1 2020-02-17 Alexandre Oliva <oliva@adacore.com>
3 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
4 handling of register_common-less targets.
6 2020-02-17 Martin Liska <mliska@suse.cz>
9 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
11 2020-02-17 Martin Liska <mliska@suse.cz>
14 * config/rs6000/rs6000.c (rs6000_option_override_internal):
17 2020-02-17 Martin Liska <mliska@suse.cz>
20 * config/rx/elf.opt: Fix typo.
22 2020-02-17 Richard Biener <rguenther@suse.de>
25 * opts-global.c (print_ignored_options): Use inform and
28 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
31 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
33 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
36 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
37 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
39 2020-02-15 Jason Merrill <jason@redhat.com>
41 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
43 2020-02-15 Jakub Jelinek <jakub@redhat.com>
45 PR tree-optimization/93744
46 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
47 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
48 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
49 sure @2 in the first and @1 in the other patterns has no side-effects.
51 2020-02-15 David Malcolm <dmalcolm@redhat.com>
52 Bernd Edlinger <bernd.edlinger@hotmail.de>
56 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
57 * configure.ac (--with-diagnostics-urls): New configuration
58 option, based on --with-diagnostics-color.
59 (DIAGNOSTICS_URLS_DEFAULT): New define.
60 * config.h: Regenerate.
61 * configure: Regenerate.
62 * diagnostic.c (diagnostic_urls_init): Handle -1 for
63 DIAGNOSTICS_URLS_DEFAULT from configure-time
64 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
65 and TERM_URLS environment variable.
66 * diagnostic-url.h (diagnostic_url_format): New enum type.
67 (diagnostic_urls_enabled_p): rename to...
68 (determine_url_format): ... this, and change return type.
69 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
70 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
71 the linux console, and mingw.
72 (diagnostic_urls_enabled_p): rename to...
73 (determine_url_format): ... this, and adjust.
74 * pretty-print.h (pretty_printer::show_urls): rename to...
75 (pretty_printer::url_format): ... this, and change to enum.
76 * pretty-print.c (pretty_printer::pretty_printer,
77 pp_begin_url, pp_end_url, test_urls): Adjust.
78 * doc/install.texi (--with-diagnostics-urls): Document the new
80 (--with-diagnostics-color): Document the existing interaction
81 with GCC_COLORS better.
82 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
83 vindex reference. Update description of defaults based on the above.
84 (-fdiagnostics-color): Update description of how -fdiagnostics-color
85 interacts with GCC_COLORS.
87 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
90 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
91 conjunction with TARGET_GNU_TLS in early return.
93 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
95 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
96 the mode is not wider than UNITS_PER_WORD.
98 2020-02-14 Martin Jambor <mjambor@suse.cz>
100 PR tree-optimization/93516
101 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
102 access of the same type as the parent.
103 (propagate_subaccesses_from_lhs): Likewise.
105 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
108 * config/i386/avx512vbmi2intrin.h
109 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
110 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
111 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
112 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
113 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
114 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
115 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
116 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
117 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
118 of lacking a closing parenthesis.
119 * config/i386/avx512vbmi2vlintrin.h
120 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
121 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
122 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
123 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
124 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
125 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
126 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
127 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
128 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
129 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
130 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
131 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
132 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
133 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
134 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
135 _mm_shldi_epi32, _mm_mask_shldi_epi32,
136 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
137 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
139 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
142 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
143 the target function entry.
145 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
147 * common/config/arc/arc-common.c (arc_option_optimization_table):
148 Disable if-conversion step when optimized for size.
150 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
152 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
153 R12-R15 are always in ARCOMPACT16_REGS register class.
154 * config/arc/arc.opt (mq-class): Deprecate.
155 * config/arc/constraint.md ("q"): Remove dependency on mq-class
157 * doc/invoke.texi (mq-class): Update text.
158 * common/config/arc/arc-common.c (arc_option_optimization_table):
161 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
163 * config/arc/arc.c (arc_insn_cost): New function.
164 (TARGET_INSN_COST): Define.
165 * config/arc/arc.md (cost): New attribute.
166 (add_n): Use arc_nonmemory_operand.
167 (ashlsi3_insn): Likewise, also update constraints.
168 (ashrsi3_insn): Likewise.
170 (add_shift): Likewise.
171 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
173 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
175 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
177 (umulsidi_600): Likewise.
179 2020-02-13 Jakub Jelinek <jakub@redhat.com>
182 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
183 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
184 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
185 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
186 pass __A to the builtin followed by __W instead of __A followed by
188 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
189 _mm512_mask_popcnt_epi64): Likewise.
190 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
191 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
192 _mm256_mask_popcnt_epi64): Likewise.
194 PR tree-optimization/93582
195 * fold-const.h (shift_bytes_in_array_left,
196 shift_bytes_in_array_right): Declare.
197 * fold-const.c (shift_bytes_in_array_left,
198 shift_bytes_in_array_right): New function, moved from
199 gimple-ssa-store-merging.c, no longer static.
200 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
201 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
202 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
203 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
204 shift_bytes_in_array.
205 (verify_shift_bytes_in_array): Rename to ...
206 (verify_shift_bytes_in_array_left): ... this. Use
207 shift_bytes_in_array_left instead of shift_bytes_in_array.
208 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
209 instead of verify_shift_bytes_in_array.
210 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
211 / native_interpret_expr where the store covers all needed bits,
212 punt on PDP-endian, otherwise allow all involved offsets and sizes
213 not to be byte-aligned.
216 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
217 use const_0_to_255_operand predicate instead of immediate_operand.
218 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
219 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
220 vgf2p8affineinvqb_<mode><mask_name>,
221 vgf2p8affineqb_<mode><mask_name>): Drop mode from
222 const_0_to_255_operand predicated operands.
224 2020-02-12 Jeff Law <law@redhat.com>
226 * config/h8300/h8300.md (comparison shortening peepholes): Use
227 a mode iterator to merge the HImode and SImode peepholes.
229 2020-02-12 Jakub Jelinek <jakub@redhat.com>
232 * real.c (is_even): Make static. Function comment fix.
233 (is_halfway_below): Make static, don't assert R is not inf/nan,
234 instead return false for those. Small formatting fixes.
236 2020-02-12 Martin Sebor <msebor@redhat.com>
239 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
240 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
241 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
242 (strlen_check_and_optimize_call): Adjust callee name.
244 2020-02-12 Jeff Law <law@redhat.com>
246 * config/h8300/h8300.md (comparison shortening peepholes): Drop
247 (and (xor)) variant. Combine other two into single peephole.
249 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
251 PR rtl-optimization/93565
252 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
254 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
256 * config/aarch64/aarch64-simd.md
257 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
258 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
259 generating separate ADDV and zero_extend patterns.
260 * config/aarch64/iterators.md (VDQV_E): New iterator.
262 2020-02-12 Jeff Law <law@redhat.com>
264 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
265 expanders, splits, etc.
266 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
267 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
268 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
269 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
270 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
271 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
273 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
275 2020-02-12 Jakub Jelinek <jakub@redhat.com>
278 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
279 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
280 TARGET_AVX512DQ from condition.
281 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
282 instead of <mask_mode512bit_condition> in condition. If
283 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
285 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
288 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
291 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
293 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
295 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
296 where strlen is more legible.
297 (rs6000_builtin_vectorized_libmass): Ditto.
298 (rs6000_print_options_internal): Ditto.
300 2020-02-11 Martin Sebor <msebor@redhat.com>
302 PR tree-optimization/93683
303 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
305 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
307 * config/rs6000/predicates.md (cint34_operand): Rename the
308 -mprefixed-addr option to be -mprefixed.
309 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
310 the -mprefixed-addr option to be -mprefixed.
311 (OTHER_FUTURE_MASKS): Likewise.
312 (POWERPC_MASKS): Likewise.
313 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
314 the -mprefixed-addr option to be -mprefixed. Change error
315 messages to refer to -mprefixed.
316 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
318 (rs6000_legitimate_offset_address_p): Likewise.
319 (rs6000_mode_dependent_address): Likewise.
320 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
321 "-mprefixed" for target attributes and pragmas.
322 (address_to_insn_form): Rename the -mprefixed-addr option to be
324 (rs6000_adjust_insn_length): Likewise.
325 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
326 -mprefixed-addr option to be -mprefixed.
327 (ASM_OUTPUT_OPCODE): Likewise.
328 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
329 -mprefixed-addr option to be -mprefixed.
330 * config/rs6000/rs6000.opt (-mprefixed): Rename the
331 -mprefixed-addr option to be prefixed. Change the option from
332 being undocumented to being documented.
333 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
334 -mprefixed option. Update the -mpcrel documentation to mention
337 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
339 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
340 including FIRST_PSEUDO_REGISTER - 1.
341 * ira-color.c (print_hard_reg_set): Ditto.
343 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
345 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
346 (USTERNOP_QUALIFIERS): New define.
347 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
348 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
349 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
350 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
351 * config/arm/arm_neon.h (vusdot_s32): New.
352 (vusdot_lane_s32): New.
353 (vusdotq_lane_s32): New.
354 (vsudot_lane_s32): New.
355 (vsudotq_lane_s32): New.
356 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
357 * config/arm/iterators.md (DOTPROD_I8MM): New.
358 (sup, opsuffix): Add <us/su>.
359 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
360 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
362 2020-02-11 Richard Biener <rguenther@suse.de>
364 PR tree-optimization/93661
365 PR tree-optimization/93662
366 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
368 * tree-sra.c (get_access_for_expr): Likewise.
370 2020-02-10 Jakub Jelinek <jakub@redhat.com>
373 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
374 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
375 Change condition from TARGET_AVX2 to TARGET_AVX.
377 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
380 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
383 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
385 Try to generate zero-based comparisons.
386 * config/cris/cris.c (cris_reduce_compare): New function.
387 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
388 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
389 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
391 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
394 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
395 in Thumb state and also as a destination in Arm state. Add T16
398 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
400 * md.texi (Define Subst): Match closing paren in example.
402 2020-02-10 Jakub Jelinek <jakub@redhat.com>
406 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
407 arguments of strncmp.
409 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
412 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
413 but different source value.
414 (adjust_callers_for_value_intersection): New function.
415 (gather_edges_for_value): Adjust order of callers to let a
416 non-self-recursive caller be the first element.
417 (self_recursive_pass_through_p): Add a new parameter "simple", and
418 check generalized self-recursive pass-through jump function.
419 (self_recursive_agg_pass_through_p): Likewise.
420 (find_more_scalar_values_for_callers_subset): Compute value from
421 pass-through jump function for self-recursive.
422 (intersect_with_plats): Cleanup previous implementation code for value
423 itersection with self-recursive call edge.
424 (intersect_with_agg_replacements): Likewise.
425 (intersect_aggregates_with_edge): Deduce value from pass-through jump
426 function for self-recursive call edge. Cleanup previous implementation
427 code for value intersection with self-recursive call edge.
428 (decide_whether_version_node): Remove dead callers and adjust order
429 to let a non-self-recursive caller be the first element.
431 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
433 * recog.c: Move pass_split_before_sched2 code in front of
434 pass_split_before_regstack.
435 (pass_data_split_before_sched2): Rename pass to split3 from split4.
436 (pass_data_split_before_regstack): Rename pass to split4 from split3.
437 (rest_of_handle_split_before_sched2): Remove.
438 (pass_split_before_sched2::execute): Unconditionally call
440 (enable_split_before_sched2): New function.
441 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
442 (pass_split_before_regstack::gate): Ditto.
443 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
444 Update name check for renamed split4 pass.
445 * config/sh/sh.c (register_sh_passes): Update pass insertion
446 point for renamed split4 pass.
448 2020-02-09 Jakub Jelinek <jakub@redhat.com>
450 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
451 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
452 copying them around between host and target.
454 2020-02-08 Andrew Pinski <apinski@marvell.com>
457 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
458 STRICT_ALIGNMENT also.
460 2020-02-08 Jim Wilson <jimw@sifive.com>
463 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
465 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
466 Jakub Jelinek <jakub@redhat.com>
469 * config/i386/i386.h (CALL_USED_REGISTERS): Make
470 xmm16-xmm31 call-used even in 64-bit ms-abi.
472 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
474 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
475 (simd_ummla, simd_usmmla): Likewise.
476 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
477 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
480 2020-02-07 Richard Biener <rguenther@suse.de>
483 * tree-inline.c (fold_marked_statements): Do a PRE walk,
484 skipping unreachable regions.
485 (optimize_inline_calls): Skip folding stmts when we didn't
488 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
491 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
492 Don't return aggregates with only SFmode and DFmode in SSE
494 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
496 2020-02-07 Jakub Jelinek <jakub@redhat.com>
499 * config/rs6000/rs6000-logue.c
500 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
501 if it fails, move rs into end_addr and retry. Add
502 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
503 the insn pattern doesn't describe well what exactly happens to
507 * config/i386/predicates.md (avx_identity_operand): Remove.
508 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
509 (avx_<castmode><avxsizesuffix>_<castmode>,
510 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
511 a VEC_CONCAT of the operand and UNSPEC_CAST.
512 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
513 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
517 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
518 recog_data.insn if distance_non_agu_define changed it.
520 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
523 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
524 we only had X-FORM (reg+reg) addressing for vectors. Also before
525 ISA 3.0, we only had X-FORM addressing for scalars in the
526 traditional Altivec registers.
528 2020-02-06 <zhongyunde@huawei.com>
529 Vladimir Makarov <vmakarov@redhat.com>
531 PR rtl-optimization/93561
532 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
535 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
537 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
540 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
542 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
543 where the low and the high 32 bits are equal to each other specially,
544 with an rldimi instruction.
546 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
548 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
550 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
552 * config/arm/arm-tables.opt: Regenerate.
554 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
557 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
558 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
559 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
561 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
563 PR rtl-optimization/87763
564 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
566 2020-02-06 Delia Burduv <delia.burduv@arm.com>
568 * config/aarch64/aarch64-simd-builtins.def
569 (bfmlaq): New built-in function.
570 (bfmlalb): New built-in function.
571 (bfmlalt): New built-in function.
572 (bfmlalb_lane): New built-in function.
573 (bfmlalt_lane): New built-in function.
574 * config/aarch64/aarch64-simd.md
575 (aarch64_bfmmlaqv4sf): New pattern.
576 (aarch64_bfmlal<bt>v4sf): New pattern.
577 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
578 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
579 (vbfmlalbq_f32): New intrinsic.
580 (vbfmlaltq_f32): New intrinsic.
581 (vbfmlalbq_lane_f32): New intrinsic.
582 (vbfmlaltq_lane_f32): New intrinsic.
583 (vbfmlalbq_laneq_f32): New intrinsic.
584 (vbfmlaltq_laneq_f32): New intrinsic.
585 * config/aarch64/iterators.md (BF_MLA): New int iterator.
586 (bt): New int attribute.
588 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
590 * config/i386/i386.md (*pushtf): Emit "#" instead of
591 calling gcc_unreachable in insn output.
594 (*pushsf_rex64): Ditto for alternatives other than 1.
595 (*pushsf): Ditto for alternatives other than 1.
597 2020-02-06 Martin Liska <mliska@suse.cz>
599 PR gcov-profile/91971
600 PR gcov-profile/93466
601 * coverage.c (coverage_init): Revert mangling of
602 path into filename. It can lead to huge filename length.
603 Creation of subfolders seem more natural.
605 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
608 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
609 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
610 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
612 2020-02-06 Jakub Jelinek <jakub@redhat.com>
615 * config/i386/predicates.md (avx_identity_operand): New predicate.
616 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
617 define_insn_and_split.
620 * omp-low.c (use_pointer_for_field): For nested constructs, also
621 look for map clauses on target construct.
622 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
623 taskreg_nesting_level.
626 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
627 shared clause, call omp_notice_variable on outer context if any.
629 2020-02-05 Jason Merrill <jason@redhat.com>
632 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
633 non-zero address even if weak and not yet defined.
635 2020-02-05 Martin Sebor <msebor@redhat.com>
637 PR tree-optimization/92765
638 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
639 * tree-ssa-strlen.c (compute_string_length): Remove.
640 (determine_min_objsize): Remove.
641 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
642 Avoid using type size as the upper bound on string length.
643 (handle_builtin_string_cmp): Add an argument. Adjust.
644 (strlen_check_and_optimize_call): Pass additional argument to
645 handle_builtin_string_cmp.
647 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
649 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
650 (*pushdi2_rex64 peephole2): Unconditionally split after
652 (*ashl<mode>3_doubleword): Ditto.
653 (*<shift_insn><mode>3_doubleword): Ditto.
655 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
658 * config/rs6000/rs6000.c (get_vector_offset): Fix
660 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
662 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
664 2020-02-05 David Malcolm <dmalcolm@redhat.com>
667 (Special Functions for Debugging the Analyzer): Update description
668 of __analyzer_dump_exploded_nodes.
670 2020-02-05 Jakub Jelinek <jakub@redhat.com>
673 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
674 include sets and not clobbers in the vzeroupper pattern.
675 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
676 the parallel has 17 (64-bit) or 9 (32-bit) elts.
677 (*avx_vzeroupper_1): New define_insn_and_split.
680 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
681 don't run when !optimize.
682 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
685 2020-02-05 Richard Biener <rguenther@suse.de>
688 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
689 checks before matching calls.
691 2020-02-05 Jakub Jelinek <jakub@redhat.com>
693 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
694 function comment typo.
697 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
698 simd_clone_create failed when i == 0, adjust clone->nargs by
701 2020-02-05 Martin Liska <mliska@suse.cz>
704 * doc/invoke.texi: Document that one should
705 not combine ASLR and -fpch.
707 2020-02-04 Richard Biener <rguenther@suse.de>
709 PR tree-optimization/93538
710 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
712 2020-02-04 Richard Biener <rguenther@suse.de>
714 PR tree-optimization/91123
715 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
716 (vn_walk_cb_data::last_vuse): New member.
717 (vn_walk_cb_data::saved_operands): Likewsie.
718 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
719 (vn_walk_cb_data::push_partial_def): Use finish.
720 (vn_reference_lookup_2): Update last_vuse and use finish if
721 we've saved operands.
722 (vn_reference_lookup_3): Use finish and update calls to
723 push_partial_defs everywhere. When translating through
724 memcpy or aggregate copies save off operands and alias-set.
725 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
726 operation for redundant store removal.
728 2020-02-04 Richard Biener <rguenther@suse.de>
730 PR tree-optimization/92819
731 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
732 generating more stmts than before.
734 2020-02-04 Martin Liska <mliska@suse.cz>
736 * config/arm/arm.c (arm_gen_far_branch): Move the function
737 outside of selftests.
739 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
741 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
742 function to adjust PC-relative vector addresses.
743 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
744 handle vectors with PC-relative addresses.
746 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
748 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
750 (hard_reg_and_mode_to_addr_mask): Delete.
751 (rs6000_adjust_vec_address): If the original vector address
752 was REG+REG or REG+OFFSET and the element is not zero, do the add
753 of the elements in the original address before adding the offset
754 for the vector element. Use address_to_insn_form to validate the
755 address using the register being loaded, rather than guessing
756 whether the address is a DS-FORM or DQ-FORM address.
758 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
760 * config/rs6000/rs6000.c (get_vector_offset): New helper function
761 to calculate the offset in memory from the start of a vector of a
762 particular element. Add code to keep the element number in
763 bounds if the element number is variable.
764 (rs6000_adjust_vec_address): Move calculation of offset of the
765 vector element to get_vector_offset.
766 (rs6000_split_vec_extract_var): Do not do the initial AND of
767 element here, move the code to get_vector_offset.
769 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
771 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
774 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
776 * config/rs6000/constraints.md: Improve documentation.
778 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
781 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
782 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
784 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
786 * config.gcc: Remove "carrizo" support.
787 * config/gcn/gcn-opts.h (processor_type): Likewise.
788 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
789 * config/gcn/gcn.opt (gpu_type): Likewise.
790 * config/gcn/t-omp-device: Likewise.
792 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
795 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
796 * config/arm/arm.c (arm_gen_far_branch): New function
798 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
800 2020-02-03 Julian Brown <julian@codesourcery.com>
801 Tobias Burnus <tobias@codesourcery.com>
803 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
805 2020-02-03 Jakub Jelinek <jakub@redhat.com>
808 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
809 valid RTL to sum up the lowest and second lowest bytes of the popcnt
812 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
814 PR rtl-optimization/91333
815 * ira-color.c (struct allocno_color_data): Add member
817 (init_allocno_threads): Set the member up.
818 (bucket_allocno_compare_func): Add compare hard reg
821 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
823 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
825 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
826 * config.in: Regenerated.
827 * configure: Regenerated.
828 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
829 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
830 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
832 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
834 * configure: Regenerate.
836 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
838 PR rtl-optimization/91333
839 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
840 reg preferences comparison up.
842 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
844 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
845 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
846 aarch64-sve-builtins-base.h.
847 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
848 aarch64-sve-builtins-base.cc.
849 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
850 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
852 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
853 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
854 (svcvtnt): New functions.
855 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
856 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
857 (svcvtnt): New functions.
858 (svcvt): Add a form that converts f32 to bf16.
859 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
860 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
862 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
863 Treat B as bfloat16_t.
864 (ternary_bfloat_lane_base): New class.
865 (ternary_bfloat_def): Likewise.
866 (ternary_bfloat): New shape.
867 (ternary_bfloat_lane_def): New class.
868 (ternary_bfloat_lane): New shape.
869 (ternary_bfloat_lanex2_def): New class.
870 (ternary_bfloat_lanex2): New shape.
871 (ternary_bfloat_opt_n_def): New class.
872 (ternary_bfloat_opt_n): New shape.
873 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
874 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
875 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
876 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
877 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
878 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
879 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
880 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
881 the pattern off the narrow mode instead of the wider one.
882 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
883 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
884 (sve_fp_op): Handle them.
885 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
886 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
888 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
890 * config/aarch64/arm_sve.h: Include arm_bf16.h.
891 * config/aarch64/aarch64-modes.def (BF): Move definition before
892 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
893 (SVE_MODES): Handle BF modes.
894 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
896 (aarch64_full_sve_mode): Likewise.
897 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
899 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
900 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
901 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
902 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
904 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
906 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
907 (TYPES_all_data): Add bf16.
908 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
909 (register_tuple_type): Increase buffer size.
910 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
911 (bf16): New type suffix.
912 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
913 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
914 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
915 Change type from all_data to all_arith.
916 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
919 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
920 Matthew Malcomson <matthew.malcomson@arm.com>
921 Richard Sandiford <richard.sandiford@arm.com>
923 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
924 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
925 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
926 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
927 __ARM_FEATURE_MATMUL_FP64.
928 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
929 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
930 be disabled at the same time.
931 (f32mm): New extension.
932 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
933 (AARCH64_FL_F64MM): Bump to the next bit up.
934 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
935 (TARGET_SVE_F64MM): New macros.
936 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
937 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
938 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
939 (UNSPEC_ZIP2Q): New unspeccs.
940 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
941 (optab, sur, perm_insn): Handle the new unspecs.
942 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
943 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
944 TARGET_SVE_F64MM instead of separate tests.
945 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
946 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
947 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
948 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
949 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
950 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
951 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
952 (TYPES_s_signed): New macro.
953 (TYPES_s_integer): Use it.
954 (TYPES_d_float): New macro.
955 (TYPES_d_data): Use it.
956 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
957 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
958 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
959 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
961 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
963 (ternary_resize2_lane_base): Likewise.
964 (ternary_resize2_base): New class.
965 (ternary_qq_lane_base): Likewise.
966 (ternary_intq_uintq_lane_def): Likewise.
967 (ternary_intq_uintq_lane): New shape.
968 (ternary_intq_uintq_opt_n_def): New class
969 (ternary_intq_uintq_opt_n): New shape.
970 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
971 (ternary_uintq_intq_def): New class.
972 (ternary_uintq_intq): New shape.
973 (ternary_uintq_intq_lane_def): New class.
974 (ternary_uintq_intq_lane): New shape.
975 (ternary_uintq_intq_opt_n_def): New class.
976 (ternary_uintq_intq_opt_n): New shape.
977 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
978 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
979 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
980 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
982 (svdotprod_lane_impl): ...this new class.
983 (svmmla_impl, svusdot_impl): New classes.
984 (svdot_lane): Update to use svdotprod_lane_impl.
985 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
986 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
988 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
989 function, with no types defined.
990 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
991 AARCH64_FL_I8MM functions.
992 (svmmla): New AARCH64_FL_F32MM function.
993 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
994 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
995 AARCH64_FL_F64MM function.
996 (REQUIRED_EXTENSIONS):
998 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1000 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1003 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1005 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1006 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1007 (*movti_internal): Do not check for
1008 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1009 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1010 just after check for TARGET_AVX.
1011 (*movdf_internal): Ditto.
1012 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1013 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1014 * config/i386/sse.md (mov<mode>_internal): Only check
1015 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1016 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1017 (<sse>_andnot<mode>3<mask_name>): Move check for
1018 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1019 (<code><mode>3<mask_name>): Ditto.
1020 (*andnot<mode>3): Ditto.
1021 (*andnottf3): Ditto.
1022 (*<code><mode>3): Ditto.
1023 (*<code>tf3): Ditto.
1024 (*andnot<VI:mode>3): Remove
1025 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1026 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1027 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1028 (sse4_1_blendv<ssemodesuffix>): Ditto.
1029 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1030 Explain that tune applies to 128bit instructions only.
1032 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1034 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1035 to definition of hsa_kernel_description. Parse assembly to find SGPR
1036 and VGPR count of kernel and store in hsa_kernel_description.
1038 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1040 PR rtl-optimization/91838
1041 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1042 to truncate if allowed or reject combination.
1044 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1046 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1047 (find_inv_vars_cb): Likewise.
1049 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1051 * calls.c (special_function_p): Split out the check for DECL_NAME
1052 being non-NULL and fndecl being extern at file scope into a
1053 new maybe_special_function_p and call it. Drop check for fndecl
1054 being non-NULL that was after a usage of DECL_NAME (fndecl).
1055 * tree.h (maybe_special_function_p): New inline function.
1057 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1059 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1060 (mask_gather_load<mode>): ... here, and zero-initialize the
1062 (maskload<mode>di): Zero-initialize the destination.
1065 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1068 * doc/analyzer.texi (Limitations): Note that constraints on
1069 floating-point values are currently ignored.
1071 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1074 * symtab.c (symtab_node::noninterposable_alias): If localalias
1075 already exists, but is not usable, append numbers after it until
1076 a unique name is found. Formatting fix.
1079 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1082 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1084 * config/gcn/gcn.c (print_operand): Handle LTGT.
1085 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1087 2020-01-30 Richard Biener <rguenther@suse.de>
1089 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1090 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1092 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1094 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1095 without a DECL in .data.rel.ro.local.
1097 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1100 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1104 * config/i386/sse.md
1105 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1106 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1107 any_extend code iterator instead of always zero_extend.
1108 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1109 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1110 Use any_extend code iterator instead of always zero_extend.
1111 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1112 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1113 Use any_extend code iterator instead of always zero_extend.
1114 (*sse2_pmovmskb_ext): New define_insn.
1115 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1118 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1119 (*popcountsi2_zext_falsedep): New define_insn.
1121 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1123 * config.in: Regenerated.
1124 * configure: Regenerated.
1126 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1129 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1130 LLVM's assembler changed the default in version 9.
1132 2020-01-24 Jeff Law <law@redhat.com>
1134 PR tree-optimization/89689
1135 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1137 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1141 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1143 PR rtl-optimization/87763
1144 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1145 simplification to handle subregs as well as bare regs.
1146 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1148 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1151 * ira.c (ira): Revert use of simplified LRA algorithm.
1153 2020-01-29 Martin Jambor <mjambor@suse.cz>
1155 PR tree-optimization/92706
1156 * tree-sra.c (struct access): Fields first_link, last_link,
1157 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1158 next_rhs_queued and grp_rhs_queued respectively, new fields
1159 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1160 (struct assign_link): Field next renamed to next_rhs, new field
1161 next_lhs. Updated comment.
1162 (work_queue_head): Renamed to rhs_work_queue_head.
1163 (lhs_work_queue_head): New variable.
1164 (add_link_to_lhs): New function.
1165 (relink_to_new_repr): Also relink LHS lists.
1166 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1167 (add_access_to_lhs_work_queue): New function.
1168 (pop_access_from_work_queue): Renamed to
1169 pop_access_from_rhs_work_queue.
1170 (pop_access_from_lhs_work_queue): New function.
1171 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1173 (child_would_conflict_in_lacc): Renamed to
1174 child_would_conflict_in_acc. Adjusted parameter names.
1175 (create_artificial_child_access): New parameter set_grp_read, use it.
1176 (subtree_mark_written_and_enqueue): Renamed to
1177 subtree_mark_written_and_rhs_enqueue.
1178 (propagate_subaccesses_across_link): Renamed to
1179 propagate_subaccesses_from_rhs.
1180 (propagate_subaccesses_from_lhs): New function.
1181 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1184 2020-01-29 Martin Jambor <mjambor@suse.cz>
1186 PR tree-optimization/92706
1187 * tree-sra.c (struct access): Adjust comment of
1188 grp_total_scalarization.
1189 (find_access_in_subtree): Look for single children spanning an entire
1191 (scalarizable_type_p): Allow register accesses, adjust callers.
1192 (completely_scalarize): Remove function.
1193 (scalarize_elem): Likewise.
1194 (create_total_scalarization_access): Likewise.
1195 (sort_and_splice_var_accesses): Do not track total scalarization
1197 (analyze_access_subtree): New parameter totally, adjust to new meaning
1198 of grp_total_scalarization.
1199 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1200 (can_totally_scalarize_forest_p): New function.
1201 (create_total_scalarization_access): Likewise.
1202 (create_total_access_and_reshape): Likewise.
1203 (total_should_skip_creating_access): Likewise.
1204 (totally_scalarize_subtree): Likewise.
1205 (analyze_all_variable_accesses): Perform total scalarization after
1206 subaccess propagation using the new functions above.
1207 (initialize_constant_pool_replacements): Output initializers by
1208 traversing the access tree.
1210 2020-01-29 Martin Jambor <mjambor@suse.cz>
1212 * tree-sra.c (verify_sra_access_forest): New function.
1213 (verify_all_sra_access_forests): Likewise.
1214 (create_artificial_child_access): Set parent.
1215 (analyze_all_variable_accesses): Call the verifier.
1217 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1219 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1220 if called on indirect edge.
1221 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1222 speculative call if needed.
1224 2020-01-29 Richard Biener <rguenther@suse.de>
1226 PR tree-optimization/93428
1227 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1228 permutation when the load node is created.
1229 (vect_analyze_slp_instance): Re-use it here.
1231 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1233 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1235 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1237 PR rtl-optimization/93272
1238 * ira-lives.c (process_out_of_region_eh_regs): New function.
1239 (process_bb_node_lives): Call it.
1241 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1243 * coverage.c (read_counts_file): Make error message lowercase.
1245 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1247 * profile-count.c (profile_quality_display_names): Fix ordering.
1249 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1252 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1253 hash only when edge is first within the sequence.
1254 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1255 (symbol_table::create_edge): Do not set target_prob.
1256 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1258 (cgraph_edge::make_speculative): Drop target_prob parameter.
1259 (cgraph_edge::speculative_call_info): Remove.
1260 (cgraph_edge::first_speculative_call_target): New member function.
1261 (update_call_stmt_hash_for_removing_direct_edge): New function.
1262 (cgraph_edge::resolve_speculation): Rewrite to new API.
1263 (cgraph_edge::speculative_call_for_target): New member function.
1264 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1265 multiple speculation targets.
1266 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1268 (verify_speculative_call): Verify that targets form an interval.
1269 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1270 (cgraph_edge::first_speculative_call_target): New member function.
1271 (cgraph_edge::next_speculative_call_target): New member function.
1272 (cgraph_edge::speculative_call_target_ref): New member function.
1273 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1274 (cgraph_edge): Remove target_prob.
1275 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1276 Fix handling of speculative calls.
1277 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1278 * ipa-fnsummary.c (analyze_function_body): Likewise.
1279 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1280 * ipa-profile.c (dump_histogram): Fix formating.
1281 (ipa_profile_generate_summary): Watch for overflows.
1282 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1283 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1284 (update_indirect_edges_after_inlining): Update to new API.
1285 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1287 * profile-count.h: (profile_probability::adjusted): New.
1288 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1289 updating of profile.
1290 * value-prof.c (gimple_ic_transform): Rename to ...
1291 (dump_ic_profile): ... this one; update dumping.
1292 (stream_in_histogram_value): Fix formating.
1293 (gimple_value_profile_transformations): Update.
1295 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1298 * config/i386/i386.md (*movoi_internal_avx): Remove
1299 TARGET_SSE_TYPELESS_STORES check.
1300 (*movti_internal): Prefer TARGET_AVX over
1301 TARGET_SSE_TYPELESS_STORES.
1302 (*movtf_internal): Likewise.
1303 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1304 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1305 from TARGET_SSE_TYPELESS_STORES.
1307 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1309 * diagnostic-core.h (warning_at): Rename overload to...
1310 (warning_meta): ...this.
1311 (emit_diagnostic_valist): Delete decl of overload taking
1312 diagnostic_metadata.
1313 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1314 (warning_at): Rename overload taking diagnostic_metadata to...
1315 (warning_meta): ...this.
1317 2020-01-28 Richard Biener <rguenther@suse.de>
1319 PR tree-optimization/93439
1320 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1321 * tree-cfg.c (move_sese_region_to_fn): ... here.
1322 (verify_types_in_gimple_reference): Verify used cliques are
1325 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1328 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1329 argument of a pointer to struct gcc_options and pass it to
1330 parse_mtune_ctrl_str.
1331 (ix86_function_specific_restore): Pass opts to
1332 set_ix86_tune_features.
1333 (ix86_option_override_internal): Likewise.
1334 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1335 gcc_options and use it for x_ix86_tune_ctrl_string.
1337 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1339 PR rtl-optimization/87763
1340 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1341 simplification to handle subregs as well as bare regs.
1342 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1344 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1346 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1347 for reduction chains that (now) include a call.
1349 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1351 PR tree-optimization/92822
1352 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1353 out the don't-care elements of a vector whose significant elements
1354 are duplicates, make the don't-care elements duplicates too.
1356 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1358 PR tree-optimization/93434
1359 * tree-predcom.c (split_data_refs_to_components): Record which
1360 components have had aliasing loads removed. Prevent store-store
1361 commoning for all such components.
1363 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1366 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1367 -1 or is_vshift is true, use new_vector with number of elts npatterns
1368 rather than new_unary_operation.
1370 PR tree-optimization/93454
1371 * gimple-fold.c (fold_array_ctor_reference): Perform
1372 elt_size.to_uhwi () just once, instead of calling it in every
1373 iteration. Punt if that value is above size of the temporary
1374 buffer. Decrease third native_encode_expr argument when
1375 bufoff + elt_sz is above size of buf.
1377 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1379 * config/mips/mips.c (mips_declare_object_name)
1380 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1382 2020-01-27 Martin Liska <mliska@suse.cz>
1384 PR gcov-profile/93403
1385 * tree-profile.c (gimple_init_gcov_profiler): Generate
1386 both __gcov_indirect_call_profiler_v4 and
1387 __gcov_indirect_call_profiler_v4_atomic.
1389 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1392 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1394 (@aarch64_split_simd_mov<mode>): Use it.
1395 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1396 Leave the vec_extract patterns to handle 2-element vectors.
1397 (aarch64_simd_mov_from_<mode>high): Likewise.
1398 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1399 (vec_extractv2dfv1df): Likewise.
1401 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1403 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1404 jump conditions for *compare_condjump<GPI:mode>.
1406 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1409 * digraph.cc (test_edge::test_edge): Specify template for base
1412 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1414 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1416 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1418 * config/arc/arc-protos.h (gen_mlo): Remove.
1419 (gen_mhi): Likewise.
1420 * config/arc/arc.c (AUX_MULHI): Define.
1421 (arc_must_save_reister): Special handling for r58/59.
1422 (arc_compute_frame_size): Consider mlo/mhi registers.
1423 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1425 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1426 mlo/mhi name selection.
1427 (arc_restore_callee_saves): Don't early restore blink when ISR.
1428 (arc_expand_prologue): Add mlo/mhi saving.
1429 (arc_expand_epilogue): Add mlo/mhi restoring.
1432 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1433 numbering when MUL64 option is used.
1434 (DWARF2_FRAME_REG_OUT): Define.
1435 * config/arc/arc.md (arc600_stall): New pattern.
1436 (VUNSPEC_ARC_ARC600_STALL): Define.
1437 (mulsi64): Use correct mlo/mhi registers.
1438 (mulsi_600): Clean it up.
1439 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1441 (mhi_operand): Likewise.
1443 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1444 Petro Karashchenko <petro.karashchenko@ring.com>
1446 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1447 attributes if needed.
1448 (prepare_move_operands): Generate special unspec instruction for
1450 (arc_isuncached_mem_p): Propagate uncached attribute to each
1452 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1453 (VUNSPEC_ARC_STDI): Likewise.
1454 (ALLI): New mode iterator.
1455 (mALLI): New mode attribute.
1456 (lddi): New instruction pattern.
1458 (stdidi_split): Split instruction for architectures which are not
1459 supporting ll64 option.
1460 (lddidi_split): Likewise.
1462 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1464 PR rtl-optimization/92989
1465 * lra-lives.c (process_bb_lives): Update the live-in set before
1466 processing additional clobbers.
1468 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1470 PR rtl-optimization/93170
1471 * cselib.c (cselib_invalidate_regno_val): New function, split out
1473 (cselib_invalidate_regno): ...here.
1474 (cselib_invalidated_by_call_p): New function.
1475 (cselib_process_insn): Iterate over all the hard-register entries in
1476 REG_VALUES and invalidate any that cross call-clobbered registers.
1478 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1480 * dojump.c (split_comparison): Use HONOR_NANS rather than
1481 HONOR_SNANS when splitting LTGT.
1483 2020-01-27 Martin Liska <mliska@suse.cz>
1486 * opts.c (print_filtered_help): Exclude language-specific
1487 options from --help=common unless enabled in all FEs.
1489 2020-01-27 Martin Liska <mliska@suse.cz>
1491 * opts.c (print_help): Exclude params from
1492 all except --help=param.
1494 2020-01-27 Martin Liska <mliska@suse.cz>
1497 * config/i386/i386-features.c (make_resolver_func):
1498 Align the code with ppc64 target implementation.
1499 Do not generate a unique name for resolver function.
1501 2020-01-27 Richard Biener <rguenther@suse.de>
1503 PR tree-optimization/93397
1504 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1505 converted reduction chain SLP graph adjustment.
1507 2020-01-26 Marek Polacek <polacek@redhat.com>
1510 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1513 2020-01-26 Jason Merrill <jason@redhat.com>
1516 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1519 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1521 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1522 (rx_setmem): Likewise.
1524 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1527 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1528 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1529 drop <di> from constraint of last operand.
1532 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1533 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1534 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1536 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1539 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1542 2020-01-24 Jeff Law <law@redhat.com>
1544 PR tree-optimization/92788
1545 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1548 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1551 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1552 *avx_vperm_broadcast_<mode>,
1553 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1554 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1555 Move before avx2_perm<mode>/avx512f_perm<mode>.
1558 * simplify-rtx.c (simplify_const_unary_operation,
1559 simplify_const_binary_operation): Punt for mode precision above
1560 MAX_BITSIZE_MODE_ANY_INT.
1562 2020-01-24 Andrew Pinski <apinski@marvell.com>
1564 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1567 2020-01-24 Jeff Law <law@redhat.com>
1570 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1571 for REGs. Call output_operand_lossage to get more reasonable
1574 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1576 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1577 gcn_fp_compare_operator.
1578 (vec_cmpu<mode>di): Use gcn_compare_operator.
1579 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1580 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1581 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1582 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1583 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1584 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1585 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1586 gcn_fp_compare_operator.
1587 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1588 gcn_fp_compare_operator.
1589 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1590 gcn_fp_compare_operator.
1591 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1592 gcn_fp_compare_operator.
1594 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1596 * doc/install.texi (Cross-Compiler-Specific Options): Document
1597 `--with-toolexeclibdir' option.
1599 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1601 * target.def (flags_regnum): Also mention effect on delay slot filling.
1602 * doc/tm.texi: Regenerate.
1604 2020-01-23 Jeff Law <law@redhat.com>
1606 PR translation/90162
1607 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1609 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1612 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1615 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1617 PR rtl-optimization/93402
1618 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1621 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1623 * config.in: Regenerated.
1624 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1625 for TARGET_LIBC_GNUSTACK.
1626 * configure: Regenerated.
1627 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1628 found to be 2.31 or greater.
1630 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1632 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1634 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1635 (mips_asm_file_end): New function. Delegate to
1636 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1637 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1639 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1642 * config/i386/i386-modes.def (POImode): New mode.
1643 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1644 * config/i386/i386.md (DPWI): New mode attribute.
1645 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1647 (QPWI): ... this. Use POI instead of OI for TImode.
1648 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1649 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1652 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1655 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1657 (speculation_tracker_rev): New pattern.
1658 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1659 Use speculation_tracker_rev to track the inverse condition.
1661 2020-01-23 Richard Biener <rguenther@suse.de>
1663 PR tree-optimization/93381
1664 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1665 alias-set of the def as argument and record the first one.
1666 (vn_walk_cb_data::first_set): New member.
1667 (vn_reference_lookup_3): Pass the alias-set of the current def
1668 to push_partial_def. Fix alias-set used in the aggregate copy
1670 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1671 * real.c (clear_significand_below): Fix out-of-bound access.
1673 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1676 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1677 New define_insn patterns.
1679 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1681 * doc/sourcebuild.texi (check-function-bodies): Add an
1682 optional target/xfail selector.
1684 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1686 PR rtl-optimization/93124
1687 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1688 bare USE and CLOBBER insns.
1690 2020-01-22 Andrew Pinski <apinski@marvell.com>
1692 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1694 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1697 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1698 diagnostic_manager into "ana" namespace.
1699 * selftest-run-tests.c (selftest::run_tests): Update for move of
1700 selftest::run_analyzer_selftests to
1701 ana::selftest::run_analyzer_selftests.
1703 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1705 * cfgexpand.c (union_stack_vars): Update the size.
1707 2020-01-22 Richard Biener <rguenther@suse.de>
1709 PR tree-optimization/93381
1710 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1711 throughout, handle all conversions the same.
1713 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1716 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1717 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1718 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1719 Call force_reg on high_in2 unconditionally.
1721 2020-01-22 Martin Liska <mliska@suse.cz>
1723 PR tree-optimization/92924
1724 * profile.c (compute_value_histograms): Divide
1727 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1730 * output.h (assemble_name_resolve): Declare.
1731 * varasm.c (assemble_name_resolve): New function.
1732 (assemble_name): Use it.
1733 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1735 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1737 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1738 update_web_docs_git instead of update_web_docs_svn.
1740 2020-01-21 Andrew Pinski <apinski@marvell.com>
1743 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1744 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1745 (*tlsgd_small_<mode>): Likewise.
1746 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1747 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1748 register. Convert that register back to dest using convert_mode.
1750 2020-01-21 Jim Wilson <jimw@sifive.com>
1752 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1755 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1756 Uros Bizjak <ubizjak@gmail.com>
1759 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1761 (legitimize_tls_address): Do GNU2 TLS address computation in
1762 ptr_mode and zero-extend result to Pmode.
1763 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1764 :P with :PTR and Pmode with ptr_mode.
1765 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1766 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1767 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1769 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1772 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1773 the last two operands are CONST_INT_P before using them as such.
1775 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1777 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1778 to get the integer element types.
1780 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1782 * config/aarch64/aarch64-sve-builtins.h
1783 (function_expander::convert_to_pmode): Declare.
1784 * config/aarch64/aarch64-sve-builtins.cc
1785 (function_expander::convert_to_pmode): New function.
1786 (function_expander::get_contiguous_base): Use it.
1787 (function_expander::prepare_gather_address_operands): Likewise.
1788 * config/aarch64/aarch64-sve-builtins-sve2.cc
1789 (svwhilerw_svwhilewr_impl::expand): Likewise.
1791 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1794 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1795 cfun->machine->label_is_assembled.
1796 (aarch64_print_patchable_function_entry): New.
1797 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1798 * config/aarch64/aarch64.h (struct machine_function): New field,
1801 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1804 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1807 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1810 * cgraph.c (cgraph_edge::resolve_speculation,
1811 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1812 call_stmt_site_hash.
1814 2020-01-21 Martin Liska <mliska@suse.cz>
1816 * config/rs6000/rs6000.c (common_mode_defined): Remove
1819 2020-01-21 Richard Biener <rguenther@suse.de>
1821 PR tree-optimization/92328
1822 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1823 type when value-numbering same-sized store by inserting a
1825 (eliminate_dom_walker::eliminate_stmt): When eliminating
1826 a redundant store handle bit-reinterpretation of the same value.
1828 2020-01-21 Andrew Pinski <apinski@marvel.com>
1831 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1833 (prepare_block_for_update): This. Use a worklist instead of
1836 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1838 * gcc/config/arm/arm.c (clear_operation_p):
1839 Initialise last_regno, skip first iteration
1840 based on the first_set value and use ints instead
1841 of the unnecessary HOST_WIDE_INTs.
1843 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1846 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1847 compare_mode other than SFmode or DFmode.
1849 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1852 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1853 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1854 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1856 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1858 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1860 2020-01-20 Andrew Pinski <apinski@marvell.com>
1863 * targhooks.c (default_print_patchable_function_entry): Use
1864 output_asm_insn to emit the nop instruction.
1866 2020-01-20 Fangrui Song <maskray@google.com>
1869 * targhooks.c (default_print_patchable_function_entry): Align to
1872 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
1875 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1876 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
1877 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1878 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
1879 (*tls_dynamic_gnu2_lea_64): Renamed to ...
1880 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
1881 Remove the {q} suffix from lea.
1882 (*tls_dynamic_gnu2_call_64): Renamed to ...
1883 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
1884 (*tls_dynamic_gnu2_combine_64): Renamed to ...
1885 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
1886 Pass Pmode to gen_tls_dynamic_gnu2_64.
1888 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1890 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1892 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
1894 * config/aarch64/aarch64-sve-builtins-base.cc
1895 (svld1ro_impl::memory_vector_mode): Remove parameter name.
1897 2020-01-20 Richard Biener <rguenther@suse.de>
1900 * dwarf2out.c (prune_unused_types): Unconditionally mark
1901 called function DIEs.
1903 2020-01-20 Martin Liska <mliska@suse.cz>
1905 PR tree-optimization/93199
1906 * tree-eh.c (struct leh_state): Add
1907 new field outer_non_cleanup.
1908 (cleanup_is_dead_in): Pass leh_state instead
1909 of eh_region. Add a checking that state->outer_non_cleanup
1910 points to outer non-clean up region.
1911 (lower_try_finally): Record outer_non_cleanup
1913 (lower_catch): Likewise.
1914 (lower_eh_filter): Likewise.
1915 (lower_eh_must_not_throw): Likewise.
1916 (lower_cleanup): Likewise.
1918 2020-01-20 Richard Biener <rguenther@suse.de>
1920 PR tree-optimization/93094
1921 * tree-vectorizer.h (vect_loop_versioning): Adjust.
1922 (vect_transform_loop): Likewise.
1923 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1924 loop_vectorized_call to vect_transform_loop.
1925 * tree-vect-loop.c (vect_transform_loop): Pass down
1926 loop_vectorized_call to vect_loop_versioning.
1927 * tree-vect-loop-manip.c (vect_loop_versioning): Use
1928 the earlier discovered loop_vectorized_call.
1930 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
1932 * doc/contribute.texi: Update for SVN -> Git transition.
1933 * doc/install.texi: Likewise.
1935 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1938 * cgraph.c (cgraph_edge::make_speculative): Increase number of
1939 speculative targets.
1940 (verify_speculative_call): New function
1941 (cgraph_node::verify_node): Use it.
1942 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
1945 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1948 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
1949 (cgraph_edge::make_direct): Remove all indirect targets.
1950 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
1951 (cgraph_node::verify_node): Verify that only one call_stmt or
1952 lto_stmt_uid is set.
1953 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
1955 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
1956 (lto_output_ref): Simplify streaming of stmt.
1957 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
1959 2020-01-18 Tamar Christina <tamar.christina@arm.com>
1961 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
1962 Mark parameter unused.
1964 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
1966 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
1968 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
1970 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
1972 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
1974 * Makefile.in: Add coroutine-passes.o.
1975 * builtin-types.def (BT_CONST_SIZE): New.
1976 (BT_FN_BOOL_PTR): New.
1977 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
1978 * builtins.def (DEF_COROUTINE_BUILTIN): New.
1979 * coroutine-builtins.def: New file.
1980 * coroutine-passes.cc: New file.
1981 * function.h (struct GTY function): Add a bit to indicate that the
1982 function is a coroutine component.
1983 * internal-fn.c (expand_CO_FRAME): New.
1984 (expand_CO_YIELD): New.
1985 (expand_CO_SUSPN): New.
1986 (expand_CO_ACTOR): New.
1987 * internal-fn.def (CO_ACTOR): New.
1991 * passes.def: Add pass_coroutine_lower_builtins,
1992 pass_coroutine_early_expand_ifns.
1993 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
1994 (make_pass_coroutine_early_expand_ifns): New.
1995 * doc/invoke.texi: Document the fcoroutines command line
1998 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2000 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2003 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2004 after checking the argument is a REG. Don't use REGNO (reg)
2005 again to set last_regno, reuse regno variable instead.
2007 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2009 * doc/analyzer.texi (Limitations): Add note about NaN.
2011 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2012 Sudakshina Das <sudi.das@arm.com>
2014 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2015 and valid immediate.
2016 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2017 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2018 * config/arm/constraints.md (Pg): New.
2019 * config/arm/predicates.md (long_shift_imm): New.
2020 (arm_reg_or_long_shift_imm): Likewise.
2021 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2022 (thumb2_lsll): Likewise.
2025 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2026 Sudakshina Das <sudi.das@arm.com>
2028 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2029 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2030 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2031 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2032 * config/arm/thumb2.md (thumb2_asrl): New.
2033 (thumb2_lsll): Likewise.
2035 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2037 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2040 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2042 * gdbinit.in (help-gcc-hooks): New command.
2043 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2044 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2047 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2049 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2050 correct target macro.
2052 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2054 * config/aarch64/aarch64-protos.h
2055 (aarch64_sve_ld1ro_operand_p): New.
2056 * config/aarch64/aarch64-sve-builtins-base.cc
2057 (class load_replicate): New.
2058 (class svld1ro_impl): New.
2059 (class svld1rq_impl): Change to inherit from load_replicate.
2060 (svld1ro): New sve intrinsic function base.
2061 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2062 New DEF_SVE_FUNCTION.
2063 * config/aarch64/aarch64-sve-builtins-base.h
2064 (svld1ro): New decl.
2065 * config/aarch64/aarch64-sve-builtins.cc
2066 (function_expander::add_mem_operand): Modify assert to allow
2068 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2070 * config/aarch64/aarch64.c
2071 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2072 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2073 (aarch64_sve_ld1ro_operand_p): New.
2074 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2075 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2076 * config/aarch64/predicates.md
2077 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2079 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2081 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2082 Introduce this ACLE specified predefined macro.
2083 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2084 (fp): Disabling this disables f64mm.
2085 (simd): Disabling this disables f64mm.
2086 (fp16): Disabling this disables f64mm.
2087 (sve): Disabling this disables f64mm.
2088 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2089 (AARCH64_ISA_F64MM): New.
2090 (TARGET_F64MM): New.
2091 * doc/invoke.texi (f64mm): Document new option.
2093 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2095 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2096 (neoversen1_tunings): Likewise.
2098 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2101 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2102 Add assert to ensure prolog has been emitted.
2103 (aarch64_split_atomic_op): Likewise.
2104 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2105 Use epilogue_completed rather than reload_completed.
2106 (aarch64_atomic_exchange<mode>): Likewise.
2107 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2108 (atomic_nand<mode>): Likewise.
2109 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2110 (atomic_fetch_nand<mode>): Likewise.
2111 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2112 (atomic_nand_fetch<mode>): Likewise.
2114 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2117 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2119 (REVERSE_CONDITION): Delete.
2120 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2121 (CCFP_CCFPE): Likewise.
2122 (e): New mode attribute.
2123 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2124 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2125 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2126 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2127 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2128 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2129 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2130 name of generator from gen_ccmpdi to gen_ccmpccdi.
2131 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2132 the previous comparison but aren't able to, use the new ccmp_rev
2135 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2137 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2138 than testing directly for INTEGER_CST.
2139 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2141 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2143 PR tree-optimization/93292
2144 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2145 get_vectype_for_scalar_type returns NULL.
2147 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2149 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2150 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2152 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2154 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2156 * params.opt: (max-predicted-iterations): Set bounds.
2157 * predict.c (real_almost_one, real_br_prob_base,
2158 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2159 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2160 probabilities; do not truncate to reg_br_prob_bases.
2161 (estimate_loops_at_level): Pass max_cyclic_prob.
2162 (estimate_loops): Compute max_cyclic_prob.
2163 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2165 * profile-count.c (profile_probability::to_sreal): New.
2166 * profile-count.h (class sreal): Move up in file.
2167 (profile_probability::to_sreal): Declare.
2169 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2172 (arm_invalid_conversion): New function for target hook.
2173 (arm_invalid_unary_op): New function for target hook.
2174 (arm_invalid_binary_op): New function for target hook.
2176 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2178 * config.gcc: Add arm_bf16.h.
2179 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2180 (arm_simd_builtin_std_type): Add BFmode.
2181 (arm_init_simd_builtin_types): Define element types for vector types.
2182 (arm_init_bf16_types): New function.
2183 (arm_init_builtins): Add arm_init_bf16_types function call.
2184 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2185 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2186 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2187 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2188 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2189 (arm_mangle_type): Add __bf16.
2190 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2191 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2192 arm_bf16_ptr_type_node.
2193 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2194 define_split between ARM registers.
2195 * config/arm/arm_bf16.h: New file.
2196 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2197 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2199 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2200 * config/arm/vfp.md: Add BFmode to movhf patterns.
2202 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2203 Andre Vieira <andre.simoesdiasvieira@arm.com>
2205 * config/arm/arm-cpus.in (mve, mve_float): New features.
2206 (dsp, mve, mve.fp): New options.
2207 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2208 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2209 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2211 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2212 Thomas Preud'homme <thomas.preudhomme@arm.com>
2214 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2216 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2217 error for using -mcmse when targeting Armv8.1-M Mainline.
2219 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2220 Thomas Preud'homme <thomas.preudhomme@arm.com>
2222 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2223 address in r4 when targeting Armv8.1-M Mainline.
2224 (nonsecure_call_value_internal): Likewise.
2225 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2226 a register match_operand again. Emit BLXNS when targeting
2228 (nonsecure_call_value_reg_thumb2): Likewise.
2230 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2231 Thomas Preud'homme <thomas.preudhomme@arm.com>
2233 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2234 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2235 variable as true when floating-point ABI is not hard. Replace
2236 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2237 Generate VLSTM and VLLDM instruction respectively before and
2238 after a function call to cmse_nonsecure_call function.
2239 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2240 (VUNSPEC_VLLDM): Likewise.
2241 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2242 (lazy_load_multiple_insn): Likewise.
2244 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2245 Thomas Preud'homme <thomas.preudhomme@arm.com>
2247 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2248 (arm_emit_vfp_multi_reg_pop): Likewise.
2249 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2250 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2251 restore callee-saved VFP registers.
2253 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2254 Thomas Preud'homme <thomas.preudhomme@arm.com>
2256 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2257 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2258 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2259 callee-saved GPRs as well as clear ip register before doing a nonsecure
2260 call then restore callee-saved GPRs after it when targeting
2262 (arm_reorg): Adapt to function rename.
2264 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2265 Thomas Preud'homme <thomas.preudhomme@arm.com>
2267 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2268 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2269 clear_vfp_multiple pattern based on a new vfp parameter.
2270 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2271 targeting Armv8.1-M Mainline.
2272 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2273 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2274 whether VFP registers are available before looking call_used_regs for a
2276 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2277 of prototype of clear_operation_p.
2278 (clear_vfp_multiple_operation): New predicate.
2279 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2280 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2282 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2283 Thomas Preud'homme <thomas.preudhomme@arm.com>
2285 * config/arm/arm-protos.h (clear_operation_p): Declare.
2286 * config/arm/arm.c (clear_operation_p): New function.
2287 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2288 targeting Armv8.1-M Mainline or successor.
2289 (output_return_instruction): Only output APSR register clearing if
2290 Armv8.1-M Mainline instructions not available.
2291 (thumb_exit): Likewise.
2292 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2293 * config/arm/thumb2.md (clear_apsr): New define_insn.
2294 (clear_multiple): Likewise.
2295 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2297 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2298 Thomas Preud'homme <thomas.preudhomme@arm.com>
2300 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2301 (use_return_insn): Also return false for Armv8.1-M Mainline.
2302 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2303 Mainline instructions are available.
2304 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2305 when targeting Armv8.1-M Mainline Security Extensions.
2306 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2307 Mainline entry function.
2308 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2309 targeting Armv8.1-M Mainline or successor.
2310 (arm_expand_epilogue): Fix indentation of caller-saved register
2311 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2313 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2314 (FP_SYSREGS): Likewise.
2315 (enum vfp_sysregs_encoding): Define enum.
2316 (fp_sysreg_names): Declare.
2317 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2318 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2319 (pop_fpsysreg_insn): Likewise.
2321 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2322 Thomas Preud'homme <thomas.preudhomme@arm.com>
2324 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2325 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2326 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2327 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2328 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2329 (ARMv8_1m_main): New feature group.
2330 (armv8.1-m.main): New architecture.
2331 * config/arm/arm-tables.opt: Regenerate.
2332 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2333 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2334 (arm_options_perform_arch_sanity_checks): Error out when targeting
2335 Armv8.1-M Mainline Security Extensions.
2336 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2338 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2340 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2341 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2342 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2343 aarch64_bfdot_laneq): New.
2344 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2345 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2346 vbfdotq_laneq_f32): New.
2347 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2348 VBFMLA_W, VBF): New.
2349 (isquadop): Add V4BF, V8BF.
2351 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2353 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2354 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2355 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2356 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2357 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2358 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2359 usdot_laneq, sudot_lane,sudot_laneq): New.
2360 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2361 (aarch64_<sur>dot_lane): New.
2362 * config/aarch64/arm_neon.h (vusdot_s32): New.
2364 (vusdot_lane_s32): New.
2365 (vsudot_lane_s32): New.
2366 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2367 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2369 2020-01-16 Martin Liska <mliska@suse.cz>
2371 * value-prof.c (dump_histogram_value): Fix
2372 obvious spacing issue.
2374 2020-01-16 Andrew Pinski <apinski@marvell.com>
2376 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2377 !storage_order_barrier_p.
2379 2020-01-16 Andrew Pinski <apinski@marvell.com>
2381 * sched-int.h (_dep): Add unused bit-field field for the padding.
2382 * sched-deps.c (init_dep_1): Init unused field.
2384 2020-01-16 Andrew Pinski <apinski@marvell.com>
2386 * optabs.h (create_expand_operand): Initialize target field also.
2388 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2390 PR tree-optimization/92429
2391 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2392 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2394 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2397 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2399 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2400 aarch64_sve_int_mode to each mode.
2402 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2404 * doc/analyzer.texi (Overview): Add note about
2405 -fdump-ipa-analyzer.
2407 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2409 PR tree-optimization/93231
2410 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2411 input_type is unsigned. Use tree_to_shwi for shift constant.
2412 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2413 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2414 inputs more efficiently.
2416 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2418 * config/i386/i386.md (*movsf_internal): Do not require
2419 SSE2 ISA for alternatives 14 and 15.
2421 2020-01-15 Richard Biener <rguenther@suse.de>
2424 * tree-eh.c (sink_clobbers): If we already visited the destination
2425 block do not defer insertion.
2426 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2427 the purpose of defered insertion.
2429 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2431 * BASE-VER: Bump to 10.0.1.
2433 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2435 PR tree-optimization/93247
2436 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2437 type of the stmt that we're going to vectorize.
2439 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2441 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2442 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2445 2020-01-15 Martin Liska <mliska@suse.cz>
2447 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2448 2 calls of streamer_read_hwi in a function call.
2450 2020-01-15 Richard Biener <rguenther@suse.de>
2452 * alias.c (record_alias_subset): Avoid redundant work when
2453 subset is already recorded.
2455 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2457 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2458 the analyzer options provide CWE identifiers.
2460 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2462 * tree-diagnostic-path.cc (path_summary::event_range::print):
2463 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2464 using get_pure_location.
2466 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2468 PR tree-optimization/93262
2469 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2470 perform head trimming only if the last argument is constant,
2471 either all ones, or larger or equal to head trim, in the latter
2472 case decrease the last argument by head_trim.
2474 PR tree-optimization/93249
2475 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2476 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2477 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2478 perform head trim unless we can prove there are no '\0' chars
2479 from the source among the first head_trim chars.
2481 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2483 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2485 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2488 * config/i386/sse.md
2489 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2490 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2491 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2492 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2493 just a single alternative instead of two, make operands 1 and 2
2496 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2499 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2502 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2504 * Makefile.in (lang_opt_files): Add analyzer.opt.
2505 (ANALYZER_OBJS): New.
2506 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2507 tristate.o and ANALYZER_OBJS.
2508 (TEXI_GCCINT_FILES): Add analyzer.texi.
2509 * common.opt (-fanalyzer): New driver option.
2510 * config.in: Regenerate.
2511 * configure: Regenerate.
2512 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2513 (gccdepdir): Also create depdir for "analyzer" subdir.
2514 * digraph.cc: New file.
2515 * digraph.h: New file.
2516 * doc/analyzer.texi: New file.
2517 * doc/gccint.texi ("Static Analyzer") New menu item.
2518 (analyzer.texi): Include it.
2519 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2520 ("Warning Options"): Add static analysis warnings to the list.
2521 (-Wno-analyzer-double-fclose): New option.
2522 (-Wno-analyzer-double-free): New option.
2523 (-Wno-analyzer-exposure-through-output-file): New option.
2524 (-Wno-analyzer-file-leak): New option.
2525 (-Wno-analyzer-free-of-non-heap): New option.
2526 (-Wno-analyzer-malloc-leak): New option.
2527 (-Wno-analyzer-possible-null-argument): New option.
2528 (-Wno-analyzer-possible-null-dereference): New option.
2529 (-Wno-analyzer-null-argument): New option.
2530 (-Wno-analyzer-null-dereference): New option.
2531 (-Wno-analyzer-stale-setjmp-buffer): New option.
2532 (-Wno-analyzer-tainted-array-index): New option.
2533 (-Wno-analyzer-use-after-free): New option.
2534 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2535 (-Wno-analyzer-use-of-uninitialized-value): New option.
2536 (-Wanalyzer-too-complex): New option.
2537 (-fanalyzer-call-summaries): New warning.
2538 (-fanalyzer-checker=): New warning.
2539 (-fanalyzer-fine-grained): New warning.
2540 (-fno-analyzer-state-merge): New warning.
2541 (-fno-analyzer-state-purge): New warning.
2542 (-fanalyzer-transitivity): New warning.
2543 (-fanalyzer-verbose-edges): New warning.
2544 (-fanalyzer-verbose-state-changes): New warning.
2545 (-fanalyzer-verbosity=): New warning.
2546 (-fdump-analyzer): New warning.
2547 (-fdump-analyzer-callgraph): New warning.
2548 (-fdump-analyzer-exploded-graph): New warning.
2549 (-fdump-analyzer-exploded-nodes): New warning.
2550 (-fdump-analyzer-exploded-nodes-2): New warning.
2551 (-fdump-analyzer-exploded-nodes-3): New warning.
2552 (-fdump-analyzer-supergraph): New warning.
2553 * doc/sourcebuild.texi (dg-require-dot): New.
2554 (dg-check-dot): New.
2555 * gdbinit.in (break-on-saved-diagnostic): New command.
2556 * graphviz.cc: New file.
2557 * graphviz.h: New file.
2558 * ordered-hash-map-tests.cc: New file.
2559 * ordered-hash-map.h: New file.
2560 * passes.def (pass_analyzer): Add before
2561 pass_ipa_whole_program_visibility.
2562 * selftest-run-tests.c (selftest::run_tests): Call
2563 selftest::ordered_hash_map_tests_cc_tests.
2564 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2566 * shortest-paths.h: New file.
2567 * timevar.def (TV_ANALYZER): New timevar.
2568 (TV_ANALYZER_SUPERGRAPH): Likewise.
2569 (TV_ANALYZER_STATE_PURGE): Likewise.
2570 (TV_ANALYZER_PLAN): Likewise.
2571 (TV_ANALYZER_SCC): Likewise.
2572 (TV_ANALYZER_WORKLIST): Likewise.
2573 (TV_ANALYZER_DUMP): Likewise.
2574 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2575 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2576 * tree-pass.h (make_pass_analyzer): New decl.
2577 * tristate.cc: New file.
2578 * tristate.h: New file.
2580 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2583 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2584 alternatives 9 and 10.
2586 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2588 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2589 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2590 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2591 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2592 (selftest::hash_map_tests_c_tests): Call it.
2593 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2594 New static constant, using the value of = H::empty_zero_p.
2595 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2596 from default_hash_traits <Value>.
2597 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2599 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2600 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2601 calls to mark_empty with !Descriptor::empty_zero_p.
2602 (hash_table::empty_slow): Conditionalize the memset call with a
2603 check that Descriptor::empty_zero_p; otherwise, loop through the
2604 entries calling mark_empty on them.
2605 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2606 (pointer_hash::empty_zero_p): Likewise.
2607 (pair_hash::empty_zero_p): Likewise.
2608 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2610 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2611 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2612 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2613 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2614 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2615 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2616 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2617 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2619 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2622 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2624 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2625 fix typo on return value.
2627 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2630 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2632 (cgraph_edge::make_speculative): Add param for setting speculative_id
2634 (cgraph_edge::speculative_call_info): Update comments and find reference
2635 by speculative_id for multiple indirect targets.
2636 (cgraph_edge::resolve_speculation): Decrease the speculations
2637 for indirect edge, drop it's speculative if not direct target
2638 left. Update comments.
2639 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2640 (cgraph_node::dump): Print num_speculative_call_targets.
2641 (cgraph_node::verify_node): Don't report error if speculative
2642 edge not include statement.
2643 (cgraph_edge::num_speculative_call_targets_p): New function.
2644 * cgraph.h (int common_target_id): Remove.
2645 (int common_target_probability): Remove.
2646 (num_speculative_call_targets): New variable.
2647 (make_speculative): Add param for setting speculative_id.
2648 (cgraph_edge::num_speculative_call_targets_p): New declare.
2649 (target_prob): New variable.
2650 (speculative_id): New variable.
2651 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2652 call summaries for multiple speculative call targets.
2653 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2654 * ipa-profile.c (struct speculative_call_target): New struct.
2655 (class speculative_call_summary): New class.
2656 (class speculative_call_summaries): New class.
2657 (call_sums): New variable.
2658 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2659 (ipa_profile_write_edge_summary): New function.
2660 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2661 (ipa_profile_dump_all_summaries): New function.
2662 (ipa_profile_read_edge_summary): New function.
2663 (ipa_profile_read_summary_section): New function.
2664 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2665 (ipa_profile): Generate num_speculative_call_targets from
2667 * ipa-ref.h (speculative_id): New variable.
2668 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2669 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2670 common_target_probability. Stream out speculative_id and
2671 num_speculative_call_targets.
2672 (input_edge): Likewise.
2673 * predict.c (dump_prediction): Remove edges count assert to be
2675 * symtab.c (symtab_node::create_reference): Init speculative_id.
2676 (symtab_node::clone_references): Clone speculative_id.
2677 (symtab_node::clone_referring): Clone speculative_id.
2678 (symtab_node::clone_reference): Clone speculative_id.
2679 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2680 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2681 if indirect call contains multiple speculative targets.
2682 * value-prof.h (check_ic_target): Remove.
2683 * value-prof.c (gimple_value_profile_transformations):
2684 Use void function gimple_ic_transform.
2685 * value-prof.c (gimple_ic_transform): Handle topn case.
2686 Fix comment typos. Change it to a void function.
2688 2020-01-13 Andrew Pinski <apinski@marvell.com>
2690 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2691 (octeontx2t98): New define.
2692 (octeontx2t96): New define.
2693 (octeontx2t93): New define.
2694 (octeontx2f95): New define.
2695 (octeontx2f95n): New define.
2696 (octeontx2f95mm): New define.
2697 * config/aarch64/aarch64-tune.md: Regenerate.
2698 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2700 2020-01-13 Jason Merrill <jason@redhat.com>
2702 PR c++/33799 - destroy return value if local cleanup throws.
2703 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2705 2020-01-13 Martin Liska <mliska@suse.cz>
2707 * ipa-cp.c (get_max_overall_size): Use newly
2708 renamed param param_ipa_cp_unit_growth.
2709 * params.opt: Remove legacy param name.
2711 2020-01-13 Martin Sebor <msebor@redhat.com>
2713 PR tree-optimization/93213
2714 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2715 stores to be eliminated.
2717 2020-01-13 Martin Liska <mliska@suse.cz>
2719 * opts.c (print_help): Do not print CL_PARAM
2720 and CL_WARNING for CL_OPTIMIZATION.
2722 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2725 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2726 depending on optimization settings.
2728 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2730 PR tree-optimization/90838
2731 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2732 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2733 argument rather than to initialize temporary for targets that
2734 don't use the mode argument at all. Initialize ctzval to avoid
2737 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2739 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2740 * tree-core.h: Document it.
2741 * gimplify.c (gimplify_omp_workshare): Set it.
2742 * omp-low.c (lower_omp_target): Use it.
2743 * tree-pretty-print.c (dump_omp_clause): Print it.
2745 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2746 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2748 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2750 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2751 * common.opt (fdiagnostics-path-format=): New option.
2752 (diagnostic_path_format): New enum.
2753 (fdiagnostics-show-path-depths): New option.
2754 * coretypes.h (diagnostic_event_id_t): New forward decl.
2755 * diagnostic-color.c (color_dict): Add "path".
2756 * diagnostic-event-id.h: New file.
2757 * diagnostic-format-json.cc (json_from_expanded_location): Make
2759 (json_end_diagnostic): Call context->make_json_for_path if it
2760 exists and the diagnostic has a path.
2761 (diagnostic_output_format_init): Clear context->print_path.
2762 * diagnostic-path.h: New file.
2763 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2764 when printing a run of events in a diagnostic_path so that they
2765 all get the same color.
2766 (layout::m_diagnostic_path_p): New field.
2767 (layout::layout): Initialize it.
2768 (layout::print_any_labels): Don't colorize the label text for an
2769 event in a diagnostic_path.
2770 (gcc_rich_location::add_location_if_nearby): Add
2771 "restrict_to_current_line_spans" and "label" params. Pass the
2772 former to layout.maybe_add_location_range; pass the latter
2773 when calling add_range.
2774 * diagnostic.c: Include "diagnostic-path.h".
2775 (diagnostic_initialize): Initialize context->path_format and
2776 context->show_path_depths.
2777 (diagnostic_show_any_path): New function.
2778 (diagnostic_path::interprocedural_p): New function.
2779 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2780 (simple_diagnostic_path::num_events): New function.
2781 (simple_diagnostic_path::get_event): New function.
2782 (simple_diagnostic_path::add_event): New function.
2783 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2784 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2785 (debug): New overload taking a diagnostic_path *.
2786 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2787 * diagnostic.h (enum diagnostic_path_format): New enum.
2788 (json::value): New forward decl.
2789 (diagnostic_context::path_format): New field.
2790 (diagnostic_context::show_path_depths): New field.
2791 (diagnostic_context::print_path): New callback field.
2792 (diagnostic_context::make_json_for_path): New callback field.
2793 (diagnostic_show_any_path): New decl.
2794 (json_from_expanded_location): New decl.
2795 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2796 (-fdiagnostics-show-path-depths): New option.
2797 (-fdiagnostics-color): Add "path" to description of default
2798 GCC_COLORS; describe it.
2799 (-fdiagnostics-format=json): Document how diagnostic paths are
2800 represented in the JSON output format.
2801 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2802 Add optional params "restrict_to_current_line_spans" and "label".
2803 * opts.c (common_handle_option): Handle
2804 OPT_fdiagnostics_path_format_ and
2805 OPT_fdiagnostics_show_path_depths.
2806 * pretty-print.c: Include "diagnostic-event-id.h".
2807 (pp_format): Implement "%@" format code for printing
2808 diagnostic_event_id_t *.
2809 (selftest::test_pp_format): Add tests for "%@".
2810 * selftest-run-tests.c (selftest::run_tests): Call
2811 selftest::tree_diagnostic_path_cc_tests.
2812 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2813 * toplev.c (general_init): Initialize global_dc->path_format and
2814 global_dc->show_path_depths.
2815 * tree-diagnostic-path.cc: New file.
2816 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2817 non-static. Drop "diagnostic" param in favor of storing the
2818 original value of "where" and re-using it.
2819 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2820 maybe_unwind_expanded_macro_loc.
2821 (tree_diagnostics_defaults): Initialize context->print_path and
2822 context->make_json_for_path.
2823 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2825 (default_tree_make_json_for_path): New decl.
2826 (maybe_unwind_expanded_macro_loc): New decl.
2828 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2830 PR tree-optimization/93210
2831 * fold-const.h (native_encode_initializer,
2832 can_native_interpret_type_p): Declare.
2833 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2835 (native_encode_initializer): New function, moved from dwarf2out.c.
2836 Adjust to native_encode_expr compatible arguments, including dry-run
2837 and partial extraction modes. Don't handle STRING_CST.
2838 (can_native_interpret_type_p): No longer static.
2839 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2840 offset / BITS_PER_UNIT fits into int and don't call it if
2841 can_native_interpret_type_p fails. If suboff is NULL and for
2842 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2843 native_encode_initializer.
2844 (fold_const_aggregate_ref_1): Formatting fix.
2845 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2846 (tree_add_const_value_attribute): Adjust caller.
2848 PR tree-optimization/90838
2849 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2850 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2851 CTZ_DEFINED_VALUE_AT_ZERO.
2853 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2856 * lra-constraints.c (match_reload): Permit input operands have the
2857 same mode as output while other input operands have a different
2860 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
2862 PR tree-optimization/90838
2863 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2864 (check_ctz_string): Likewise.
2865 (optimize_count_trailing_zeroes): Likewise.
2866 (simplify_count_trailing_zeroes): Likewise.
2867 (pass_forwprop::execute): Try ctz simplification.
2868 * match.pd: Add matching for ctz idioms.
2870 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2872 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2874 (aarch64_invalid_unary_op): New function for target hook.
2875 (aarch64_invalid_binary_op): New function for target hook.
2877 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2879 * config.gcc: Add arm_bf16.h.
2880 * config/aarch64/aarch64-builtins.c
2881 (aarch64_simd_builtin_std_type): Add BFmode.
2882 (aarch64_init_simd_builtin_types): Define element types for vector
2884 (aarch64_init_bf16_types): New function.
2885 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2886 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2888 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2889 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2891 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2892 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2893 * config/aarch64/aarch64.c
2894 (aarch64_classify_vector_mode): Add support for BF types.
2895 (aarch64_gimplify_va_arg_expr): Add support for BF types.
2896 (aarch64_vq_mode): Add support for BF types.
2897 (aarch64_simd_container_mode): Add support for BF types.
2898 (aarch64_mangle_type): Add support for BF scalar type.
2899 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2900 * config/aarch64/arm_bf16.h: New file.
2901 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2902 * config/aarch64/iterators.md: Add BF types to mode attributes.
2903 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2905 2020-01-10 Jason Merrill <jason@redhat.com>
2907 PR c++/93173 - incorrect tree sharing.
2908 * gimplify.c (copy_if_shared): No longer static.
2909 * gimplify.h: Declare it.
2911 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2913 * doc/invoke.texi (-msve-vector-bits=): Document that
2914 -msve-vector-bits=128 now generates VL-specific code for
2915 little-endian targets.
2916 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2917 build_vector_type_for_mode to construct the data vector types.
2918 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2919 VL-specific code for -msve-vector-bits=128 on little-endian targets.
2920 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2921 for 128-bit vectors.
2923 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2925 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2928 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2930 * config/aarch64/aarch64-builtins.c
2931 (aarch64_builtin_vectorized_function): Check for specific vector modes,
2932 rather than checking the number of elements and the element mode.
2934 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2936 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2937 get_related_vectype_for_scalar_type rather than build_vector_type
2938 to create the index type for a conditional reduction.
2940 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2942 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
2943 for any type of gather or scatter, including strided accesses.
2945 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2947 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
2950 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2952 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
2954 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
2955 parameter and its use to reset DR_OFFSET's.
2956 (vect_transform_loop): Remove orig_drs_init argument.
2957 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
2958 member of dr_vec_info rather than the offset of the associated
2959 data_reference's innermost_loop_behavior.
2960 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
2961 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
2962 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
2963 get_dr_vinfo_offset.
2964 (vectorizable_store): Likewise.
2965 (vectorizable_load): Likewise.
2967 2020-01-10 Richard Biener <rguenther@suse.de>
2969 * gimple-ssa-store-merging
2970 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
2972 2020-01-10 Martin Liska <mliska@suse.cz>
2975 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
2976 encapsulation that was there before r280040.
2978 2020-01-10 Richard Biener <rguenther@suse.de>
2981 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
2982 sequences to avoid walking them again for secondary opportunities.
2983 (pass_lower_eh_dispatch::execute): Instead actually insert
2986 2020-01-10 Richard Biener <rguenther@suse.de>
2989 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
2990 (cleanup_all_empty_eh): Walk landing pads in reverse order to
2991 avoid quadraticness.
2993 2020-01-10 Martin Jambor <mjambor@suse.cz>
2995 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
2996 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
2997 to get param_ipa_sra_max_replacements.
2998 (param_splitting_across_edge): Pass the caller to
2999 pull_accesses_from_callee.
3001 2020-01-10 Martin Jambor <mjambor@suse.cz>
3003 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3004 * ipa-cp.c (max_new_size): Removed.
3005 (orig_overall_size): New variable.
3006 (get_max_overall_size): New function.
3007 (estimate_local_effects): Use it. Adjust dump.
3008 (decide_about_value): Likewise.
3009 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3010 orig_overall_size. Adjust dump.
3011 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3013 2020-01-10 Martin Jambor <mjambor@suse.cz>
3015 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3016 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3017 instead of param_ipa_max_agg_items.
3018 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3019 optimization info for the callee.
3021 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3023 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3024 markers if debug_inline_points is false.
3026 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3028 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3030 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3031 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3032 aarch64-sve-builtins-sve2.h.
3033 (aarch64-sve-builtins-sve2.o): New rule.
3034 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3035 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3036 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3037 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3038 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3039 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3041 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3042 instructions that are handled here.
3043 (@cond_asrd<mode>): Generalize to...
3044 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3045 (*cond_asrd<mode>_2): Generalize to...
3046 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3047 (*cond_asrd<mode>_z): Generalize to...
3048 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3049 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3050 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3051 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3052 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3054 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3055 (@aarch64_scatter_stnt<mode>): Likewise.
3056 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3057 (@aarch64_mul_lane_<mode>): Likewise.
3058 (@aarch64_sve_suqadd<mode>_const): Likewise.
3059 (*<sur>h<addsub><mode>): Generalize to...
3060 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3062 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3063 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3064 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3065 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3066 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3067 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3068 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3069 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3070 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3071 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3072 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3073 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3074 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3075 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3076 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3077 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3078 (@aarch64_sve2_xar<mode>): Likewise.
3079 (@aarch64_sve2_bcax<mode>): Likewise.
3080 (*aarch64_sve2_eor3<mode>): Rename to...
3081 (@aarch64_sve2_eor3<mode>): ...this.
3082 (@aarch64_sve2_bsl<mode>): New expander.
3083 (@aarch64_sve2_nbsl<mode>): Likewise.
3084 (@aarch64_sve2_bsl1n<mode>): Likewise.
3085 (@aarch64_sve2_bsl2n<mode>): Likewise.
3086 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3087 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3088 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3089 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3090 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3091 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3092 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3093 (<su>mull<bt><Vwide>): Generalize to...
3094 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3096 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3097 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3098 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3099 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3100 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3101 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3102 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3103 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3104 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3105 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3106 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3107 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3108 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3109 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3110 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3111 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3112 (<SHRNB:r>shrnb<mode>): Generalize to...
3113 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3115 (<SHRNT:r>shrnt<mode>): Generalize to...
3116 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3118 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3119 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3120 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3121 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3122 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3123 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3124 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3125 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3126 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3127 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3128 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3129 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3130 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3131 (@aarch64_sve2_cvtnt<mode>): Likewise.
3132 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3133 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3134 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3135 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3136 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3137 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3138 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3139 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3140 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3141 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3142 (@aarch64_sve2_pmul<mode>): Likewise.
3143 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3144 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3145 (@aarch64_sve2_tbl2<mode>): Likewise.
3146 (@aarch64_sve2_tbx<mode>): Likewise.
3147 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3148 (@aarch64_sve2_histcnt<mode>): Likewise.
3149 (@aarch64_sve2_histseg<mode>): Likewise.
3150 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3151 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3152 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3153 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3154 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3155 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3156 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3157 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3158 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3159 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3160 (SVE2_PMULL_PAIR_I): New mode iterators.
3161 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3162 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3163 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3164 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3165 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3166 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3167 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3168 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3169 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3170 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3171 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3172 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3173 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3174 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3175 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3176 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3177 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3178 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3179 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3180 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3181 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3182 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3183 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3184 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3185 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3186 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3187 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3188 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3189 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3190 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3191 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3193 (VNARROW, Ventype): New mode attributes.
3194 (Vewtype): Handle VNx2DI. Fix typo in comment.
3195 (VDOUBLE): New mode attribute.
3196 (sve_lane_con): Handle VNx8HI.
3197 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3198 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3199 (sve_int_op, sve_int_op_rev): Handle the above codes.
3200 (sve_pred_int_rhs2_operand): Likewise.
3201 (MULLBT, SHRNB, SHRNT): Delete.
3202 (SVE_INT_SHIFT_IMM): New int iterator.
3203 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3204 and UNSPEC_WHILEHS for TARGET_SVE2.
3205 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3206 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3207 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3208 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3209 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3210 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3211 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3212 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3213 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3214 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3215 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3216 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3217 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3218 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3219 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3220 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3221 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3222 (optab): Handle the new unspecs.
3223 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3225 (lr): Handle the new unspecs.
3227 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3228 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3229 (sve_int_qsub_op): New int attributes.
3230 (sve_fp_op, rot): Handle the new unspecs.
3231 * config/aarch64/aarch64-sve-builtins.h
3232 (function_resolver::require_matching_pointer_type): Declare.
3233 (function_resolver::resolve_unary): Add an optional boolean argument.
3234 (function_resolver::finish_opt_n_resolution): Add an optional
3235 type_suffix_index argument.
3236 (gimple_folder::redirect_call): Declare.
3237 (gimple_expander::prepare_gather_address_operands): Add an optional
3239 * config/aarch64/aarch64-sve-builtins.cc: Include
3240 aarch64-sve-builtins-sve2.h.
3241 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3242 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3243 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3244 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3245 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3246 (TYPES_s_unsigned): Likewise.
3247 (TYPES_s_integer): Use TYPES_s_unsigned.
3248 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3249 (TYPES_sd_integer): Use them.
3250 (TYPES_d_unsigned): New macro.
3251 (TYPES_d_integer): Use it.
3252 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3253 (TYPES_cvt_narrow): Likewise.
3254 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3255 (preds_mx): New variable.
3256 (function_builder::add_overloaded_function): Allow the new feature
3257 set to be more restrictive than the original one.
3258 (function_resolver::infer_pointer_type): Remove qualifiers from
3259 the pointer type before printing it.
3260 (function_resolver::require_matching_pointer_type): New function.
3261 (function_resolver::resolve_sv_displacement): Handle functions
3262 that don't support 32-bit vector indices or svint32_t vector offsets.
3263 (function_resolver::finish_opt_n_resolution): Take the inferred type
3264 as a separate argument.
3265 (function_resolver::resolve_unary): Optionally treat all forms in
3266 the same way as normal merging functions.
3267 (gimple_folder::redirect_call): New function.
3268 (function_expander::prepare_gather_address_operands): Add an argument
3269 that says whether scaled forms are available. If they aren't,
3270 handle scaling of vector indices and don't add the extension and
3272 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3273 fall back to using cond_* instead.
3274 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3275 Split out the member variables into...
3276 (rtx_code_function_base): ...this new base class.
3277 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3278 (unspec_based_function): Split out the member variables into...
3279 (unspec_based_function_base): ...this new base class.
3280 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3281 (unspec_based_function_exact_insn): New class.
3282 (unspec_based_add_function, unspec_based_add_lane_function)
3283 (unspec_based_lane_function, unspec_based_pred_function)
3284 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3285 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3286 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3288 (unspec_based_fused_function): New class.
3289 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3290 (unspec_based_fused_lane_function): New class.
3291 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3293 (CODE_FOR_MODE1): New macro.
3294 (fixed_insn_function): New class.
3295 (while_comparison): Likewise.
3296 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3297 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3298 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3299 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3300 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3301 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3302 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3303 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3304 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3305 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3306 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3307 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3308 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3309 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3310 Also add an initial argument for unary_convert_narrowt, regardless
3311 of the predication type.
3312 (build_32_64): Allow loads and stores to specify MODE_none.
3313 (build_sv_index64, build_sv_uint_offset): New functions.
3314 (long_type_suffix): New function.
3315 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3316 (binary_imm_long_base, load_gather_sv_base): Likewise.
3317 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3318 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3319 (unary_narrowb_base, unary_narrowt_base): Likewise.
3320 (binary_long_lane_def, binary_long_lane): New shape.
3321 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3322 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3323 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3324 (binary_to_uint_def, binary_to_uint): Likewise.
3325 (binary_wide_def, binary_wide): Likewise.
3326 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3327 (compare_def, compare): Likewise.
3328 (compare_ptr_def, compare_ptr): Likewise.
3329 (load_ext_gather_index_restricted_def,
3330 load_ext_gather_index_restricted): Likewise.
3331 (load_ext_gather_offset_restricted_def,
3332 load_ext_gather_offset_restricted): Likewise.
3333 (load_gather_sv_def): Inherit from load_gather_sv_base.
3334 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3335 (shift_left_imm_def, shift_left_imm): Likewise.
3336 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3337 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3338 (store_scatter_index_restricted_def,
3339 store_scatter_index_restricted): Likewise.
3340 (store_scatter_offset_restricted_def,
3341 store_scatter_offset_restricted): Likewise.
3342 (tbl_tuple_def, tbl_tuple): Likewise.
3343 (ternary_long_lane_def, ternary_long_lane): Likewise.
3344 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3345 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3346 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3347 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3348 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3349 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3350 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3351 (ternary_uint_def, ternary_uint): Likewise.
3352 (unary_convert): Fix typo in comment.
3353 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3354 (unary_long_def, unary_long): Likewise.
3355 (unary_narrowb_def, unary_narrowb): Likewise.
3356 (unary_narrowt_def, unary_narrowt): Likewise.
3357 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3358 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3359 (unary_to_int_def, unary_to_int): Likewise.
3360 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3361 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3362 (svasrd_impl): Delete.
3363 (svcadd_impl::expand): Handle integer operations too.
3364 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3365 new functions to derive the unspec numbers.
3366 (svmla_svmls_lane_impl): Replace with...
3367 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3368 integer operations too.
3369 (svwhile_impl): Rename to...
3370 (svwhilelx_impl): ...this and inherit from while_comparison.
3371 (svasrd): Use unspec_based_function.
3372 (svmla_lane): Use svmla_lane_impl.
3373 (svmls_lane): Use svmls_lane_impl.
3374 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3375 (svwhilele, svwhilelt): Use svwhilelx_impl.
3376 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3377 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3378 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3379 * config/aarch64/aarch64-sve-builtins.def: Include
3380 aarch64-sve-builtins-sve2.def.
3382 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3384 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3385 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3386 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3387 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3388 immediates as well as vector ones.
3389 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3390 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3391 (aarch64_sve_qsub_immediate): Update calls accordingly.
3393 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3395 * config/aarch64/aarch64-sve2.md: Add banner comments.
3396 (<su>mulh<r>s<mode>3): Move further up file.
3397 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3398 (*aarch64_sve2_sra<mode>): Move further down file.
3399 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3401 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3403 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3405 (while_optab_cmp): Handle them.
3406 * config/aarch64/aarch64-sve.md
3407 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3408 and add a "@" marker.
3409 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3410 instead of gen_aarch64_sve2_while_ptest.
3411 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3413 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3415 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3416 (UNSPEC_WHILELE): ...this.
3417 (UNSPEC_WHILE_LO): Rename to...
3418 (UNSPEC_WHILELO): ...this.
3419 (UNSPEC_WHILE_LS): Rename to...
3420 (UNSPEC_WHILELS): ...this.
3421 (UNSPEC_WHILE_LT): Rename to...
3422 (UNSPEC_WHILELT): ...this.
3423 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3424 (cmp_op, while_optab_cmp): Likewise.
3425 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3426 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3427 (svwhilelt): Likewise.
3429 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3431 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3432 (unary_to_uint): Define.
3433 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3434 (unary_count): Rename to...
3435 (unary_to_uint_def, unary_to_uint): ...this.
3436 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3438 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3440 * config/aarch64/aarch64-sve-builtins-functions.h
3441 (code_for_mode_function): New class.
3442 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3443 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3444 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3445 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3446 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3448 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3450 * config/aarch64/iterators.md (addsub): New code attribute.
3451 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3453 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3454 in the asm string and attributes. Fix indentation.
3455 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3457 (@aarch64_sve_<optab><mode>): ...this.
3458 * config/aarch64/aarch64-sve-builtins.h
3459 (function_expander::expand_signed_unpred_op): Delete.
3460 * config/aarch64/aarch64-sve-builtins.cc
3461 (function_expander::expand_signed_unpred_op): Likewise.
3462 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3463 try using code_for_aarch64_sve instead.
3464 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3465 (svqsub_impl): Likewise.
3466 (svqadd, svqsub): Use rtx_code_function instead.
3468 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3470 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3471 (HADDSUB, sur, addsub): Remove them.
3473 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3475 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3476 functions the same way, rather than singling out those that
3477 aren't mapped directly to optabs.
3479 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3481 * target.def (compatible_vector_types_p): New target hook.
3482 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3483 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3484 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3485 * doc/tm.texi: Regenerate.
3486 * gimple-expr.c: Include target.h.
3487 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3488 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3490 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3491 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3492 Use the original predicate if it already has a suitable type.
3494 2020-01-09 Martin Jambor <mjambor@suse.cz>
3496 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3497 resolve_speculation and redirect_call_stmt_to_callee static. Change
3498 return type of set_call_stmt to cgraph_edge *.
3499 * auto-profile.c (afdo_indirect_call): Adjust call to
3500 redirect_call_stmt_to_callee.
3501 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3502 make the this pointer explicit, adjust self-recursive calls and the
3503 call top make_direct. Return the resulting edge.
3504 (cgraph_edge::remove): Make this pointer explicit.
3505 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3506 (cgraph_edge::make_direct): Likewise, adjust call to
3507 resolve_speculation.
3508 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3509 call to set_call_stmt.
3510 (cgraph_update_edges_for_call_stmt_node): Update call to
3511 set_call_stmt and remove.
3512 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3513 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3514 (cgraph_node::create_edge_including_clones): Moved "first" definition
3515 of edge to the block where it was used. Adjusted calls to
3517 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3518 cgraph_edge::remove.
3519 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3520 make_direct and redirect_call_stmt_to_callee.
3521 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3522 resolve_speculation and make_direct.
3523 * ipa-inline-transform.c (inline_transform): Adjust call to
3524 redirect_call_stmt_to_callee.
3525 (check_speculations_1):: Adjust call to resolve_speculation.
3526 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3527 resolve-speculation.
3528 (inline_small_functions): Adjust call to resolve_speculation.
3529 (ipa_inline): Likewise.
3530 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3532 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3533 safe with regards to edge removal, adjust calls to
3534 redirect_call_stmt_to_callee.
3535 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3536 and redirect_call_stmt_to_callee.
3537 * multiple_target.c (create_dispatcher_calls): Adjust call to
3538 redirect_call_stmt_to_callee
3539 (redirect_to_specific_clone): Likewise.
3540 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3541 Adjust calls to cgraph_edge::remove.
3542 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3543 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3544 (expand_call_inline): Adjust call to cgraph_edge::remove.
3546 2020-01-09 Martin Liska <mliska@suse.cz>
3548 * params.opt: Set Optimization for
3549 param_max_speculative_devirt_maydefs.
3551 2020-01-09 Martin Sebor <msebor@redhat.com>
3555 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3557 2020-01-09 Martin Liska <mliska@suse.cz>
3559 * auto-profile.c (auto_profile): Use opt_for_fn
3561 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3562 (propagate_vals_across_arith_jfunc): Likewise.
3563 (hint_time_bonus): Likewise.
3564 (incorporate_penalties): Likewise.
3565 (good_cloning_opportunity_p): Likewise.
3566 (perform_estimation_of_a_value): Likewise.
3567 (estimate_local_effects): Likewise.
3568 (ipcp_propagate_stage): Likewise.
3569 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3570 (set_switch_stmt_execution_predicate): Likewise.
3571 (analyze_function_body): Likewise.
3572 * ipa-inline-analysis.c (offline_size): Likewise.
3573 * ipa-inline.c (early_inliner): Likewise.
3574 * ipa-prop.c (ipa_analyze_node): Likewise.
3575 (ipcp_transform_function): Likewise.
3576 * ipa-sra.c (process_scan_results): Likewise.
3577 (ipa_sra_summarize_function): Likewise.
3578 * params.opt: Rename ipcp-unit-growth to
3579 ipa-cp-unit-growth. Add Optimization for various
3580 IPA-related parameters.
3582 2020-01-09 Richard Biener <rguenther@suse.de>
3585 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3587 2020-01-09 Richard Biener <rguenther@suse.de>
3589 PR tree-optimization/93040
3590 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3592 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3594 * common/config/avr/avr-common.c (avr_option_optimization_table)
3595 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3597 2020-01-09 Martin Liska <mliska@suse.cz>
3599 * cgraphclones.c (symbol_table::materialize_all_clones):
3600 Use cgraph_node::dump_name.
3602 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3605 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3606 output_operand_lossage instead of gcc_unreachable.
3607 * doc/md.texi (riscv f constraint): Fix typo.
3610 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3611 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3612 CONST_SCALAR_INT_P instead of CONST_INT_P.
3613 (*subv<mode>4_1): Rename to ...
3614 (subv<mode>4_1): ... this.
3615 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3616 define_insn_and_split patterns.
3617 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3620 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3622 * vec.c (class selftest::count_dtor): New class.
3623 (selftest::test_auto_delete_vec): New test.
3624 (selftest::vec_c_tests): Call it.
3625 * vec.h (class auto_delete_vec): New class template.
3626 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3628 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3630 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3632 2020-01-08 Jim Wilson <jimw@sifive.com>
3634 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3635 use of TLS_MODEL_LOCAL_EXEC when not pic.
3637 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3639 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3642 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3645 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3646 *stack_protect_set_3 peephole2): Also check that the second
3647 insns source is general_operand.
3650 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3651 predicate for output operand instead of register_operand.
3652 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3653 memory destination and non-memory operands[2].
3655 2020-01-08 Martin Liska <mliska@suse.cz>
3657 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3658 ::dump_asm_name instead of (::name or ::asm_name).
3659 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3660 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3661 (analyze_functions): Likewise.
3662 (expand_all_functions): Likewise.
3663 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3664 (propagate_bits_across_jump_function): Likewise.
3665 (dump_profile_updates): Likewise.
3666 (ipcp_store_bits_results): Likewise.
3667 (ipcp_store_vr_results): Likewise.
3668 * ipa-devirt.c (dump_targets): Likewise.
3669 * ipa-fnsummary.c (analyze_function_body): Likewise.
3670 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3671 (process_hsa_functions): Likewise.
3672 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3673 (set_alias_uids): Likewise.
3674 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3675 * ipa-inline.c (recursive_inlining): Likewise.
3676 (inline_to_all_callers_1): Likewise.
3677 (ipa_inline): Likewise.
3678 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3679 (ipa_propagate_frequency): Likewise.
3680 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3681 (remove_described_reference): Likewise.
3682 * ipa-pure-const.c (worse_state): Likewise.
3683 (check_retval_uses): Likewise.
3684 (analyze_function): Likewise.
3685 (propagate_pure_const): Likewise.
3686 (propagate_nothrow): Likewise.
3687 (dump_malloc_lattice): Likewise.
3688 (propagate_malloc): Likewise.
3689 (pass_local_pure_const::execute): Likewise.
3690 * ipa-visibility.c (optimize_weakref): Likewise.
3691 (function_and_variable_visibility): Likewise.
3692 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3693 (ipa_discover_variable_flags): Likewise.
3694 * lto-streamer-out.c (output_function): Likewise.
3695 (output_constructor): Likewise.
3696 * tree-inline.c (copy_bb): Likewise.
3697 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3698 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3700 2020-01-08 Richard Biener <rguenther@suse.de>
3703 * tree-eh.c (sink_clobbers): Update virtual operands for
3704 the first and last stmt only. Add a dry-run capability.
3705 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3706 after CFG manipulations and in RPO order to catch all
3707 secondary opportunities reliably.
3709 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3712 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3714 2019-01-08 Richard Biener <rguenther@suse.de>
3717 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3718 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3719 virtual operand, also updating SSA use.
3720 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3721 Update stmt after resetting virtual operand.
3722 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3723 * gimple-iterator.c (gsi_remove): When not removing the stmt
3724 permanently do not delink immediate uses or mark the stmt modified.
3726 2020-01-08 Martin Liska <mliska@suse.cz>
3728 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3729 (ipa_call_context::estimate_size_and_time): Likewise.
3730 (inline_analyze_function): Likewise.
3732 2020-01-08 Martin Liska <mliska@suse.cz>
3734 * cgraph.c (cgraph_node::dump): Use systematically
3737 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3739 Add -nodevicespecs option for avr.
3742 * config/avr/avr.opt (-nodevicespecs): New driver option.
3743 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3744 "-specs=device-specs/..." if that option is not set.
3745 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3747 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3749 Implement 64-bit double functions for avr.
3752 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3753 --with-double-comparison.
3754 * doc/install.texi: Document them.
3755 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3756 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3757 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3758 * doc/invoke.texi (AVR Built-in Macros): Document them.
3759 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3760 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3761 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3763 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3766 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3767 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3768 when only building rm-profile multilibs.
3770 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3773 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3774 lattice for a value to check.
3775 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3776 finite propagation in self-recursive scc.
3778 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3780 * ipa-inline.c (caller_growth_limits): Restore the AND.
3782 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3784 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3785 (VEC_ALLREG_ALT): New iterator.
3786 (VEC_ALLREG_INT_MODE): New iterator.
3787 (VCMP_MODE): New iterator.
3788 (VCMP_MODE_INT): New iterator.
3789 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3790 (vec_cmp<u>v64qidi): New define_expand.
3791 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3792 (vec_cmpu<mode>di_exec): New define_expand.
3793 (vec_cmp<u>v64qidi_exec): New define_expand.
3794 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3795 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3796 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3797 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3798 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3799 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3800 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3801 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3802 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3803 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3805 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3806 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3808 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3810 * config/gcn/constraints.md (DA): Update description and match.
3812 (Db): New constraint.
3813 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3815 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3816 Implement 'Db' mixed immediate type.
3817 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3818 (addcv64si3_dup<exec_vcc>): Delete.
3819 (subcv64si3<exec_vcc>): Rework constraints.
3820 (addv64di3): Rework constraints.
3821 (addv64di3_exec): Rework constraints.
3822 (subv64di3): Rework constraints.
3823 (addv64di3_dup): Delete.
3824 (addv64di3_dup_exec): Delete.
3825 (addv64di3_zext): Rework constraints.
3826 (addv64di3_zext_exec): Rework constraints.
3827 (addv64di3_zext_dup): Rework constraints.
3828 (addv64di3_zext_dup_exec): Rework constraints.
3829 (addv64di3_zext_dup2): Rework constraints.
3830 (addv64di3_zext_dup2_exec): Rework constraints.
3831 (addv64di3_sext_dup2): Rework constraints.
3832 (addv64di3_sext_dup2_exec): Rework constraints.
3834 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3836 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3837 existing target checks.
3839 2020-01-07 Richard Biener <rguenther@suse.de>
3841 * doc/install.texi: Bump minimal supported MPC version.
3843 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3845 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3846 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3847 * langhooks.c: Include stor-layout.h.
3848 (lhd_simulate_enum_decl): New function.
3849 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3850 handle_arm_sve_h for the LTO frontend.
3851 (register_vector_type): Cope with null returns from pushdecl.
3853 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3855 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3856 (aarch64_sve::nvectors_if_data_type): Replace with...
3857 (aarch64_sve::builtin_type_p): ...this.
3858 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3859 (find_vector_type): Delete.
3860 (add_sve_type_attribute): New function.
3861 (lookup_sve_type_attribute): Likewise.
3862 (register_builtin_types): Add an "SVE type" attribute to each type.
3863 (register_tuple_type): Likewise.
3864 (svbool_type_p, nvectors_if_data_type): Delete.
3865 (mangle_builtin_type): Use lookup_sve_type_attribute.
3866 (builtin_type_p): Likewise. Add an overload that returns the
3867 number of constituent vector and predicate registers.
3868 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3869 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3870 instead of aarch64_sve_argument_p.
3871 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3872 (aarch64_pass_by_reference): Likewise.
3873 (aarch64_function_value_1): Likewise.
3874 (aarch64_return_in_memory): Likewise.
3875 (aarch64_layout_arg): Likewise.
3877 2020-01-07 Jakub Jelinek <jakub@redhat.com>
3879 PR tree-optimization/93156
3880 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3881 least significant bit is always clear.
3883 PR tree-optimization/93118
3884 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
3885 simplifier with two intermediate conversions.
3887 2020-01-07 Martin Liska <mliska@suse.cz>
3889 * params.opt: Add Optimization for various parameters.
3891 2020-01-07 Martin Liska <mliska@suse.cz>
3894 * doc/extend.texi: Explain cloning for target_clone
3897 2020-01-07 Martin Liska <mliska@suse.cz>
3899 PR tree-optimization/92860
3900 * common.opt: Make in Optimization option
3901 as it is affected by -O0, which is an Optimization
3903 * tree-inline.c (tree_inlinable_function_p):
3904 Use opt_for_fn for warn_inline.
3905 (expand_call_inline): Likewise.
3907 2020-01-07 Martin Liska <mliska@suse.cz>
3909 PR tree-optimization/92860
3910 * common.opt: Make flag_ree as optimization
3913 2020-01-07 Martin Liska <mliska@suse.cz>
3915 PR optimization/92860
3916 * params.opt: Mark param_min_crossjump_insns with Optimization
3919 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
3921 * ipa-inline-analysis.c (estimate_growth): Fix typo.
3922 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3924 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
3926 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3927 helper function to return the valid addressing formats for a given
3928 hard register and mode.
3929 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3931 * config/rs6000/constraints.md (Q constraint): Update
3933 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3936 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3937 Use 'Q' for doing vector extract from memory.
3938 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
3940 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
3941 doing vector extract from memory.
3942 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
3943 extract from memory.
3945 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
3946 for the offset being 34-bits when -mcpu=future is used.
3948 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
3950 * config/pa/pa.md: Revert change to use ordered_comparison_operator
3951 instead of cmpib_comparison_operator in cmpib patterns.
3952 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
3953 of cmpib_comparison_operator. Revise comment.
3955 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3957 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
3958 in an IFN_DIV_POW2 node to be equal.
3960 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3962 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
3963 (vect_check_scalar_mask): ...this.
3964 (vectorizable_store, vectorizable_load): Update call accordingly.
3965 (vectorizable_call): Use vect_check_scalar_mask to check the mask
3966 argument in calls to conditional internal functions.
3968 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3970 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
3971 '0' matching inputs.
3972 (subv64di3_exec): Likewise.
3974 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
3976 * config/mips/mips.c (vr4130_align_insns): Fix typo.
3977 * doc/md.texi (movstr): Likewise.
3979 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3981 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
3984 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3986 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
3988 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
3989 to a temporary file and use move-if-change to update the real
3990 file where necessary.
3992 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3994 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
3995 rather than Upa for CPY /M.
3997 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3999 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4002 2020-01-06 Martin Liska <mliska@suse.cz>
4004 PR tree-optimization/92860
4005 * params.opt: Mark param_max_combine_insns with Optimization
4008 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4011 * config/i386/i386.md (SWIDWI): New mode iterator.
4012 (DWI, dwi): Add TImode variants.
4013 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4014 <general_hilo_operand> instead of <general_operand>. Use
4015 CONST_SCALAR_INT_P instead of CONST_INT_P.
4016 (*addv<mode>4_1): Rename to ...
4017 (addv<mode>4_1): ... this.
4018 (QWI): New mode attribute.
4019 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4020 define_insn_and_split patterns.
4021 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4023 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4024 <general_hilo_operand> instead of <general_operand>.
4025 (*addcarry<mode>_1): New define_insn.
4026 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4028 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4030 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4031 Use "call" instead of "set".
4033 2020-01-03 Martin Jambor <mjambor@suse.cz>
4036 * ipa-cp.c (print_all_lattices): Skip functions without info.
4038 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4041 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4042 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4043 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4044 for 'e' simd clones.
4047 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4049 (mprefer-vector-width=): Add Save.
4050 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4051 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4052 (ix86_debug_options, ix86_function_specific_print): Adjust
4053 ix86_target_string callers.
4054 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4055 (ix86_valid_target_attribute_tree): Likewise.
4056 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4057 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4058 ix86_target_string caller.
4061 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4062 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4063 instead of gen_int_shift_amount + convert_modes.
4065 PR rtl-optimization/93088
4066 * loop-iv.c (find_single_def_src): Punt after looking through
4067 128 reg copies for regs with single definitions. Move definitions
4070 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4072 * config/arm/arm-c.c (arm_cpu_builtins): Define
4073 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4074 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4075 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4076 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4077 * config/arm/arm-tables.opt: Regenerated.
4078 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4079 arm_arch_i8mm and arm_arch_bf16 when enabled.
4080 * config/arm/arm.h (TARGET_I8MM): New macro.
4081 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4082 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4083 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4084 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4085 (v8_6_a_simd_variants): New.
4086 (v8_*_a_simd_variants): Add i8mm and bf16.
4087 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4089 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4092 * predict.c (compute_function_frequency): Don't call
4093 warn_function_cold on functions that already have cold attribute.
4095 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4098 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4099 COMDAT group function labels in .data.rel.ro.local section.
4100 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4103 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4104 comparison_operator in B and S integer comparisons. Likewise, use
4105 ordered_comparison_operator instead of cmpib_comparison_operator in
4107 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4109 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4111 Update copyright years.
4113 * gcc.c (process_command): Update copyright notice dates.
4114 * gcov-dump.c (print_version): Ditto.
4115 * gcov.c (print_version): Ditto.
4116 * gcov-tool.c (print_version): Ditto.
4117 * gengtype.c (create_file): Ditto.
4118 * doc/cpp.texi: Bump @copying's copyright year.
4119 * doc/cppinternals.texi: Ditto.
4120 * doc/gcc.texi: Ditto.
4121 * doc/gccint.texi: Ditto.
4122 * doc/gcov.texi: Ditto.
4123 * doc/install.texi: Ditto.
4124 * doc/invoke.texi: Ditto.
4126 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4128 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4131 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4133 PR tree-optimization/93098
4134 * match.pd (popcount): For shift amounts, use integer_onep
4135 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4136 tests. Make sure that precision is power of two larger than or equal
4137 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4138 instead of ULL suffixed constants. Formatting fixes.
4140 Copyright (C) 2020 Free Software Foundation, Inc.
4142 Copying and distribution of this file, with or without modification,
4143 are permitted in any medium without royalty provided the copyright
4144 notice and this notice are preserved.