do GNU2 TLS address computation in
1823 ptr_mode and zero-extend result to Pmode.
1824 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1825 :P with :PTR and Pmode with ptr_mode.
1826 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1827 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1828 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1830 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1833 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1834 the last two operands are CONST_INT_P before using them as such.
1836 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1838 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1839 to get the integer element types.
1841 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1843 * config/aarch64/aarch64-sve-builtins.h
1844 (function_expander::convert_to_pmode): Declare.
1845 * config/aarch64/aarch64-sve-builtins.cc
1846 (function_expander::convert_to_pmode): New function.
1847 (function_expander::get_contiguous_base): Use it.
1848 (function_expander::prepare_gather_address_operands): Likewise.
1849 * config/aarch64/aarch64-sve-builtins-sve2.cc
1850 (svwhilerw_svwhilewr_impl::expand): Likewise.
1852 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1855 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1856 cfun->machine->label_is_assembled.
1857 (aarch64_print_patchable_function_entry): New.
1858 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1859 * config/aarch64/aarch64.h (struct machine_function): New field,
1862 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1865 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1868 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1871 * cgraph.c (cgraph_edge::resolve_speculation,
1872 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1873 call_stmt_site_hash.
1875 2020-01-21 Martin Liska <mliska@suse.cz>
1877 * config/rs6000/rs6000.c (common_mode_defined): Remove
1880 2020-01-21 Richard Biener <rguenther@suse.de>
1882 PR tree-optimization/92328
1883 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1884 type when value-numbering same-sized store by inserting a
1886 (eliminate_dom_walker::eliminate_stmt): When eliminating
1887 a redundant store handle bit-reinterpretation of the same value.
1889 2020-01-21 Andrew Pinski <apinski@marvel.com>
1892 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1894 (prepare_block_for_update): This. Use a worklist instead of
1897 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1899 * gcc/config/arm/arm.c (clear_operation_p):
1900 Initialise last_regno, skip first iteration
1901 based on the first_set value and use ints instead
1902 of the unnecessary HOST_WIDE_INTs.
1904 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1907 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1908 compare_mode other than SFmode or DFmode.
1910 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1913 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1914 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1915 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1917 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1919 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1921 2020-01-20 Andrew Pinski <apinski@marvell.com>
1924 * targhooks.c (default_print_patchable_function_entry): Use
1925 output_asm_insn to emit the nop instruction.
1927 2020-01-20 Fangrui Song <maskray@google.com>
1930 * targhooks.c (default_print_patchable_function_entry): Align to
1933 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
1936 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1937 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
1938 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1939 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
1940 (*tls_dynamic_gnu2_lea_64): Renamed to ...
1941 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
1942 Remove the {q} suffix from lea.
1943 (*tls_dynamic_gnu2_call_64): Renamed to ...
1944 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
1945 (*tls_dynamic_gnu2_combine_64): Renamed to ...
1946 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
1947 Pass Pmode to gen_tls_dynamic_gnu2_64.
1949 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1951 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1953 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
1955 * config/aarch64/aarch64-sve-builtins-base.cc
1956 (svld1ro_impl::memory_vector_mode): Remove parameter name.
1958 2020-01-20 Richard Biener <rguenther@suse.de>
1961 * dwarf2out.c (prune_unused_types): Unconditionally mark
1962 called function DIEs.
1964 2020-01-20 Martin Liska <mliska@suse.cz>
1966 PR tree-optimization/93199
1967 * tree-eh.c (struct leh_state): Add
1968 new field outer_non_cleanup.
1969 (cleanup_is_dead_in): Pass leh_state instead
1970 of eh_region. Add a checking that state->outer_non_cleanup
1971 points to outer non-clean up region.
1972 (lower_try_finally): Record outer_non_cleanup
1974 (lower_catch): Likewise.
1975 (lower_eh_filter): Likewise.
1976 (lower_eh_must_not_throw): Likewise.
1977 (lower_cleanup): Likewise.
1979 2020-01-20 Richard Biener <rguenther@suse.de>
1981 PR tree-optimization/93094
1982 * tree-vectorizer.h (vect_loop_versioning): Adjust.
1983 (vect_transform_loop): Likewise.
1984 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1985 loop_vectorized_call to vect_transform_loop.
1986 * tree-vect-loop.c (vect_transform_loop): Pass down
1987 loop_vectorized_call to vect_loop_versioning.
1988 * tree-vect-loop-manip.c (vect_loop_versioning): Use
1989 the earlier discovered loop_vectorized_call.
1991 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
1993 * doc/contribute.texi: Update for SVN -> Git transition.
1994 * doc/install.texi: Likewise.
1996 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1999 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2000 speculative targets.
2001 (verify_speculative_call): New function
2002 (cgraph_node::verify_node): Use it.
2003 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2006 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2009 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2010 (cgraph_edge::make_direct): Remove all indirect targets.
2011 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2012 (cgraph_node::verify_node): Verify that only one call_stmt or
2013 lto_stmt_uid is set.
2014 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2016 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2017 (lto_output_ref): Simplify streaming of stmt.
2018 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2020 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2022 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2023 Mark parameter unused.
2025 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2027 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2029 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2031 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2033 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2035 * Makefile.in: Add coroutine-passes.o.
2036 * builtin-types.def (BT_CONST_SIZE): New.
2037 (BT_FN_BOOL_PTR): New.
2038 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2039 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2040 * coroutine-builtins.def: New file.
2041 * coroutine-passes.cc: New file.
2042 * function.h (struct GTY function): Add a bit to indicate that the
2043 function is a coroutine component.
2044 * internal-fn.c (expand_CO_FRAME): New.
2045 (expand_CO_YIELD): New.
2046 (expand_CO_SUSPN): New.
2047 (expand_CO_ACTOR): New.
2048 * internal-fn.def (CO_ACTOR): New.
2052 * passes.def: Add pass_coroutine_lower_builtins,
2053 pass_coroutine_early_expand_ifns.
2054 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2055 (make_pass_coroutine_early_expand_ifns): New.
2056 * doc/invoke.texi: Document the fcoroutines command line
2059 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2061 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2064 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2065 after checking the argument is a REG. Don't use REGNO (reg)
2066 again to set last_regno, reuse regno variable instead.
2068 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2070 * doc/analyzer.texi (Limitations): Add note about NaN.
2072 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2073 Sudakshina Das <sudi.das@arm.com>
2075 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2076 and valid immediate.
2077 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2078 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2079 * config/arm/constraints.md (Pg): New.
2080 * config/arm/predicates.md (long_shift_imm): New.
2081 (arm_reg_or_long_shift_imm): Likewise.
2082 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2083 (thumb2_lsll): Likewise.
2086 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2087 Sudakshina Das <sudi.das@arm.com>
2089 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2090 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2091 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2092 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2093 * config/arm/thumb2.md (thumb2_asrl): New.
2094 (thumb2_lsll): Likewise.
2096 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2098 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2101 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2103 * gdbinit.in (help-gcc-hooks): New command.
2104 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2105 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2108 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2110 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2111 correct target macro.
2113 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2115 * config/aarch64/aarch64-protos.h
2116 (aarch64_sve_ld1ro_operand_p): New.
2117 * config/aarch64/aarch64-sve-builtins-base.cc
2118 (class load_replicate): New.
2119 (class svld1ro_impl): New.
2120 (class svld1rq_impl): Change to inherit from load_replicate.
2121 (svld1ro): New sve intrinsic function base.
2122 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2123 New DEF_SVE_FUNCTION.
2124 * config/aarch64/aarch64-sve-builtins-base.h
2125 (svld1ro): New decl.
2126 * config/aarch64/aarch64-sve-builtins.cc
2127 (function_expander::add_mem_operand): Modify assert to allow
2129 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2131 * config/aarch64/aarch64.c
2132 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2133 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2134 (aarch64_sve_ld1ro_operand_p): New.
2135 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2136 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2137 * config/aarch64/predicates.md
2138 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2140 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2142 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2143 Introduce this ACLE specified predefined macro.
2144 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2145 (fp): Disabling this disables f64mm.
2146 (simd): Disabling this disables f64mm.
2147 (fp16): Disabling this disables f64mm.
2148 (sve): Disabling this disables f64mm.
2149 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2150 (AARCH64_ISA_F64MM): New.
2151 (TARGET_F64MM): New.
2152 * doc/invoke.texi (f64mm): Document new option.
2154 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2156 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2157 (neoversen1_tunings): Likewise.
2159 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2162 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2163 Add assert to ensure prolog has been emitted.
2164 (aarch64_split_atomic_op): Likewise.
2165 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2166 Use epilogue_completed rather than reload_completed.
2167 (aarch64_atomic_exchange<mode>): Likewise.
2168 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2169 (atomic_nand<mode>): Likewise.
2170 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2171 (atomic_fetch_nand<mode>): Likewise.
2172 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2173 (atomic_nand_fetch<mode>): Likewise.
2175 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2178 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2180 (REVERSE_CONDITION): Delete.
2181 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2182 (CCFP_CCFPE): Likewise.
2183 (e): New mode attribute.
2184 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2185 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2186 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2187 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2188 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2189 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2190 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2191 name of generator from gen_ccmpdi to gen_ccmpccdi.
2192 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2193 the previous comparison but aren't able to, use the new ccmp_rev
2196 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2198 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2199 than testing directly for INTEGER_CST.
2200 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2202 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2204 PR tree-optimization/93292
2205 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2206 get_vectype_for_scalar_type returns NULL.
2208 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2210 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2211 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2213 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2215 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2217 * params.opt: (max-predicted-iterations): Set bounds.
2218 * predict.c (real_almost_one, real_br_prob_base,
2219 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2220 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2221 probabilities; do not truncate to reg_br_prob_bases.
2222 (estimate_loops_at_level): Pass max_cyclic_prob.
2223 (estimate_loops): Compute max_cyclic_prob.
2224 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2226 * profile-count.c (profile_probability::to_sreal): New.
2227 * profile-count.h (class sreal): Move up in file.
2228 (profile_probability::to_sreal): Declare.
2230 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2233 (arm_invalid_conversion): New function for target hook.
2234 (arm_invalid_unary_op): New function for target hook.
2235 (arm_invalid_binary_op): New function for target hook.
2237 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2239 * config.gcc: Add arm_bf16.h.
2240 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2241 (arm_simd_builtin_std_type): Add BFmode.
2242 (arm_init_simd_builtin_types): Define element types for vector types.
2243 (arm_init_bf16_types): New function.
2244 (arm_init_builtins): Add arm_init_bf16_types function call.
2245 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2246 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2247 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2248 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2249 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2250 (arm_mangle_type): Add __bf16.
2251 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2252 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2253 arm_bf16_ptr_type_node.
2254 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2255 define_split between ARM registers.
2256 * config/arm/arm_bf16.h: New file.
2257 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2258 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2260 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2261 * config/arm/vfp.md: Add BFmode to movhf patterns.
2263 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2264 Andre Vieira <andre.simoesdiasvieira@arm.com>
2266 * config/arm/arm-cpus.in (mve, mve_float): New features.
2267 (dsp, mve, mve.fp): New options.
2268 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2269 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2270 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2272 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2273 Thomas Preud'homme <thomas.preudhomme@arm.com>
2275 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2277 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2278 error for using -mcmse when targeting Armv8.1-M Mainline.
2280 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2281 Thomas Preud'homme <thomas.preudhomme@arm.com>
2283 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2284 address in r4 when targeting Armv8.1-M Mainline.
2285 (nonsecure_call_value_internal): Likewise.
2286 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2287 a register match_operand again. Emit BLXNS when targeting
2289 (nonsecure_call_value_reg_thumb2): Likewise.
2291 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2292 Thomas Preud'homme <thomas.preudhomme@arm.com>
2294 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2295 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2296 variable as true when floating-point ABI is not hard. Replace
2297 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2298 Generate VLSTM and VLLDM instruction respectively before and
2299 after a function call to cmse_nonsecure_call function.
2300 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2301 (VUNSPEC_VLLDM): Likewise.
2302 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2303 (lazy_load_multiple_insn): Likewise.
2305 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2306 Thomas Preud'homme <thomas.preudhomme@arm.com>
2308 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2309 (arm_emit_vfp_multi_reg_pop): Likewise.
2310 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2311 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2312 restore callee-saved VFP registers.
2314 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2315 Thomas Preud'homme <thomas.preudhomme@arm.com>
2317 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2318 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2319 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2320 callee-saved GPRs as well as clear ip register before doing a nonsecure
2321 call then restore callee-saved GPRs after it when targeting
2323 (arm_reorg): Adapt to function rename.
2325 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2326 Thomas Preud'homme <thomas.preudhomme@arm.com>
2328 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2329 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2330 clear_vfp_multiple pattern based on a new vfp parameter.
2331 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2332 targeting Armv8.1-M Mainline.
2333 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2334 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2335 whether VFP registers are available before looking call_used_regs for a
2337 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2338 of prototype of clear_operation_p.
2339 (clear_vfp_multiple_operation): New predicate.
2340 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2341 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2343 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2344 Thomas Preud'homme <thomas.preudhomme@arm.com>
2346 * config/arm/arm-protos.h (clear_operation_p): Declare.
2347 * config/arm/arm.c (clear_operation_p): New function.
2348 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2349 targeting Armv8.1-M Mainline or successor.
2350 (output_return_instruction): Only output APSR register clearing if
2351 Armv8.1-M Mainline instructions not available.
2352 (thumb_exit): Likewise.
2353 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2354 * config/arm/thumb2.md (clear_apsr): New define_insn.
2355 (clear_multiple): Likewise.
2356 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2358 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2359 Thomas Preud'homme <thomas.preudhomme@arm.com>
2361 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2362 (use_return_insn): Also return false for Armv8.1-M Mainline.
2363 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2364 Mainline instructions are available.
2365 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2366 when targeting Armv8.1-M Mainline Security Extensions.
2367 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2368 Mainline entry function.
2369 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2370 targeting Armv8.1-M Mainline or successor.
2371 (arm_expand_epilogue): Fix indentation of caller-saved register
2372 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2374 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2375 (FP_SYSREGS): Likewise.
2376 (enum vfp_sysregs_encoding): Define enum.
2377 (fp_sysreg_names): Declare.
2378 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2379 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2380 (pop_fpsysreg_insn): Likewise.
2382 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2383 Thomas Preud'homme <thomas.preudhomme@arm.com>
2385 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2386 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2387 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2388 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2389 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2390 (ARMv8_1m_main): New feature group.
2391 (armv8.1-m.main): New architecture.
2392 * config/arm/arm-tables.opt: Regenerate.
2393 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2394 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2395 (arm_options_perform_arch_sanity_checks): Error out when targeting
2396 Armv8.1-M Mainline Security Extensions.
2397 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2399 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2401 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2402 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2403 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2404 aarch64_bfdot_laneq): New.
2405 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2406 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2407 vbfdotq_laneq_f32): New.
2408 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2409 VBFMLA_W, VBF): New.
2410 (isquadop): Add V4BF, V8BF.
2412 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2414 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2415 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2416 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2417 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2418 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2419 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2420 usdot_laneq, sudot_lane,sudot_laneq): New.
2421 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2422 (aarch64_<sur>dot_lane): New.
2423 * config/aarch64/arm_neon.h (vusdot_s32): New.
2425 (vusdot_lane_s32): New.
2426 (vsudot_lane_s32): New.
2427 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2428 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2430 2020-01-16 Martin Liska <mliska@suse.cz>
2432 * value-prof.c (dump_histogram_value): Fix
2433 obvious spacing issue.
2435 2020-01-16 Andrew Pinski <apinski@marvell.com>
2437 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2438 !storage_order_barrier_p.
2440 2020-01-16 Andrew Pinski <apinski@marvell.com>
2442 * sched-int.h (_dep): Add unused bit-field field for the padding.
2443 * sched-deps.c (init_dep_1): Init unused field.
2445 2020-01-16 Andrew Pinski <apinski@marvell.com>
2447 * optabs.h (create_expand_operand): Initialize target field also.
2449 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2451 PR tree-optimization/92429
2452 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2453 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2455 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2458 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2460 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2461 aarch64_sve_int_mode to each mode.
2463 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2465 * doc/analyzer.texi (Overview): Add note about
2466 -fdump-ipa-analyzer.
2468 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2470 PR tree-optimization/93231
2471 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2472 input_type is unsigned. Use tree_to_shwi for shift constant.
2473 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2474 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2475 inputs more efficiently.
2477 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2479 * config/i386/i386.md (*movsf_internal): Do not require
2480 SSE2 ISA for alternatives 14 and 15.
2482 2020-01-15 Richard Biener <rguenther@suse.de>
2485 * tree-eh.c (sink_clobbers): If we already visited the destination
2486 block do not defer insertion.
2487 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2488 the purpose of defered insertion.
2490 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2492 * BASE-VER: Bump to 10.0.1.
2494 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2496 PR tree-optimization/93247
2497 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2498 type of the stmt that we're going to vectorize.
2500 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2502 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2503 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2506 2020-01-15 Martin Liska <mliska@suse.cz>
2508 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2509 2 calls of streamer_read_hwi in a function call.
2511 2020-01-15 Richard Biener <rguenther@suse.de>
2513 * alias.c (record_alias_subset): Avoid redundant work when
2514 subset is already recorded.
2516 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2518 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2519 the analyzer options provide CWE identifiers.
2521 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2523 * tree-diagnostic-path.cc (path_summary::event_range::print):
2524 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2525 using get_pure_location.
2527 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2529 PR tree-optimization/93262
2530 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2531 perform head trimming only if the last argument is constant,
2532 either all ones, or larger or equal to head trim, in the latter
2533 case decrease the last argument by head_trim.
2535 PR tree-optimization/93249
2536 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2537 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2538 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2539 perform head trim unless we can prove there are no '\0' chars
2540 from the source among the first head_trim chars.
2542 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2544 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2546 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2549 * config/i386/sse.md
2550 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2551 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2552 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2553 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2554 just a single alternative instead of two, make operands 1 and 2
2557 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2560 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2563 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2565 * Makefile.in (lang_opt_files): Add analyzer.opt.
2566 (ANALYZER_OBJS): New.
2567 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2568 tristate.o and ANALYZER_OBJS.
2569 (TEXI_GCCINT_FILES): Add analyzer.texi.
2570 * common.opt (-fanalyzer): New driver option.
2571 * config.in: Regenerate.
2572 * configure: Regenerate.
2573 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2574 (gccdepdir): Also create depdir for "analyzer" subdir.
2575 * digraph.cc: New file.
2576 * digraph.h: New file.
2577 * doc/analyzer.texi: New file.
2578 * doc/gccint.texi ("Static Analyzer") New menu item.
2579 (analyzer.texi): Include it.
2580 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2581 ("Warning Options"): Add static analysis warnings to the list.
2582 (-Wno-analyzer-double-fclose): New option.
2583 (-Wno-analyzer-double-free): New option.
2584 (-Wno-analyzer-exposure-through-output-file): New option.
2585 (-Wno-analyzer-file-leak): New option.
2586 (-Wno-analyzer-free-of-non-heap): New option.
2587 (-Wno-analyzer-malloc-leak): New option.
2588 (-Wno-analyzer-possible-null-argument): New option.
2589 (-Wno-analyzer-possible-null-dereference): New option.
2590 (-Wno-analyzer-null-argument): New option.
2591 (-Wno-analyzer-null-dereference): New option.
2592 (-Wno-analyzer-stale-setjmp-buffer): New option.
2593 (-Wno-analyzer-tainted-array-index): New option.
2594 (-Wno-analyzer-use-after-free): New option.
2595 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2596 (-Wno-analyzer-use-of-uninitialized-value): New option.
2597 (-Wanalyzer-too-complex): New option.
2598 (-fanalyzer-call-summaries): New warning.
2599 (-fanalyzer-checker=): New warning.
2600 (-fanalyzer-fine-grained): New warning.
2601 (-fno-analyzer-state-merge): New warning.
2602 (-fno-analyzer-state-purge): New warning.
2603 (-fanalyzer-transitivity): New warning.
2604 (-fanalyzer-verbose-edges): New warning.
2605 (-fanalyzer-verbose-state-changes): New warning.
2606 (-fanalyzer-verbosity=): New warning.
2607 (-fdump-analyzer): New warning.
2608 (-fdump-analyzer-callgraph): New warning.
2609 (-fdump-analyzer-exploded-graph): New warning.
2610 (-fdump-analyzer-exploded-nodes): New warning.
2611 (-fdump-analyzer-exploded-nodes-2): New warning.
2612 (-fdump-analyzer-exploded-nodes-3): New warning.
2613 (-fdump-analyzer-supergraph): New warning.
2614 * doc/sourcebuild.texi (dg-require-dot): New.
2615 (dg-check-dot): New.
2616 * gdbinit.in (break-on-saved-diagnostic): New command.
2617 * graphviz.cc: New file.
2618 * graphviz.h: New file.
2619 * ordered-hash-map-tests.cc: New file.
2620 * ordered-hash-map.h: New file.
2621 * passes.def (pass_analyzer): Add before
2622 pass_ipa_whole_program_visibility.
2623 * selftest-run-tests.c (selftest::run_tests): Call
2624 selftest::ordered_hash_map_tests_cc_tests.
2625 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2627 * shortest-paths.h: New file.
2628 * timevar.def (TV_ANALYZER): New timevar.
2629 (TV_ANALYZER_SUPERGRAPH): Likewise.
2630 (TV_ANALYZER_STATE_PURGE): Likewise.
2631 (TV_ANALYZER_PLAN): Likewise.
2632 (TV_ANALYZER_SCC): Likewise.
2633 (TV_ANALYZER_WORKLIST): Likewise.
2634 (TV_ANALYZER_DUMP): Likewise.
2635 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2636 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2637 * tree-pass.h (make_pass_analyzer): New decl.
2638 * tristate.cc: New file.
2639 * tristate.h: New file.
2641 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2644 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2645 alternatives 9 and 10.
2647 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2649 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2650 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2651 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2652 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2653 (selftest::hash_map_tests_c_tests): Call it.
2654 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2655 New static constant, using the value of = H::empty_zero_p.
2656 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2657 from default_hash_traits <Value>.
2658 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2660 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2661 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2662 calls to mark_empty with !Descriptor::empty_zero_p.
2663 (hash_table::empty_slow): Conditionalize the memset call with a
2664 check that Descriptor::empty_zero_p; otherwise, loop through the
2665 entries calling mark_empty on them.
2666 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2667 (pointer_hash::empty_zero_p): Likewise.
2668 (pair_hash::empty_zero_p): Likewise.
2669 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2671 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2672 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2673 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2674 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2675 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2676 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2677 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2678 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2680 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2683 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2685 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2686 fix typo on return value.
2688 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2691 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2693 (cgraph_edge::make_speculative): Add param for setting speculative_id
2695 (cgraph_edge::speculative_call_info): Update comments and find reference
2696 by speculative_id for multiple indirect targets.
2697 (cgraph_edge::resolve_speculation): Decrease the speculations
2698 for indirect edge, drop it's speculative if not direct target
2699 left. Update comments.
2700 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2701 (cgraph_node::dump): Print num_speculative_call_targets.
2702 (cgraph_node::verify_node): Don't report error if speculative
2703 edge not include statement.
2704 (cgraph_edge::num_speculative_call_targets_p): New function.
2705 * cgraph.h (int common_target_id): Remove.
2706 (int common_target_probability): Remove.
2707 (num_speculative_call_targets): New variable.
2708 (make_speculative): Add param for setting speculative_id.
2709 (cgraph_edge::num_speculative_call_targets_p): New declare.
2710 (target_prob): New variable.
2711 (speculative_id): New variable.
2712 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2713 call summaries for multiple speculative call targets.
2714 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2715 * ipa-profile.c (struct speculative_call_target): New struct.
2716 (class speculative_call_summary): New class.
2717 (class speculative_call_summaries): New class.
2718 (call_sums): New variable.
2719 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2720 (ipa_profile_write_edge_summary): New function.
2721 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2722 (ipa_profile_dump_all_summaries): New function.
2723 (ipa_profile_read_edge_summary): New function.
2724 (ipa_profile_read_summary_section): New function.
2725 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2726 (ipa_profile): Generate num_speculative_call_targets from
2728 * ipa-ref.h (speculative_id): New variable.
2729 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2730 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2731 common_target_probability. Stream out speculative_id and
2732 num_speculative_call_targets.
2733 (input_edge): Likewise.
2734 * predict.c (dump_prediction): Remove edges count assert to be
2736 * symtab.c (symtab_node::create_reference): Init speculative_id.
2737 (symtab_node::clone_references): Clone speculative_id.
2738 (symtab_node::clone_referring): Clone speculative_id.
2739 (symtab_node::clone_reference): Clone speculative_id.
2740 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2741 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2742 if indirect call contains multiple speculative targets.
2743 * value-prof.h (check_ic_target): Remove.
2744 * value-prof.c (gimple_value_profile_transformations):
2745 Use void function gimple_ic_transform.
2746 * value-prof.c (gimple_ic_transform): Handle topn case.
2747 Fix comment typos. Change it to a void function.
2749 2020-01-13 Andrew Pinski <apinski@marvell.com>
2751 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2752 (octeontx2t98): New define.
2753 (octeontx2t96): New define.
2754 (octeontx2t93): New define.
2755 (octeontx2f95): New define.
2756 (octeontx2f95n): New define.
2757 (octeontx2f95mm): New define.
2758 * config/aarch64/aarch64-tune.md: Regenerate.
2759 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2761 2020-01-13 Jason Merrill <jason@redhat.com>
2763 PR c++/33799 - destroy return value if local cleanup throws.
2764 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2766 2020-01-13 Martin Liska <mliska@suse.cz>
2768 * ipa-cp.c (get_max_overall_size): Use newly
2769 renamed param param_ipa_cp_unit_growth.
2770 * params.opt: Remove legacy param name.
2772 2020-01-13 Martin Sebor <msebor@redhat.com>
2774 PR tree-optimization/93213
2775 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2776 stores to be eliminated.
2778 2020-01-13 Martin Liska <mliska@suse.cz>
2780 * opts.c (print_help): Do not print CL_PARAM
2781 and CL_WARNING for CL_OPTIMIZATION.
2783 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2786 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2787 depending on optimization settings.
2789 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2791 PR tree-optimization/90838
2792 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2793 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2794 argument rather than to initialize temporary for targets that
2795 don't use the mode argument at all. Initialize ctzval to avoid
2798 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2800 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2801 * tree-core.h: Document it.
2802 * gimplify.c (gimplify_omp_workshare): Set it.
2803 * omp-low.c (lower_omp_target): Use it.
2804 * tree-pretty-print.c (dump_omp_clause): Print it.
2806 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2807 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2809 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2811 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2812 * common.opt (fdiagnostics-path-format=): New option.
2813 (diagnostic_path_format): New enum.
2814 (fdiagnostics-show-path-depths): New option.
2815 * coretypes.h (diagnostic_event_id_t): New forward decl.
2816 * diagnostic-color.c (color_dict): Add "path".
2817 * diagnostic-event-id.h: New file.
2818 * diagnostic-format-json.cc (json_from_expanded_location): Make
2820 (json_end_diagnostic): Call context->make_json_for_path if it
2821 exists and the diagnostic has a path.
2822 (diagnostic_output_format_init): Clear context->print_path.
2823 * diagnostic-path.h: New file.
2824 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2825 when printing a run of events in a diagnostic_path so that they
2826 all get the same color.
2827 (layout::m_diagnostic_path_p): New field.
2828 (layout::layout): Initialize it.
2829 (layout::print_any_labels): Don't colorize the label text for an
2830 event in a diagnostic_path.
2831 (gcc_rich_location::add_location_if_nearby): Add
2832 "restrict_to_current_line_spans" and "label" params. Pass the
2833 former to layout.maybe_add_location_range; pass the latter
2834 when calling add_range.
2835 * diagnostic.c: Include "diagnostic-path.h".
2836 (diagnostic_initialize): Initialize context->path_format and
2837 context->show_path_depths.
2838 (diagnostic_show_any_path): New function.
2839 (diagnostic_path::interprocedural_p): New function.
2840 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2841 (simple_diagnostic_path::num_events): New function.
2842 (simple_diagnostic_path::get_event): New function.
2843 (simple_diagnostic_path::add_event): New function.
2844 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2845 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2846 (debug): New overload taking a diagnostic_path *.
2847 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2848 * diagnostic.h (enum diagnostic_path_format): New enum.
2849 (json::value): New forward decl.
2850 (diagnostic_context::path_format): New field.
2851 (diagnostic_context::show_path_depths): New field.
2852 (diagnostic_context::print_path): New callback field.
2853 (diagnostic_context::make_json_for_path): New callback field.
2854 (diagnostic_show_any_path): New decl.
2855 (json_from_expanded_location): New decl.
2856 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2857 (-fdiagnostics-show-path-depths): New option.
2858 (-fdiagnostics-color): Add "path" to description of default
2859 GCC_COLORS; describe it.
2860 (-fdiagnostics-format=json): Document how diagnostic paths are
2861 represented in the JSON output format.
2862 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2863 Add optional params "restrict_to_current_line_spans" and "label".
2864 * opts.c (common_handle_option): Handle
2865 OPT_fdiagnostics_path_format_ and
2866 OPT_fdiagnostics_show_path_depths.
2867 * pretty-print.c: Include "diagnostic-event-id.h".
2868 (pp_format): Implement "%@" format code for printing
2869 diagnostic_event_id_t *.
2870 (selftest::test_pp_format): Add tests for "%@".
2871 * selftest-run-tests.c (selftest::run_tests): Call
2872 selftest::tree_diagnostic_path_cc_tests.
2873 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2874 * toplev.c (general_init): Initialize global_dc->path_format and
2875 global_dc->show_path_depths.
2876 * tree-diagnostic-path.cc: New file.
2877 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2878 non-static. Drop "diagnostic" param in favor of storing the
2879 original value of "where" and re-using it.
2880 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2881 maybe_unwind_expanded_macro_loc.
2882 (tree_diagnostics_defaults): Initialize context->print_path and
2883 context->make_json_for_path.
2884 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2886 (default_tree_make_json_for_path): New decl.
2887 (maybe_unwind_expanded_macro_loc): New decl.
2889 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2891 PR tree-optimization/93210
2892 * fold-const.h (native_encode_initializer,
2893 can_native_interpret_type_p): Declare.
2894 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2896 (native_encode_initializer): New function, moved from dwarf2out.c.
2897 Adjust to native_encode_expr compatible arguments, including dry-run
2898 and partial extraction modes. Don't handle STRING_CST.
2899 (can_native_interpret_type_p): No longer static.
2900 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2901 offset / BITS_PER_UNIT fits into int and don't call it if
2902 can_native_interpret_type_p fails. If suboff is NULL and for
2903 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2904 native_encode_initializer.
2905 (fold_const_aggregate_ref_1): Formatting fix.
2906 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2907 (tree_add_const_value_attribute): Adjust caller.
2909 PR tree-optimization/90838
2910 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2911 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2912 CTZ_DEFINED_VALUE_AT_ZERO.
2914 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2917 * lra-constraints.c (match_reload): Permit input operands have the
2918 same mode as output while other input operands have a different
2921 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
2923 PR tree-optimization/90838
2924 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2925 (check_ctz_string): Likewise.
2926 (optimize_count_trailing_zeroes): Likewise.
2927 (simplify_count_trailing_zeroes): Likewise.
2928 (pass_forwprop::execute): Try ctz simplification.
2929 * match.pd: Add matching for ctz idioms.
2931 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2933 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2935 (aarch64_invalid_unary_op): New function for target hook.
2936 (aarch64_invalid_binary_op): New function for target hook.
2938 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2940 * config.gcc: Add arm_bf16.h.
2941 * config/aarch64/aarch64-builtins.c
2942 (aarch64_simd_builtin_std_type): Add BFmode.
2943 (aarch64_init_simd_builtin_types): Define element types for vector
2945 (aarch64_init_bf16_types): New function.
2946 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2947 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2949 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2950 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2952 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2953 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2954 * config/aarch64/aarch64.c
2955 (aarch64_classify_vector_mode): Add support for BF types.
2956 (aarch64_gimplify_va_arg_expr): Add support for BF types.
2957 (aarch64_vq_mode): Add support for BF types.
2958 (aarch64_simd_container_mode): Add support for BF types.
2959 (aarch64_mangle_type): Add support for BF scalar type.
2960 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2961 * config/aarch64/arm_bf16.h: New file.
2962 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2963 * config/aarch64/iterators.md: Add BF types to mode attributes.
2964 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2966 2020-01-10 Jason Merrill <jason@redhat.com>
2968 PR c++/93173 - incorrect tree sharing.
2969 * gimplify.c (copy_if_shared): No longer static.
2970 * gimplify.h: Declare it.
2972 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2974 * doc/invoke.texi (-msve-vector-bits=): Document that
2975 -msve-vector-bits=128 now generates VL-specific code for
2976 little-endian targets.
2977 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2978 build_vector_type_for_mode to construct the data vector types.
2979 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2980 VL-specific code for -msve-vector-bits=128 on little-endian targets.
2981 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2982 for 128-bit vectors.
2984 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2986 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2989 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2991 * config/aarch64/aarch64-builtins.c
2992 (aarch64_builtin_vectorized_function): Check for specific vector modes,
2993 rather than checking the number of elements and the element mode.
2995 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2997 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2998 get_related_vectype_for_scalar_type rather than build_vector_type
2999 to create the index type for a conditional reduction.
3001 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3003 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3004 for any type of gather or scatter, including strided accesses.
3006 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3008 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3011 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3013 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3015 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3016 parameter and its use to reset DR_OFFSET's.
3017 (vect_transform_loop): Remove orig_drs_init argument.
3018 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3019 member of dr_vec_info rather than the offset of the associated
3020 data_reference's innermost_loop_behavior.
3021 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3022 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3023 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3024 get_dr_vinfo_offset.
3025 (vectorizable_store): Likewise.
3026 (vectorizable_load): Likewise.
3028 2020-01-10 Richard Biener <rguenther@suse.de>
3030 * gimple-ssa-store-merging
3031 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3033 2020-01-10 Martin Liska <mliska@suse.cz>
3036 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3037 encapsulation that was there before r280040.
3039 2020-01-10 Richard Biener <rguenther@suse.de>
3042 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3043 sequences to avoid walking them again for secondary opportunities.
3044 (pass_lower_eh_dispatch::execute): Instead actually insert
3047 2020-01-10 Richard Biener <rguenther@suse.de>
3050 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3051 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3052 avoid quadraticness.
3054 2020-01-10 Martin Jambor <mjambor@suse.cz>
3056 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3057 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3058 to get param_ipa_sra_max_replacements.
3059 (param_splitting_across_edge): Pass the caller to
3060 pull_accesses_from_callee.
3062 2020-01-10 Martin Jambor <mjambor@suse.cz>
3064 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3065 * ipa-cp.c (max_new_size): Removed.
3066 (orig_overall_size): New variable.
3067 (get_max_overall_size): New function.
3068 (estimate_local_effects): Use it. Adjust dump.
3069 (decide_about_value): Likewise.
3070 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3071 orig_overall_size. Adjust dump.
3072 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3074 2020-01-10 Martin Jambor <mjambor@suse.cz>
3076 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3077 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3078 instead of param_ipa_max_agg_items.
3079 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3080 optimization info for the callee.
3082 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3084 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3085 markers if debug_inline_points is false.
3087 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3089 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3091 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3092 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3093 aarch64-sve-builtins-sve2.h.
3094 (aarch64-sve-builtins-sve2.o): New rule.
3095 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3096 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3097 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3098 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3099 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3100 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3102 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3103 instructions that are handled here.
3104 (@cond_asrd<mode>): Generalize to...
3105 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3106 (*cond_asrd<mode>_2): Generalize to...
3107 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3108 (*cond_asrd<mode>_z): Generalize to...
3109 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3110 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3111 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3112 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3113 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3115 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3116 (@aarch64_scatter_stnt<mode>): Likewise.
3117 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3118 (@aarch64_mul_lane_<mode>): Likewise.
3119 (@aarch64_sve_suqadd<mode>_const): Likewise.
3120 (*<sur>h<addsub><mode>): Generalize to...
3121 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3123 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3124 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3125 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3126 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3127 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3128 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3129 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3130 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3131 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3132 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3133 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3134 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3135 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3136 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3137 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3138 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3139 (@aarch64_sve2_xar<mode>): Likewise.
3140 (@aarch64_sve2_bcax<mode>): Likewise.
3141 (*aarch64_sve2_eor3<mode>): Rename to...
3142 (@aarch64_sve2_eor3<mode>): ...this.
3143 (@aarch64_sve2_bsl<mode>): New expander.
3144 (@aarch64_sve2_nbsl<mode>): Likewise.
3145 (@aarch64_sve2_bsl1n<mode>): Likewise.
3146 (@aarch64_sve2_bsl2n<mode>): Likewise.
3147 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3148 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3149 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3150 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3151 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3152 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3153 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3154 (<su>mull<bt><Vwide>): Generalize to...
3155 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3157 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3158 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3159 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3160 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3161 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3162 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3163 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3164 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3165 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3166 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3167 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3168 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3169 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3170 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3171 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3172 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3173 (<SHRNB:r>shrnb<mode>): Generalize to...
3174 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3176 (<SHRNT:r>shrnt<mode>): Generalize to...
3177 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3179 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3180 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3181 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3182 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3183 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3184 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3185 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3186 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3187 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3188 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3189 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3190 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3191 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3192 (@aarch64_sve2_cvtnt<mode>): Likewise.
3193 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3194 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3195 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3196 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3197 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3198 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3199 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3200 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3201 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3202 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3203 (@aarch64_sve2_pmul<mode>): Likewise.
3204 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3205 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3206 (@aarch64_sve2_tbl2<mode>): Likewise.
3207 (@aarch64_sve2_tbx<mode>): Likewise.
3208 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3209 (@aarch64_sve2_histcnt<mode>): Likewise.
3210 (@aarch64_sve2_histseg<mode>): Likewise.
3211 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3212 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3213 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3214 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3215 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3216 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3217 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3218 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3219 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3220 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3221 (SVE2_PMULL_PAIR_I): New mode iterators.
3222 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3223 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3224 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3225 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3226 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3227 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3228 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3229 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3230 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3231 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3232 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3233 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3234 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3235 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3236 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3237 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3238 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3239 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3240 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3241 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3242 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3243 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3244 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3245 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3246 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3247 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3248 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3249 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3250 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3251 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3252 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3254 (VNARROW, Ventype): New mode attributes.
3255 (Vewtype): Handle VNx2DI. Fix typo in comment.
3256 (VDOUBLE): New mode attribute.
3257 (sve_lane_con): Handle VNx8HI.
3258 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3259 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3260 (sve_int_op, sve_int_op_rev): Handle the above codes.
3261 (sve_pred_int_rhs2_operand): Likewise.
3262 (MULLBT, SHRNB, SHRNT): Delete.
3263 (SVE_INT_SHIFT_IMM): New int iterator.
3264 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3265 and UNSPEC_WHILEHS for TARGET_SVE2.
3266 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3267 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3268 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3269 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3270 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3271 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3272 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3273 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3274 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3275 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3276 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3277 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3278 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3279 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3280 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3281 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3282 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3283 (optab): Handle the new unspecs.
3284 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3286 (lr): Handle the new unspecs.
3288 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3289 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3290 (sve_int_qsub_op): New int attributes.
3291 (sve_fp_op, rot): Handle the new unspecs.
3292 * config/aarch64/aarch64-sve-builtins.h
3293 (function_resolver::require_matching_pointer_type): Declare.
3294 (function_resolver::resolve_unary): Add an optional boolean argument.
3295 (function_resolver::finish_opt_n_resolution): Add an optional
3296 type_suffix_index argument.
3297 (gimple_folder::redirect_call): Declare.
3298 (gimple_expander::prepare_gather_address_operands): Add an optional
3300 * config/aarch64/aarch64-sve-builtins.cc: Include
3301 aarch64-sve-builtins-sve2.h.
3302 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3303 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3304 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3305 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3306 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3307 (TYPES_s_unsigned): Likewise.
3308 (TYPES_s_integer): Use TYPES_s_unsigned.
3309 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3310 (TYPES_sd_integer): Use them.
3311 (TYPES_d_unsigned): New macro.
3312 (TYPES_d_integer): Use it.
3313 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3314 (TYPES_cvt_narrow): Likewise.
3315 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3316 (preds_mx): New variable.
3317 (function_builder::add_overloaded_function): Allow the new feature
3318 set to be more restrictive than the original one.
3319 (function_resolver::infer_pointer_type): Remove qualifiers from
3320 the pointer type before printing it.
3321 (function_resolver::require_matching_pointer_type): New function.
3322 (function_resolver::resolve_sv_displacement): Handle functions
3323 that don't support 32-bit vector indices or svint32_t vector offsets.
3324 (function_resolver::finish_opt_n_resolution): Take the inferred type
3325 as a separate argument.
3326 (function_resolver::resolve_unary): Optionally treat all forms in
3327 the same way as normal merging functions.
3328 (gimple_folder::redirect_call): New function.
3329 (function_expander::prepare_gather_address_operands): Add an argument
3330 that says whether scaled forms are available. If they aren't,
3331 handle scaling of vector indices and don't add the extension and
3333 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3334 fall back to using cond_* instead.
3335 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3336 Split out the member variables into...
3337 (rtx_code_function_base): ...this new base class.
3338 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3339 (unspec_based_function): Split out the member variables into...
3340 (unspec_based_function_base): ...this new base class.
3341 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3342 (unspec_based_function_exact_insn): New class.
3343 (unspec_based_add_function, unspec_based_add_lane_function)
3344 (unspec_based_lane_function, unspec_based_pred_function)
3345 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3346 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3347 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3349 (unspec_based_fused_function): New class.
3350 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3351 (unspec_based_fused_lane_function): New class.
3352 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3354 (CODE_FOR_MODE1): New macro.
3355 (fixed_insn_function): New class.
3356 (while_comparison): Likewise.
3357 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3358 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3359 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3360 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3361 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3362 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3363 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3364 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3365 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3366 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3367 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3368 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3369 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3370 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3371 Also add an initial argument for unary_convert_narrowt, regardless
3372 of the predication type.
3373 (build_32_64): Allow loads and stores to specify MODE_none.
3374 (build_sv_index64, build_sv_uint_offset): New functions.
3375 (long_type_suffix): New function.
3376 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3377 (binary_imm_long_base, load_gather_sv_base): Likewise.
3378 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3379 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3380 (unary_narrowb_base, unary_narrowt_base): Likewise.
3381 (binary_long_lane_def, binary_long_lane): New shape.
3382 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3383 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3384 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3385 (binary_to_uint_def, binary_to_uint): Likewise.
3386 (binary_wide_def, binary_wide): Likewise.
3387 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3388 (compare_def, compare): Likewise.
3389 (compare_ptr_def, compare_ptr): Likewise.
3390 (load_ext_gather_index_restricted_def,
3391 load_ext_gather_index_restricted): Likewise.
3392 (load_ext_gather_offset_restricted_def,
3393 load_ext_gather_offset_restricted): Likewise.
3394 (load_gather_sv_def): Inherit from load_gather_sv_base.
3395 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3396 (shift_left_imm_def, shift_left_imm): Likewise.
3397 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3398 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3399 (store_scatter_index_restricted_def,
3400 store_scatter_index_restricted): Likewise.
3401 (store_scatter_offset_restricted_def,
3402 store_scatter_offset_restricted): Likewise.
3403 (tbl_tuple_def, tbl_tuple): Likewise.
3404 (ternary_long_lane_def, ternary_long_lane): Likewise.
3405 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3406 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3407 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3408 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3409 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3410 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3411 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3412 (ternary_uint_def, ternary_uint): Likewise.
3413 (unary_convert): Fix typo in comment.
3414 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3415 (unary_long_def, unary_long): Likewise.
3416 (unary_narrowb_def, unary_narrowb): Likewise.
3417 (unary_narrowt_def, unary_narrowt): Likewise.
3418 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3419 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3420 (unary_to_int_def, unary_to_int): Likewise.
3421 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3422 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3423 (svasrd_impl): Delete.
3424 (svcadd_impl::expand): Handle integer operations too.
3425 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3426 new functions to derive the unspec numbers.
3427 (svmla_svmls_lane_impl): Replace with...
3428 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3429 integer operations too.
3430 (svwhile_impl): Rename to...
3431 (svwhilelx_impl): ...this and inherit from while_comparison.
3432 (svasrd): Use unspec_based_function.
3433 (svmla_lane): Use svmla_lane_impl.
3434 (svmls_lane): Use svmls_lane_impl.
3435 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3436 (svwhilele, svwhilelt): Use svwhilelx_impl.
3437 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3438 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3439 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3440 * config/aarch64/aarch64-sve-builtins.def: Include
3441 aarch64-sve-builtins-sve2.def.
3443 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3445 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3446 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3447 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3448 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3449 immediates as well as vector ones.
3450 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3451 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3452 (aarch64_sve_qsub_immediate): Update calls accordingly.
3454 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3456 * config/aarch64/aarch64-sve2.md: Add banner comments.
3457 (<su>mulh<r>s<mode>3): Move further up file.
3458 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3459 (*aarch64_sve2_sra<mode>): Move further down file.
3460 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3462 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3464 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3466 (while_optab_cmp): Handle them.
3467 * config/aarch64/aarch64-sve.md
3468 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3469 and add a "@" marker.
3470 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3471 instead of gen_aarch64_sve2_while_ptest.
3472 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3474 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3476 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3477 (UNSPEC_WHILELE): ...this.
3478 (UNSPEC_WHILE_LO): Rename to...
3479 (UNSPEC_WHILELO): ...this.
3480 (UNSPEC_WHILE_LS): Rename to...
3481 (UNSPEC_WHILELS): ...this.
3482 (UNSPEC_WHILE_LT): Rename to...
3483 (UNSPEC_WHILELT): ...this.
3484 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3485 (cmp_op, while_optab_cmp): Likewise.
3486 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3487 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3488 (svwhilelt): Likewise.
3490 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3492 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3493 (unary_to_uint): Define.
3494 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3495 (unary_count): Rename to...
3496 (unary_to_uint_def, unary_to_uint): ...this.
3497 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3499 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3501 * config/aarch64/aarch64-sve-builtins-functions.h
3502 (code_for_mode_function): New class.
3503 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3504 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3505 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3506 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3507 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3509 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3511 * config/aarch64/iterators.md (addsub): New code attribute.
3512 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3514 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3515 in the asm string and attributes. Fix indentation.
3516 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3518 (@aarch64_sve_<optab><mode>): ...this.
3519 * config/aarch64/aarch64-sve-builtins.h
3520 (function_expander::expand_signed_unpred_op): Delete.
3521 * config/aarch64/aarch64-sve-builtins.cc
3522 (function_expander::expand_signed_unpred_op): Likewise.
3523 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3524 try using code_for_aarch64_sve instead.
3525 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3526 (svqsub_impl): Likewise.
3527 (svqadd, svqsub): Use rtx_code_function instead.
3529 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3531 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3532 (HADDSUB, sur, addsub): Remove them.
3534 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3536 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3537 functions the same way, rather than singling out those that
3538 aren't mapped directly to optabs.
3540 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3542 * target.def (compatible_vector_types_p): New target hook.
3543 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3544 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3545 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3546 * doc/tm.texi: Regenerate.
3547 * gimple-expr.c: Include target.h.
3548 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3549 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3551 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3552 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3553 Use the original predicate if it already has a suitable type.
3555 2020-01-09 Martin Jambor <mjambor@suse.cz>
3557 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3558 resolve_speculation and redirect_call_stmt_to_callee static. Change
3559 return type of set_call_stmt to cgraph_edge *.
3560 * auto-profile.c (afdo_indirect_call): Adjust call to
3561 redirect_call_stmt_to_callee.
3562 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3563 make the this pointer explicit, adjust self-recursive calls and the
3564 call top make_direct. Return the resulting edge.
3565 (cgraph_edge::remove): Make this pointer explicit.
3566 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3567 (cgraph_edge::make_direct): Likewise, adjust call to
3568 resolve_speculation.
3569 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3570 call to set_call_stmt.
3571 (cgraph_update_edges_for_call_stmt_node): Update call to
3572 set_call_stmt and remove.
3573 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3574 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3575 (cgraph_node::create_edge_including_clones): Moved "first" definition
3576 of edge to the block where it was used. Adjusted calls to
3578 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3579 cgraph_edge::remove.
3580 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3581 make_direct and redirect_call_stmt_to_callee.
3582 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3583 resolve_speculation and make_direct.
3584 * ipa-inline-transform.c (inline_transform): Adjust call to
3585 redirect_call_stmt_to_callee.
3586 (check_speculations_1):: Adjust call to resolve_speculation.
3587 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3588 resolve-speculation.
3589 (inline_small_functions): Adjust call to resolve_speculation.
3590 (ipa_inline): Likewise.
3591 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3593 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3594 safe with regards to edge removal, adjust calls to
3595 redirect_call_stmt_to_callee.
3596 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3597 and redirect_call_stmt_to_callee.
3598 * multiple_target.c (create_dispatcher_calls): Adjust call to
3599 redirect_call_stmt_to_callee
3600 (redirect_to_specific_clone): Likewise.
3601 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3602 Adjust calls to cgraph_edge::remove.
3603 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3604 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3605 (expand_call_inline): Adjust call to cgraph_edge::remove.
3607 2020-01-09 Martin Liska <mliska@suse.cz>
3609 * params.opt: Set Optimization for
3610 param_max_speculative_devirt_maydefs.
3612 2020-01-09 Martin Sebor <msebor@redhat.com>
3616 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3618 2020-01-09 Martin Liska <mliska@suse.cz>
3620 * auto-profile.c (auto_profile): Use opt_for_fn
3622 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3623 (propagate_vals_across_arith_jfunc): Likewise.
3624 (hint_time_bonus): Likewise.
3625 (incorporate_penalties): Likewise.
3626 (good_cloning_opportunity_p): Likewise.
3627 (perform_estimation_of_a_value): Likewise.
3628 (estimate_local_effects): Likewise.
3629 (ipcp_propagate_stage): Likewise.
3630 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3631 (set_switch_stmt_execution_predicate): Likewise.
3632 (analyze_function_body): Likewise.
3633 * ipa-inline-analysis.c (offline_size): Likewise.
3634 * ipa-inline.c (early_inliner): Likewise.
3635 * ipa-prop.c (ipa_analyze_node): Likewise.
3636 (ipcp_transform_function): Likewise.
3637 * ipa-sra.c (process_scan_results): Likewise.
3638 (ipa_sra_summarize_function): Likewise.
3639 * params.opt: Rename ipcp-unit-growth to
3640 ipa-cp-unit-growth. Add Optimization for various
3641 IPA-related parameters.
3643 2020-01-09 Richard Biener <rguenther@suse.de>
3646 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3648 2020-01-09 Richard Biener <rguenther@suse.de>
3650 PR tree-optimization/93040
3651 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3653 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3655 * common/config/avr/avr-common.c (avr_option_optimization_table)
3656 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3658 2020-01-09 Martin Liska <mliska@suse.cz>
3660 * cgraphclones.c (symbol_table::materialize_all_clones):
3661 Use cgraph_node::dump_name.
3663 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3666 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3667 output_operand_lossage instead of gcc_unreachable.
3668 * doc/md.texi (riscv f constraint): Fix typo.
3671 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3672 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3673 CONST_SCALAR_INT_P instead of CONST_INT_P.
3674 (*subv<mode>4_1): Rename to ...
3675 (subv<mode>4_1): ... this.
3676 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3677 define_insn_and_split patterns.
3678 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3681 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3683 * vec.c (class selftest::count_dtor): New class.
3684 (selftest::test_auto_delete_vec): New test.
3685 (selftest::vec_c_tests): Call it.
3686 * vec.h (class auto_delete_vec): New class template.
3687 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3689 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3691 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3693 2020-01-08 Jim Wilson <jimw@sifive.com>
3695 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3696 use of TLS_MODEL_LOCAL_EXEC when not pic.
3698 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3700 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3703 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3706 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3707 *stack_protect_set_3 peephole2): Also check that the second
3708 insns source is general_operand.
3711 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3712 predicate for output operand instead of register_operand.
3713 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3714 memory destination and non-memory operands[2].
3716 2020-01-08 Martin Liska <mliska@suse.cz>
3718 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3719 ::dump_asm_name instead of (::name or ::asm_name).
3720 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3721 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3722 (analyze_functions): Likewise.
3723 (expand_all_functions): Likewise.
3724 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3725 (propagate_bits_across_jump_function): Likewise.
3726 (dump_profile_updates): Likewise.
3727 (ipcp_store_bits_results): Likewise.
3728 (ipcp_store_vr_results): Likewise.
3729 * ipa-devirt.c (dump_targets): Likewise.
3730 * ipa-fnsummary.c (analyze_function_body): Likewise.
3731 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3732 (process_hsa_functions): Likewise.
3733 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3734 (set_alias_uids): Likewise.
3735 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3736 * ipa-inline.c (recursive_inlining): Likewise.
3737 (inline_to_all_callers_1): Likewise.
3738 (ipa_inline): Likewise.
3739 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3740 (ipa_propagate_frequency): Likewise.
3741 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3742 (remove_described_reference): Likewise.
3743 * ipa-pure-const.c (worse_state): Likewise.
3744 (check_retval_uses): Likewise.
3745 (analyze_function): Likewise.
3746 (propagate_pure_const): Likewise.
3747 (propagate_nothrow): Likewise.
3748 (dump_malloc_lattice): Likewise.
3749 (propagate_malloc): Likewise.
3750 (pass_local_pure_const::execute): Likewise.
3751 * ipa-visibility.c (optimize_weakref): Likewise.
3752 (function_and_variable_visibility): Likewise.
3753 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3754 (ipa_discover_variable_flags): Likewise.
3755 * lto-streamer-out.c (output_function): Likewise.
3756 (output_constructor): Likewise.
3757 * tree-inline.c (copy_bb): Likewise.
3758 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3759 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3761 2020-01-08 Richard Biener <rguenther@suse.de>
3764 * tree-eh.c (sink_clobbers): Update virtual operands for
3765 the first and last stmt only. Add a dry-run capability.
3766 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3767 after CFG manipulations and in RPO order to catch all
3768 secondary opportunities reliably.
3770 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3773 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3775 2019-01-08 Richard Biener <rguenther@suse.de>
3778 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3779 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3780 virtual operand, also updating SSA use.
3781 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3782 Update stmt after resetting virtual operand.
3783 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3784 * gimple-iterator.c (gsi_remove): When not removing the stmt
3785 permanently do not delink immediate uses or mark the stmt modified.
3787 2020-01-08 Martin Liska <mliska@suse.cz>
3789 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3790 (ipa_call_context::estimate_size_and_time): Likewise.
3791 (inline_analyze_function): Likewise.
3793 2020-01-08 Martin Liska <mliska@suse.cz>
3795 * cgraph.c (cgraph_node::dump): Use systematically
3798 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3800 Add -nodevicespecs option for avr.
3803 * config/avr/avr.opt (-nodevicespecs): New driver option.
3804 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3805 "-specs=device-specs/..." if that option is not set.
3806 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3808 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3810 Implement 64-bit double functions for avr.
3813 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3814 --with-double-comparison.
3815 * doc/install.texi: Document them.
3816 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3817 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3818 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3819 * doc/invoke.texi (AVR Built-in Macros): Document them.
3820 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3821 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3822 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3824 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3827 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3828 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3829 when only building rm-profile multilibs.
3831 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3834 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3835 lattice for a value to check.
3836 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3837 finite propagation in self-recursive scc.
3839 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3841 * ipa-inline.c (caller_growth_limits): Restore the AND.
3843 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3845 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3846 (VEC_ALLREG_ALT): New iterator.
3847 (VEC_ALLREG_INT_MODE): New iterator.
3848 (VCMP_MODE): New iterator.
3849 (VCMP_MODE_INT): New iterator.
3850 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3851 (vec_cmp<u>v64qidi): New define_expand.
3852 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3853 (vec_cmpu<mode>di_exec): New define_expand.
3854 (vec_cmp<u>v64qidi_exec): New define_expand.
3855 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3856 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3857 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3858 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3859 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3860 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3861 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3862 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3863 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3864 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3866 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3867 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3869 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3871 * config/gcn/constraints.md (DA): Update description and match.
3873 (Db): New constraint.
3874 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3876 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3877 Implement 'Db' mixed immediate type.
3878 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3879 (addcv64si3_dup<exec_vcc>): Delete.
3880 (subcv64si3<exec_vcc>): Rework constraints.
3881 (addv64di3): Rework constraints.
3882 (addv64di3_exec): Rework constraints.
3883 (subv64di3): Rework constraints.
3884 (addv64di3_dup): Delete.
3885 (addv64di3_dup_exec): Delete.
3886 (addv64di3_zext): Rework constraints.
3887 (addv64di3_zext_exec): Rework constraints.
3888 (addv64di3_zext_dup): Rework constraints.
3889 (addv64di3_zext_dup_exec): Rework constraints.
3890 (addv64di3_zext_dup2): Rework constraints.
3891 (addv64di3_zext_dup2_exec): Rework constraints.
3892 (addv64di3_sext_dup2): Rework constraints.
3893 (addv64di3_sext_dup2_exec): Rework constraints.
3895 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3897 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3898 existing target checks.
3900 2020-01-07 Richard Biener <rguenther@suse.de>
3902 * doc/install.texi: Bump minimal supported MPC version.
3904 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3906 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3907 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3908 * langhooks.c: Include stor-layout.h.
3909 (lhd_simulate_enum_decl): New function.
3910 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3911 handle_arm_sve_h for the LTO frontend.
3912 (register_vector_type): Cope with null returns from pushdecl.
3914 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3916 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3917 (aarch64_sve::nvectors_if_data_type): Replace with...
3918 (aarch64_sve::builtin_type_p): ...this.
3919 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3920 (find_vector_type): Delete.
3921 (add_sve_type_attribute): New function.
3922 (lookup_sve_type_attribute): Likewise.
3923 (register_builtin_types): Add an "SVE type" attribute to each type.
3924 (register_tuple_type): Likewise.
3925 (svbool_type_p, nvectors_if_data_type): Delete.
3926 (mangle_builtin_type): Use lookup_sve_type_attribute.
3927 (builtin_type_p): Likewise. Add an overload that returns the
3928 number of constituent vector and predicate registers.
3929 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3930 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3931 instead of aarch64_sve_argument_p.
3932 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3933 (aarch64_pass_by_reference): Likewise.
3934 (aarch64_function_value_1): Likewise.
3935 (aarch64_return_in_memory): Likewise.
3936 (aarch64_layout_arg): Likewise.
3938 2020-01-07 Jakub Jelinek <jakub@redhat.com>
3940 PR tree-optimization/93156
3941 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3942 least significant bit is always clear.
3944 PR tree-optimization/93118
3945 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
3946 simplifier with two intermediate conversions.
3948 2020-01-07 Martin Liska <mliska@suse.cz>
3950 * params.opt: Add Optimization for various parameters.
3952 2020-01-07 Martin Liska <mliska@suse.cz>
3955 * doc/extend.texi: Explain cloning for target_clone
3958 2020-01-07 Martin Liska <mliska@suse.cz>
3960 PR tree-optimization/92860
3961 * common.opt: Make in Optimization option
3962 as it is affected by -O0, which is an Optimization
3964 * tree-inline.c (tree_inlinable_function_p):
3965 Use opt_for_fn for warn_inline.
3966 (expand_call_inline): Likewise.
3968 2020-01-07 Martin Liska <mliska@suse.cz>
3970 PR tree-optimization/92860
3971 * common.opt: Make flag_ree as optimization
3974 2020-01-07 Martin Liska <mliska@suse.cz>
3976 PR optimization/92860
3977 * params.opt: Mark param_min_crossjump_insns with Optimization
3980 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
3982 * ipa-inline-analysis.c (estimate_growth): Fix typo.
3983 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3985 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
3987 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3988 helper function to return the valid addressing formats for a given
3989 hard register and mode.
3990 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3992 * config/rs6000/constraints.md (Q constraint): Update
3994 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3997 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3998 Use 'Q' for doing vector extract from memory.
3999 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4001 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4002 doing vector extract from memory.
4003 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4004 extract from memory.
4006 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4007 for the offset being 34-bits when -mcpu=future is used.
4009 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4011 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4012 instead of cmpib_comparison_operator in cmpib patterns.
4013 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4014 of cmpib_comparison_operator. Revise comment.
4016 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4018 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4019 in an IFN_DIV_POW2 node to be equal.
4021 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4023 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4024 (vect_check_scalar_mask): ...this.
4025 (vectorizable_store, vectorizable_load): Update call accordingly.
4026 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4027 argument in calls to conditional internal functions.
4029 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4031 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4032 '0' matching inputs.
4033 (subv64di3_exec): Likewise.
4035 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4037 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4038 * doc/md.texi (movstr): Likewise.
4040 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4042 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4045 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4047 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4049 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4050 to a temporary file and use move-if-change to update the real
4051 file where necessary.
4053 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4055 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4056 rather than Upa for CPY /M.
4058 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4060 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4063 2020-01-06 Martin Liska <mliska@suse.cz>
4065 PR tree-optimization/92860
4066 * params.opt: Mark param_max_combine_insns with Optimization
4069 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4072 * config/i386/i386.md (SWIDWI): New mode iterator.
4073 (DWI, dwi): Add TImode variants.
4074 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4075 <general_hilo_operand> instead of <general_operand>. Use
4076 CONST_SCALAR_INT_P instead of CONST_INT_P.
4077 (*addv<mode>4_1): Rename to ...
4078 (addv<mode>4_1): ... this.
4079 (QWI): New mode attribute.
4080 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4081 define_insn_and_split patterns.
4082 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4084 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4085 <general_hilo_operand> instead of <general_operand>.
4086 (*addcarry<mode>_1): New define_insn.
4087 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4089 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4091 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4092 Use "call" instead of "set".
4094 2020-01-03 Martin Jambor <mjambor@suse.cz>
4097 * ipa-cp.c (print_all_lattices): Skip functions without info.
4099 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4102 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4103 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4104 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4105 for 'e' simd clones.
4108 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4110 (mprefer-vector-width=): Add Save.
4111 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4112 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4113 (ix86_debug_options, ix86_function_specific_print): Adjust
4114 ix86_target_string callers.
4115 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4116 (ix86_valid_target_attribute_tree): Likewise.
4117 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4118 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4119 ix86_target_string caller.
4122 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4123 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4124 instead of gen_int_shift_amount + convert_modes.
4126 PR rtl-optimization/93088
4127 * loop-iv.c (find_single_def_src): Punt after looking through
4128 128 reg copies for regs with single definitions. Move definitions
4131 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4133 * config/arm/arm-c.c (arm_cpu_builtins): Define
4134 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4135 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4136 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4137 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4138 * config/arm/arm-tables.opt: Regenerated.
4139 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4140 arm_arch_i8mm and arm_arch_bf16 when enabled.
4141 * config/arm/arm.h (TARGET_I8MM): New macro.
4142 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4143 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4144 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4145 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4146 (v8_6_a_simd_variants): New.
4147 (v8_*_a_simd_variants): Add i8mm and bf16.
4148 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4150 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4153 * predict.c (compute_function_frequency): Don't call
4154 warn_function_cold on functions that already have cold attribute.
4156 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4159 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4160 COMDAT group function labels in .data.rel.ro.local section.
4161 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4164 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4165 comparison_operator in B and S integer comparisons. Likewise, use
4166 ordered_comparison_operator instead of cmpib_comparison_operator in
4168 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4170 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4172 Update copyright years.
4174 * gcc.c (process_command): Update copyright notice dates.
4175 * gcov-dump.c (print_version): Ditto.
4176 * gcov.c (print_version): Ditto.
4177 * gcov-tool.c (print_version): Ditto.
4178 * gengtype.c (create_file): Ditto.
4179 * doc/cpp.texi: Bump @copying's copyright year.
4180 * doc/cppinternals.texi: Ditto.
4181 * doc/gcc.texi: Ditto.
4182 * doc/gccint.texi: Ditto.
4183 * doc/gcov.texi: Ditto.
4184 * doc/install.texi: Ditto.
4185 * doc/invoke.texi: Ditto.
4187 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4189 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4192 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4194 PR tree-optimization/93098
4195 * match.pd (popcount): For shift amounts, use integer_onep
4196 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4197 tests. Make sure that precision is power of two larger than or equal
4198 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4199 instead of ULL suffixed constants. Formatting fixes.
4201 Copyright (C) 2020 Free Software Foundation, Inc.
4203 Copying and distribution of this file, with or without modification,
4204 are permitted in any medium without royalty provided the copyright
4205 notice and this notice are preserved.