1 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
4 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
5 conjunction with TARGET_GNU_TLS in early return.
7 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
9 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
10 the mode is not wider than UNITS_PER_WORD.
12 2020-02-14 Martin Jambor <mjambor@suse.cz>
14 PR tree-optimization/93516
15 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
16 access of the same type as the parent.
17 (propagate_subaccesses_from_lhs): Likewise.
19 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
22 * config/i386/avx512vbmi2intrin.h
23 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
24 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
25 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
26 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
27 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
28 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
29 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
30 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
31 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
32 of lacking a closing parenthesis.
33 * config/i386/avx512vbmi2vlintrin.h
34 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
35 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
36 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
37 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
38 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
39 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
40 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
41 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
42 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
43 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
44 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
45 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
46 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
47 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
48 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
49 _mm_shldi_epi32, _mm_mask_shldi_epi32,
50 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
51 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
53 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
56 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
57 the target function entry.
59 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
61 * common/config/arc/arc-common.c (arc_option_optimization_table):
62 Disable if-conversion step when optimized for size.
64 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
66 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
67 R12-R15 are always in ARCOMPACT16_REGS register class.
68 * config/arc/arc.opt (mq-class): Deprecate.
69 * config/arc/constraint.md ("q"): Remove dependency on mq-class
71 * doc/invoke.texi (mq-class): Update text.
72 * common/config/arc/arc-common.c (arc_option_optimization_table):
75 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
77 * config/arc/arc.c (arc_insn_cost): New function.
78 (TARGET_INSN_COST): Define.
79 * config/arc/arc.md (cost): New attribute.
80 (add_n): Use arc_nonmemory_operand.
81 (ashlsi3_insn): Likewise, also update constraints.
82 (ashrsi3_insn): Likewise.
84 (add_shift): Likewise.
85 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
87 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
89 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
91 (umulsidi_600): Likewise.
93 2020-02-13 Jakub Jelinek <jakub@redhat.com>
96 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
97 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
98 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
99 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
100 pass __A to the builtin followed by __W instead of __A followed by
102 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
103 _mm512_mask_popcnt_epi64): Likewise.
104 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
105 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
106 _mm256_mask_popcnt_epi64): Likewise.
108 PR tree-optimization/93582
109 * fold-const.h (shift_bytes_in_array_left,
110 shift_bytes_in_array_right): Declare.
111 * fold-const.c (shift_bytes_in_array_left,
112 shift_bytes_in_array_right): New function, moved from
113 gimple-ssa-store-merging.c, no longer static.
114 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
115 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
116 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
117 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
118 shift_bytes_in_array.
119 (verify_shift_bytes_in_array): Rename to ...
120 (verify_shift_bytes_in_array_left): ... this. Use
121 shift_bytes_in_array_left instead of shift_bytes_in_array.
122 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
123 instead of verify_shift_bytes_in_array.
124 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
125 / native_interpret_expr where the store covers all needed bits,
126 punt on PDP-endian, otherwise allow all involved offsets and sizes
127 not to be byte-aligned.
130 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
131 use const_0_to_255_operand predicate instead of immediate_operand.
132 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
133 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
134 vgf2p8affineinvqb_<mode><mask_name>,
135 vgf2p8affineqb_<mode><mask_name>): Drop mode from
136 const_0_to_255_operand predicated operands.
138 2020-02-12 Jeff Law <law@redhat.com>
140 * config/h8300/h8300.md (comparison shortening peepholes): Use
141 a mode iterator to merge the HImode and SImode peepholes.
143 2020-02-12 Jakub Jelinek <jakub@redhat.com>
146 * real.c (is_even): Make static. Function comment fix.
147 (is_halfway_below): Make static, don't assert R is not inf/nan,
148 instead return false for those. Small formatting fixes.
150 2020-02-12 Martin Sebor <msebor@redhat.com>
153 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
154 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
155 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
156 (strlen_check_and_optimize_call): Adjust callee name.
158 2020-02-12 Jeff Law <law@redhat.com>
160 * config/h8300/h8300.md (comparison shortening peepholes): Drop
161 (and (xor)) variant. Combine other two into single peephole.
163 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
165 PR rtl-optimization/93565
166 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
168 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
170 * config/aarch64/aarch64-simd.md
171 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
172 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
173 generating separate ADDV and zero_extend patterns.
174 * config/aarch64/iterators.md (VDQV_E): New iterator.
176 2020-02-12 Jeff Law <law@redhat.com>
178 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
179 expanders, splits, etc.
180 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
181 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
182 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
183 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
184 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
185 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
187 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
189 2020-02-12 Jakub Jelinek <jakub@redhat.com>
192 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
193 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
194 TARGET_AVX512DQ from condition.
195 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
196 instead of <mask_mode512bit_condition> in condition. If
197 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
199 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
202 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
205 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
207 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
209 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
210 where strlen is more legible.
211 (rs6000_builtin_vectorized_libmass): Ditto.
212 (rs6000_print_options_internal): Ditto.
214 2020-02-11 Martin Sebor <msebor@redhat.com>
216 PR tree-optimization/93683
217 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
219 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
221 * config/rs6000/predicates.md (cint34_operand): Rename the
222 -mprefixed-addr option to be -mprefixed.
223 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
224 the -mprefixed-addr option to be -mprefixed.
225 (OTHER_FUTURE_MASKS): Likewise.
226 (POWERPC_MASKS): Likewise.
227 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
228 the -mprefixed-addr option to be -mprefixed. Change error
229 messages to refer to -mprefixed.
230 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
232 (rs6000_legitimate_offset_address_p): Likewise.
233 (rs6000_mode_dependent_address): Likewise.
234 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
235 "-mprefixed" for target attributes and pragmas.
236 (address_to_insn_form): Rename the -mprefixed-addr option to be
238 (rs6000_adjust_insn_length): Likewise.
239 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
240 -mprefixed-addr option to be -mprefixed.
241 (ASM_OUTPUT_OPCODE): Likewise.
242 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
243 -mprefixed-addr option to be -mprefixed.
244 * config/rs6000/rs6000.opt (-mprefixed): Rename the
245 -mprefixed-addr option to be prefixed. Change the option from
246 being undocumented to being documented.
247 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
248 -mprefixed option. Update the -mpcrel documentation to mention
251 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
253 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
254 including FIRST_PSEUDO_REGISTER - 1.
255 * ira-color.c (print_hard_reg_set): Ditto.
257 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
259 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
260 (USTERNOP_QUALIFIERS): New define.
261 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
262 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
263 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
264 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
265 * config/arm/arm_neon.h (vusdot_s32): New.
266 (vusdot_lane_s32): New.
267 (vusdotq_lane_s32): New.
268 (vsudot_lane_s32): New.
269 (vsudotq_lane_s32): New.
270 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
271 * config/arm/iterators.md (DOTPROD_I8MM): New.
272 (sup, opsuffix): Add <us/su>.
273 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
274 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
276 2020-02-11 Richard Biener <rguenther@suse.de>
278 PR tree-optimization/93661
279 PR tree-optimization/93662
280 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
282 * tree-sra.c (get_access_for_expr): Likewise.
284 2020-02-10 Jakub Jelinek <jakub@redhat.com>
287 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
288 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
289 Change condition from TARGET_AVX2 to TARGET_AVX.
291 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
294 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
297 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
299 Try to generate zero-based comparisons.
300 * config/cris/cris.c (cris_reduce_compare): New function.
301 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
302 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
303 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
305 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
308 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
309 in Thumb state and also as a destination in Arm state. Add T16
312 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
314 * md.texi (Define Subst): Match closing paren in example.
316 2020-02-10 Jakub Jelinek <jakub@redhat.com>
320 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
321 arguments of strncmp.
323 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
326 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
327 but different source value.
328 (adjust_callers_for_value_intersection): New function.
329 (gather_edges_for_value): Adjust order of callers to let a
330 non-self-recursive caller be the first element.
331 (self_recursive_pass_through_p): Add a new parameter "simple", and
332 check generalized self-recursive pass-through jump function.
333 (self_recursive_agg_pass_through_p): Likewise.
334 (find_more_scalar_values_for_callers_subset): Compute value from
335 pass-through jump function for self-recursive.
336 (intersect_with_plats): Cleanup previous implementation code for value
337 itersection with self-recursive call edge.
338 (intersect_with_agg_replacements): Likewise.
339 (intersect_aggregates_with_edge): Deduce value from pass-through jump
340 function for self-recursive call edge. Cleanup previous implementation
341 code for value intersection with self-recursive call edge.
342 (decide_whether_version_node): Remove dead callers and adjust order
343 to let a non-self-recursive caller be the first element.
345 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
347 * recog.c: Move pass_split_before_sched2 code in front of
348 pass_split_before_regstack.
349 (pass_data_split_before_sched2): Rename pass to split3 from split4.
350 (pass_data_split_before_regstack): Rename pass to split4 from split3.
351 (rest_of_handle_split_before_sched2): Remove.
352 (pass_split_before_sched2::execute): Unconditionally call
354 (enable_split_before_sched2): New function.
355 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
356 (pass_split_before_regstack::gate): Ditto.
357 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
358 Update name check for renamed split4 pass.
359 * config/sh/sh.c (register_sh_passes): Update pass insertion
360 point for renamed split4 pass.
362 2020-02-09 Jakub Jelinek <jakub@redhat.com>
364 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
365 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
366 copying them around between host and target.
368 2020-02-08 Andrew Pinski <apinski@marvell.com>
371 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
372 STRICT_ALIGNMENT also.
374 2020-02-08 Jim Wilson <jimw@sifive.com>
377 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
379 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
380 Jakub Jelinek <jakub@redhat.com>
383 * config/i386/i386.h (CALL_USED_REGISTERS): Make
384 xmm16-xmm31 call-used even in 64-bit ms-abi.
386 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
388 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
389 (simd_ummla, simd_usmmla): Likewise.
390 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
391 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
394 2020-02-07 Richard Biener <rguenther@suse.de>
397 * tree-inline.c (fold_marked_statements): Do a PRE walk,
398 skipping unreachable regions.
399 (optimize_inline_calls): Skip folding stmts when we didn't
402 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
405 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
406 Don't return aggregates with only SFmode and DFmode in SSE
408 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
410 2020-02-07 Jakub Jelinek <jakub@redhat.com>
413 * config/rs6000/rs6000-logue.c
414 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
415 if it fails, move rs into end_addr and retry. Add
416 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
417 the insn pattern doesn't describe well what exactly happens to
421 * config/i386/predicates.md (avx_identity_operand): Remove.
422 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
423 (avx_<castmode><avxsizesuffix>_<castmode>,
424 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
425 a VEC_CONCAT of the operand and UNSPEC_CAST.
426 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
427 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
431 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
432 recog_data.insn if distance_non_agu_define changed it.
434 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
437 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
438 we only had X-FORM (reg+reg) addressing for vectors. Also before
439 ISA 3.0, we only had X-FORM addressing for scalars in the
440 traditional Altivec registers.
442 2020-02-06 <zhongyunde@huawei.com>
443 Vladimir Makarov <vmakarov@redhat.com>
445 PR rtl-optimization/93561
446 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
449 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
451 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
454 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
456 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
457 where the low and the high 32 bits are equal to each other specially,
458 with an rldimi instruction.
460 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
462 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
464 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
466 * config/arm/arm-tables.opt: Regenerate.
468 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
471 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
472 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
473 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
475 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
477 PR rtl-optimization/87763
478 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
480 2020-02-06 Delia Burduv <delia.burduv@arm.com>
482 * config/aarch64/aarch64-simd-builtins.def
483 (bfmlaq): New built-in function.
484 (bfmlalb): New built-in function.
485 (bfmlalt): New built-in function.
486 (bfmlalb_lane): New built-in function.
487 (bfmlalt_lane): New built-in function.
488 * config/aarch64/aarch64-simd.md
489 (aarch64_bfmmlaqv4sf): New pattern.
490 (aarch64_bfmlal<bt>v4sf): New pattern.
491 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
492 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
493 (vbfmlalbq_f32): New intrinsic.
494 (vbfmlaltq_f32): New intrinsic.
495 (vbfmlalbq_lane_f32): New intrinsic.
496 (vbfmlaltq_lane_f32): New intrinsic.
497 (vbfmlalbq_laneq_f32): New intrinsic.
498 (vbfmlaltq_laneq_f32): New intrinsic.
499 * config/aarch64/iterators.md (BF_MLA): New int iterator.
500 (bt): New int attribute.
502 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
504 * config/i386/i386.md (*pushtf): Emit "#" instead of
505 calling gcc_unreachable in insn output.
508 (*pushsf_rex64): Ditto for alternatives other than 1.
509 (*pushsf): Ditto for alternatives other than 1.
511 2020-02-06 Martin Liska <mliska@suse.cz>
513 PR gcov-profile/91971
514 PR gcov-profile/93466
515 * coverage.c (coverage_init): Revert mangling of
516 path into filename. It can lead to huge filename length.
517 Creation of subfolders seem more natural.
519 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
522 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
523 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
524 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
526 2020-02-06 Jakub Jelinek <jakub@redhat.com>
529 * config/i386/predicates.md (avx_identity_operand): New predicate.
530 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
531 define_insn_and_split.
534 * omp-low.c (use_pointer_for_field): For nested constructs, also
535 look for map clauses on target construct.
536 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
537 taskreg_nesting_level.
540 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
541 shared clause, call omp_notice_variable on outer context if any.
543 2020-02-05 Jason Merrill <jason@redhat.com>
546 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
547 non-zero address even if weak and not yet defined.
549 2020-02-05 Martin Sebor <msebor@redhat.com>
551 PR tree-optimization/92765
552 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
553 * tree-ssa-strlen.c (compute_string_length): Remove.
554 (determine_min_objsize): Remove.
555 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
556 Avoid using type size as the upper bound on string length.
557 (handle_builtin_string_cmp): Add an argument. Adjust.
558 (strlen_check_and_optimize_call): Pass additional argument to
559 handle_builtin_string_cmp.
561 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
563 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
564 (*pushdi2_rex64 peephole2): Unconditionally split after
566 (*ashl<mode>3_doubleword): Ditto.
567 (*<shift_insn><mode>3_doubleword): Ditto.
569 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
572 * config/rs6000/rs6000.c (get_vector_offset): Fix
574 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
576 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
578 2020-02-05 David Malcolm <dmalcolm@redhat.com>
581 (Special Functions for Debugging the Analyzer): Update description
582 of __analyzer_dump_exploded_nodes.
584 2020-02-05 Jakub Jelinek <jakub@redhat.com>
587 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
588 include sets and not clobbers in the vzeroupper pattern.
589 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
590 the parallel has 17 (64-bit) or 9 (32-bit) elts.
591 (*avx_vzeroupper_1): New define_insn_and_split.
594 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
595 don't run when !optimize.
596 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
599 2020-02-05 Richard Biener <rguenther@suse.de>
602 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
603 checks before matching calls.
605 2020-02-05 Jakub Jelinek <jakub@redhat.com>
607 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
608 function comment typo.
611 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
612 simd_clone_create failed when i == 0, adjust clone->nargs by
615 2020-02-05 Martin Liska <mliska@suse.cz>
618 * doc/invoke.texi: Document that one should
619 not combine ASLR and -fpch.
621 2020-02-04 Richard Biener <rguenther@suse.de>
623 PR tree-optimization/93538
624 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
626 2020-02-04 Richard Biener <rguenther@suse.de>
628 PR tree-optimization/91123
629 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
630 (vn_walk_cb_data::last_vuse): New member.
631 (vn_walk_cb_data::saved_operands): Likewsie.
632 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
633 (vn_walk_cb_data::push_partial_def): Use finish.
634 (vn_reference_lookup_2): Update last_vuse and use finish if
635 we've saved operands.
636 (vn_reference_lookup_3): Use finish and update calls to
637 push_partial_defs everywhere. When translating through
638 memcpy or aggregate copies save off operands and alias-set.
639 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
640 operation for redundant store removal.
642 2020-02-04 Richard Biener <rguenther@suse.de>
644 PR tree-optimization/92819
645 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
646 generating more stmts than before.
648 2020-02-04 Martin Liska <mliska@suse.cz>
650 * config/arm/arm.c (arm_gen_far_branch): Move the function
651 outside of selftests.
653 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
655 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
656 function to adjust PC-relative vector addresses.
657 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
658 handle vectors with PC-relative addresses.
660 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
662 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
664 (hard_reg_and_mode_to_addr_mask): Delete.
665 (rs6000_adjust_vec_address): If the original vector address
666 was REG+REG or REG+OFFSET and the element is not zero, do the add
667 of the elements in the original address before adding the offset
668 for the vector element. Use address_to_insn_form to validate the
669 address using the register being loaded, rather than guessing
670 whether the address is a DS-FORM or DQ-FORM address.
672 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
674 * config/rs6000/rs6000.c (get_vector_offset): New helper function
675 to calculate the offset in memory from the start of a vector of a
676 particular element. Add code to keep the element number in
677 bounds if the element number is variable.
678 (rs6000_adjust_vec_address): Move calculation of offset of the
679 vector element to get_vector_offset.
680 (rs6000_split_vec_extract_var): Do not do the initial AND of
681 element here, move the code to get_vector_offset.
683 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
685 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
688 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
690 * config/rs6000/constraints.md: Improve documentation.
692 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
695 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
696 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
698 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
700 * config.gcc: Remove "carrizo" support.
701 * config/gcn/gcn-opts.h (processor_type): Likewise.
702 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
703 * config/gcn/gcn.opt (gpu_type): Likewise.
704 * config/gcn/t-omp-device: Likewise.
706 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
709 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
710 * config/arm/arm.c (arm_gen_far_branch): New function
712 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
714 2020-02-03 Julian Brown <julian@codesourcery.com>
715 Tobias Burnus <tobias@codesourcery.com>
717 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
719 2020-02-03 Jakub Jelinek <jakub@redhat.com>
722 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
723 valid RTL to sum up the lowest and second lowest bytes of the popcnt
726 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
728 PR rtl-optimization/91333
729 * ira-color.c (struct allocno_color_data): Add member
731 (init_allocno_threads): Set the member up.
732 (bucket_allocno_compare_func): Add compare hard reg
735 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
737 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
739 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
740 * config.in: Regenerated.
741 * configure: Regenerated.
742 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
743 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
744 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
746 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
748 * configure: Regenerate.
750 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
752 PR rtl-optimization/91333
753 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
754 reg preferences comparison up.
756 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
758 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
759 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
760 aarch64-sve-builtins-base.h.
761 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
762 aarch64-sve-builtins-base.cc.
763 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
764 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
766 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
767 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
768 (svcvtnt): New functions.
769 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
770 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
771 (svcvtnt): New functions.
772 (svcvt): Add a form that converts f32 to bf16.
773 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
774 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
776 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
777 Treat B as bfloat16_t.
778 (ternary_bfloat_lane_base): New class.
779 (ternary_bfloat_def): Likewise.
780 (ternary_bfloat): New shape.
781 (ternary_bfloat_lane_def): New class.
782 (ternary_bfloat_lane): New shape.
783 (ternary_bfloat_lanex2_def): New class.
784 (ternary_bfloat_lanex2): New shape.
785 (ternary_bfloat_opt_n_def): New class.
786 (ternary_bfloat_opt_n): New shape.
787 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
788 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
789 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
790 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
791 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
792 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
793 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
794 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
795 the pattern off the narrow mode instead of the wider one.
796 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
797 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
798 (sve_fp_op): Handle them.
799 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
800 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
802 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
804 * config/aarch64/arm_sve.h: Include arm_bf16.h.
805 * config/aarch64/aarch64-modes.def (BF): Move definition before
806 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
807 (SVE_MODES): Handle BF modes.
808 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
810 (aarch64_full_sve_mode): Likewise.
811 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
813 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
814 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
815 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
816 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
818 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
820 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
821 (TYPES_all_data): Add bf16.
822 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
823 (register_tuple_type): Increase buffer size.
824 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
825 (bf16): New type suffix.
826 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
827 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
828 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
829 Change type from all_data to all_arith.
830 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
833 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
834 Matthew Malcomson <matthew.malcomson@arm.com>
835 Richard Sandiford <richard.sandiford@arm.com>
837 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
838 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
839 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
840 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
841 __ARM_FEATURE_MATMUL_FP64.
842 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
843 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
844 be disabled at the same time.
845 (f32mm): New extension.
846 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
847 (AARCH64_FL_F64MM): Bump to the next bit up.
848 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
849 (TARGET_SVE_F64MM): New macros.
850 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
851 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
852 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
853 (UNSPEC_ZIP2Q): New unspeccs.
854 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
855 (optab, sur, perm_insn): Handle the new unspecs.
856 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
857 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
858 TARGET_SVE_F64MM instead of separate tests.
859 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
860 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
861 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
862 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
863 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
864 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
865 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
866 (TYPES_s_signed): New macro.
867 (TYPES_s_integer): Use it.
868 (TYPES_d_float): New macro.
869 (TYPES_d_data): Use it.
870 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
871 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
872 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
873 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
875 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
877 (ternary_resize2_lane_base): Likewise.
878 (ternary_resize2_base): New class.
879 (ternary_qq_lane_base): Likewise.
880 (ternary_intq_uintq_lane_def): Likewise.
881 (ternary_intq_uintq_lane): New shape.
882 (ternary_intq_uintq_opt_n_def): New class
883 (ternary_intq_uintq_opt_n): New shape.
884 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
885 (ternary_uintq_intq_def): New class.
886 (ternary_uintq_intq): New shape.
887 (ternary_uintq_intq_lane_def): New class.
888 (ternary_uintq_intq_lane): New shape.
889 (ternary_uintq_intq_opt_n_def): New class.
890 (ternary_uintq_intq_opt_n): New shape.
891 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
892 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
893 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
894 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
896 (svdotprod_lane_impl): ...this new class.
897 (svmmla_impl, svusdot_impl): New classes.
898 (svdot_lane): Update to use svdotprod_lane_impl.
899 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
900 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
902 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
903 function, with no types defined.
904 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
905 AARCH64_FL_I8MM functions.
906 (svmmla): New AARCH64_FL_F32MM function.
907 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
908 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
909 AARCH64_FL_F64MM function.
910 (REQUIRED_EXTENSIONS):
912 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
914 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
917 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
919 * config/i386/i386.md (*movoi_internal_avx): Do not check for
920 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
921 (*movti_internal): Do not check for
922 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
923 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
924 just after check for TARGET_AVX.
925 (*movdf_internal): Ditto.
926 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
927 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
928 * config/i386/sse.md (mov<mode>_internal): Only check
929 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
930 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
931 (<sse>_andnot<mode>3<mask_name>): Move check for
932 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
933 (<code><mode>3<mask_name>): Ditto.
934 (*andnot<mode>3): Ditto.
936 (*<code><mode>3): Ditto.
938 (*andnot<VI:mode>3): Remove
939 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
940 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
941 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
942 (sse4_1_blendv<ssemodesuffix>): Ditto.
943 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
944 Explain that tune applies to 128bit instructions only.
946 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
948 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
949 to definition of hsa_kernel_description. Parse assembly to find SGPR
950 and VGPR count of kernel and store in hsa_kernel_description.
952 2020-01-31 Tamar Christina <tamar.christina@arm.com>
954 PR rtl-optimization/91838
955 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
956 to truncate if allowed or reject combination.
958 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
960 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
961 (find_inv_vars_cb): Likewise.
963 2020-01-31 David Malcolm <dmalcolm@redhat.com>
965 * calls.c (special_function_p): Split out the check for DECL_NAME
966 being non-NULL and fndecl being extern at file scope into a
967 new maybe_special_function_p and call it. Drop check for fndecl
968 being non-NULL that was after a usage of DECL_NAME (fndecl).
969 * tree.h (maybe_special_function_p): New inline function.
971 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
973 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
974 (mask_gather_load<mode>): ... here, and zero-initialize the
976 (maskload<mode>di): Zero-initialize the destination.
979 2020-01-30 David Malcolm <dmalcolm@redhat.com>
982 * doc/analyzer.texi (Limitations): Note that constraints on
983 floating-point values are currently ignored.
985 2020-01-30 Jakub Jelinek <jakub@redhat.com>
988 * symtab.c (symtab_node::noninterposable_alias): If localalias
989 already exists, but is not usable, append numbers after it until
990 a unique name is found. Formatting fix.
993 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
996 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
998 * config/gcn/gcn.c (print_operand): Handle LTGT.
999 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1001 2020-01-30 Richard Biener <rguenther@suse.de>
1003 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1004 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1006 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1008 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1009 without a DECL in .data.rel.ro.local.
1011 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1014 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1018 * config/i386/sse.md
1019 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1020 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1021 any_extend code iterator instead of always zero_extend.
1022 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1023 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1024 Use any_extend code iterator instead of always zero_extend.
1025 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1026 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1027 Use any_extend code iterator instead of always zero_extend.
1028 (*sse2_pmovmskb_ext): New define_insn.
1029 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1032 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1033 (*popcountsi2_zext_falsedep): New define_insn.
1035 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1037 * config.in: Regenerated.
1038 * configure: Regenerated.
1040 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1043 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1044 LLVM's assembler changed the default in version 9.
1046 2020-01-24 Jeff Law <law@redhat.com>
1048 PR tree-optimization/89689
1049 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1051 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1055 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1057 PR rtl-optimization/87763
1058 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1059 simplification to handle subregs as well as bare regs.
1060 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1062 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1065 * ira.c (ira): Revert use of simplified LRA algorithm.
1067 2020-01-29 Martin Jambor <mjambor@suse.cz>
1069 PR tree-optimization/92706
1070 * tree-sra.c (struct access): Fields first_link, last_link,
1071 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1072 next_rhs_queued and grp_rhs_queued respectively, new fields
1073 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1074 (struct assign_link): Field next renamed to next_rhs, new field
1075 next_lhs. Updated comment.
1076 (work_queue_head): Renamed to rhs_work_queue_head.
1077 (lhs_work_queue_head): New variable.
1078 (add_link_to_lhs): New function.
1079 (relink_to_new_repr): Also relink LHS lists.
1080 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1081 (add_access_to_lhs_work_queue): New function.
1082 (pop_access_from_work_queue): Renamed to
1083 pop_access_from_rhs_work_queue.
1084 (pop_access_from_lhs_work_queue): New function.
1085 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1087 (child_would_conflict_in_lacc): Renamed to
1088 child_would_conflict_in_acc. Adjusted parameter names.
1089 (create_artificial_child_access): New parameter set_grp_read, use it.
1090 (subtree_mark_written_and_enqueue): Renamed to
1091 subtree_mark_written_and_rhs_enqueue.
1092 (propagate_subaccesses_across_link): Renamed to
1093 propagate_subaccesses_from_rhs.
1094 (propagate_subaccesses_from_lhs): New function.
1095 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1098 2020-01-29 Martin Jambor <mjambor@suse.cz>
1100 PR tree-optimization/92706
1101 * tree-sra.c (struct access): Adjust comment of
1102 grp_total_scalarization.
1103 (find_access_in_subtree): Look for single children spanning an entire
1105 (scalarizable_type_p): Allow register accesses, adjust callers.
1106 (completely_scalarize): Remove function.
1107 (scalarize_elem): Likewise.
1108 (create_total_scalarization_access): Likewise.
1109 (sort_and_splice_var_accesses): Do not track total scalarization
1111 (analyze_access_subtree): New parameter totally, adjust to new meaning
1112 of grp_total_scalarization.
1113 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1114 (can_totally_scalarize_forest_p): New function.
1115 (create_total_scalarization_access): Likewise.
1116 (create_total_access_and_reshape): Likewise.
1117 (total_should_skip_creating_access): Likewise.
1118 (totally_scalarize_subtree): Likewise.
1119 (analyze_all_variable_accesses): Perform total scalarization after
1120 subaccess propagation using the new functions above.
1121 (initialize_constant_pool_replacements): Output initializers by
1122 traversing the access tree.
1124 2020-01-29 Martin Jambor <mjambor@suse.cz>
1126 * tree-sra.c (verify_sra_access_forest): New function.
1127 (verify_all_sra_access_forests): Likewise.
1128 (create_artificial_child_access): Set parent.
1129 (analyze_all_variable_accesses): Call the verifier.
1131 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1133 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1134 if called on indirect edge.
1135 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1136 speculative call if needed.
1138 2020-01-29 Richard Biener <rguenther@suse.de>
1140 PR tree-optimization/93428
1141 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1142 permutation when the load node is created.
1143 (vect_analyze_slp_instance): Re-use it here.
1145 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1147 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1149 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1151 PR rtl-optimization/93272
1152 * ira-lives.c (process_out_of_region_eh_regs): New function.
1153 (process_bb_node_lives): Call it.
1155 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1157 * coverage.c (read_counts_file): Make error message lowercase.
1159 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1161 * profile-count.c (profile_quality_display_names): Fix ordering.
1163 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1166 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1167 hash only when edge is first within the sequence.
1168 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1169 (symbol_table::create_edge): Do not set target_prob.
1170 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1172 (cgraph_edge::make_speculative): Drop target_prob parameter.
1173 (cgraph_edge::speculative_call_info): Remove.
1174 (cgraph_edge::first_speculative_call_target): New member function.
1175 (update_call_stmt_hash_for_removing_direct_edge): New function.
1176 (cgraph_edge::resolve_speculation): Rewrite to new API.
1177 (cgraph_edge::speculative_call_for_target): New member function.
1178 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1179 multiple speculation targets.
1180 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1182 (verify_speculative_call): Verify that targets form an interval.
1183 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1184 (cgraph_edge::first_speculative_call_target): New member function.
1185 (cgraph_edge::next_speculative_call_target): New member function.
1186 (cgraph_edge::speculative_call_target_ref): New member function.
1187 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1188 (cgraph_edge): Remove target_prob.
1189 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1190 Fix handling of speculative calls.
1191 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1192 * ipa-fnsummary.c (analyze_function_body): Likewise.
1193 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1194 * ipa-profile.c (dump_histogram): Fix formating.
1195 (ipa_profile_generate_summary): Watch for overflows.
1196 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1197 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1198 (update_indirect_edges_after_inlining): Update to new API.
1199 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1201 * profile-count.h: (profile_probability::adjusted): New.
1202 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1203 updating of profile.
1204 * value-prof.c (gimple_ic_transform): Rename to ...
1205 (dump_ic_profile): ... this one; update dumping.
1206 (stream_in_histogram_value): Fix formating.
1207 (gimple_value_profile_transformations): Update.
1209 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1212 * config/i386/i386.md (*movoi_internal_avx): Remove
1213 TARGET_SSE_TYPELESS_STORES check.
1214 (*movti_internal): Prefer TARGET_AVX over
1215 TARGET_SSE_TYPELESS_STORES.
1216 (*movtf_internal): Likewise.
1217 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1218 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1219 from TARGET_SSE_TYPELESS_STORES.
1221 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1223 * diagnostic-core.h (warning_at): Rename overload to...
1224 (warning_meta): ...this.
1225 (emit_diagnostic_valist): Delete decl of overload taking
1226 diagnostic_metadata.
1227 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1228 (warning_at): Rename overload taking diagnostic_metadata to...
1229 (warning_meta): ...this.
1231 2020-01-28 Richard Biener <rguenther@suse.de>
1233 PR tree-optimization/93439
1234 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1235 * tree-cfg.c (move_sese_region_to_fn): ... here.
1236 (verify_types_in_gimple_reference): Verify used cliques are
1239 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1242 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1243 argument of a pointer to struct gcc_options and pass it to
1244 parse_mtune_ctrl_str.
1245 (ix86_function_specific_restore): Pass opts to
1246 set_ix86_tune_features.
1247 (ix86_option_override_internal): Likewise.
1248 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1249 gcc_options and use it for x_ix86_tune_ctrl_string.
1251 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1253 PR rtl-optimization/87763
1254 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1255 simplification to handle subregs as well as bare regs.
1256 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1258 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1260 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1261 for reduction chains that (now) include a call.
1263 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1265 PR tree-optimization/92822
1266 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1267 out the don't-care elements of a vector whose significant elements
1268 are duplicates, make the don't-care elements duplicates too.
1270 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1272 PR tree-optimization/93434
1273 * tree-predcom.c (split_data_refs_to_components): Record which
1274 components have had aliasing loads removed. Prevent store-store
1275 commoning for all such components.
1277 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1280 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1281 -1 or is_vshift is true, use new_vector with number of elts npatterns
1282 rather than new_unary_operation.
1284 PR tree-optimization/93454
1285 * gimple-fold.c (fold_array_ctor_reference): Perform
1286 elt_size.to_uhwi () just once, instead of calling it in every
1287 iteration. Punt if that value is above size of the temporary
1288 buffer. Decrease third native_encode_expr argument when
1289 bufoff + elt_sz is above size of buf.
1291 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1293 * config/mips/mips.c (mips_declare_object_name)
1294 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1296 2020-01-27 Martin Liska <mliska@suse.cz>
1298 PR gcov-profile/93403
1299 * tree-profile.c (gimple_init_gcov_profiler): Generate
1300 both __gcov_indirect_call_profiler_v4 and
1301 __gcov_indirect_call_profiler_v4_atomic.
1303 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1306 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1308 (@aarch64_split_simd_mov<mode>): Use it.
1309 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1310 Leave the vec_extract patterns to handle 2-element vectors.
1311 (aarch64_simd_mov_from_<mode>high): Likewise.
1312 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1313 (vec_extractv2dfv1df): Likewise.
1315 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1317 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1318 jump conditions for *compare_condjump<GPI:mode>.
1320 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1323 * digraph.cc (test_edge::test_edge): Specify template for base
1326 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1328 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1330 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1332 * config/arc/arc-protos.h (gen_mlo): Remove.
1333 (gen_mhi): Likewise.
1334 * config/arc/arc.c (AUX_MULHI): Define.
1335 (arc_must_save_reister): Special handling for r58/59.
1336 (arc_compute_frame_size): Consider mlo/mhi registers.
1337 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1339 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1340 mlo/mhi name selection.
1341 (arc_restore_callee_saves): Don't early restore blink when ISR.
1342 (arc_expand_prologue): Add mlo/mhi saving.
1343 (arc_expand_epilogue): Add mlo/mhi restoring.
1346 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1347 numbering when MUL64 option is used.
1348 (DWARF2_FRAME_REG_OUT): Define.
1349 * config/arc/arc.md (arc600_stall): New pattern.
1350 (VUNSPEC_ARC_ARC600_STALL): Define.
1351 (mulsi64): Use correct mlo/mhi registers.
1352 (mulsi_600): Clean it up.
1353 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1355 (mhi_operand): Likewise.
1357 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1358 Petro Karashchenko <petro.karashchenko@ring.com>
1360 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1361 attributes if needed.
1362 (prepare_move_operands): Generate special unspec instruction for
1364 (arc_isuncached_mem_p): Propagate uncached attribute to each
1366 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1367 (VUNSPEC_ARC_STDI): Likewise.
1368 (ALLI): New mode iterator.
1369 (mALLI): New mode attribute.
1370 (lddi): New instruction pattern.
1372 (stdidi_split): Split instruction for architectures which are not
1373 supporting ll64 option.
1374 (lddidi_split): Likewise.
1376 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1378 PR rtl-optimization/92989
1379 * lra-lives.c (process_bb_lives): Update the live-in set before
1380 processing additional clobbers.
1382 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1384 PR rtl-optimization/93170
1385 * cselib.c (cselib_invalidate_regno_val): New function, split out
1387 (cselib_invalidate_regno): ...here.
1388 (cselib_invalidated_by_call_p): New function.
1389 (cselib_process_insn): Iterate over all the hard-register entries in
1390 REG_VALUES and invalidate any that cross call-clobbered registers.
1392 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1394 * dojump.c (split_comparison): Use HONOR_NANS rather than
1395 HONOR_SNANS when splitting LTGT.
1397 2020-01-27 Martin Liska <mliska@suse.cz>
1400 * opts.c (print_filtered_help): Exclude language-specific
1401 options from --help=common unless enabled in all FEs.
1403 2020-01-27 Martin Liska <mliska@suse.cz>
1405 * opts.c (print_help): Exclude params from
1406 all except --help=param.
1408 2020-01-27 Martin Liska <mliska@suse.cz>
1411 * config/i386/i386-features.c (make_resolver_func):
1412 Align the code with ppc64 target implementation.
1413 Do not generate a unique name for resolver function.
1415 2020-01-27 Richard Biener <rguenther@suse.de>
1417 PR tree-optimization/93397
1418 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1419 converted reduction chain SLP graph adjustment.
1421 2020-01-26 Marek Polacek <polacek@redhat.com>
1424 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1427 2020-01-26 Jason Merrill <jason@redhat.com>
1430 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1433 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1435 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1436 (rx_setmem): Likewise.
1438 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1441 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1442 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1443 drop <di> from constraint of last operand.
1446 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1447 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1448 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1450 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1453 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1456 2020-01-24 Jeff Law <law@redhat.com>
1458 PR tree-optimization/92788
1459 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1462 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1465 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1466 *avx_vperm_broadcast_<mode>,
1467 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1468 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1469 Move before avx2_perm<mode>/avx512f_perm<mode>.
1472 * simplify-rtx.c (simplify_const_unary_operation,
1473 simplify_const_binary_operation): Punt for mode precision above
1474 MAX_BITSIZE_MODE_ANY_INT.
1476 2020-01-24 Andrew Pinski <apinski@marvell.com>
1478 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1481 2020-01-24 Jeff Law <law@redhat.com>
1484 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1485 for REGs. Call output_operand_lossage to get more reasonable
1488 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1490 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1491 gcn_fp_compare_operator.
1492 (vec_cmpu<mode>di): Use gcn_compare_operator.
1493 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1494 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1495 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1496 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1497 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1498 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1499 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1500 gcn_fp_compare_operator.
1501 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1502 gcn_fp_compare_operator.
1503 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1504 gcn_fp_compare_operator.
1505 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1506 gcn_fp_compare_operator.
1508 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1510 * doc/install.texi (Cross-Compiler-Specific Options): Document
1511 `--with-toolexeclibdir' option.
1513 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1515 * target.def (flags_regnum): Also mention effect on delay slot filling.
1516 * doc/tm.texi: Regenerate.
1518 2020-01-23 Jeff Law <law@redhat.com>
1520 PR translation/90162
1521 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1523 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1526 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1529 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1531 PR rtl-optimization/93402
1532 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1535 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1537 * config.in: Regenerated.
1538 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1539 for TARGET_LIBC_GNUSTACK.
1540 * configure: Regenerated.
1541 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1542 found to be 2.31 or greater.
1544 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1546 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1548 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1549 (mips_asm_file_end): New function. Delegate to
1550 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1551 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1553 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1556 * config/i386/i386-modes.def (POImode): New mode.
1557 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1558 * config/i386/i386.md (DPWI): New mode attribute.
1559 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1561 (QPWI): ... this. Use POI instead of OI for TImode.
1562 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1563 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1566 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1569 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1571 (speculation_tracker_rev): New pattern.
1572 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1573 Use speculation_tracker_rev to track the inverse condition.
1575 2020-01-23 Richard Biener <rguenther@suse.de>
1577 PR tree-optimization/93381
1578 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1579 alias-set of the def as argument and record the first one.
1580 (vn_walk_cb_data::first_set): New member.
1581 (vn_reference_lookup_3): Pass the alias-set of the current def
1582 to push_partial_def. Fix alias-set used in the aggregate copy
1584 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1585 * real.c (clear_significand_below): Fix out-of-bound access.
1587 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1590 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1591 New define_insn patterns.
1593 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1595 * doc/sourcebuild.texi (check-function-bodies): Add an
1596 optional target/xfail selector.
1598 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1600 PR rtl-optimization/93124
1601 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1602 bare USE and CLOBBER insns.
1604 2020-01-22 Andrew Pinski <apinski@marvell.com>
1606 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1608 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1611 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1612 diagnostic_manager into "ana" namespace.
1613 * selftest-run-tests.c (selftest::run_tests): Update for move of
1614 selftest::run_analyzer_selftests to
1615 ana::selftest::run_analyzer_selftests.
1617 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1619 * cfgexpand.c (union_stack_vars): Update the size.
1621 2020-01-22 Richard Biener <rguenther@suse.de>
1623 PR tree-optimization/93381
1624 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1625 throughout, handle all conversions the same.
1627 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1630 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1631 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1632 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1633 Call force_reg on high_in2 unconditionally.
1635 2020-01-22 Martin Liska <mliska@suse.cz>
1637 PR tree-optimization/92924
1638 * profile.c (compute_value_histograms): Divide
1641 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1644 * output.h (assemble_name_resolve): Declare.
1645 * varasm.c (assemble_name_resolve): New function.
1646 (assemble_name): Use it.
1647 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1649 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1651 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1652 update_web_docs_git instead of update_web_docs_svn.
1654 2020-01-21 Andrew Pinski <apinski@marvell.com>
1657 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1658 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1659 (*tlsgd_small_<mode>): Likewise.
1660 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1661 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1662 register. Convert that register back to dest using convert_mode.
1664 2020-01-21 Jim Wilson <jimw@sifive.com>
1666 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1669 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1670 Uros Bizjak <ubizjak@gmail.com>
1673 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1675 (legitimize_tls_address): Do GNU2 TLS address computation in
1676 ptr_mode and zero-extend result to Pmode.
1677 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1678 :P with :PTR and Pmode with ptr_mode.
1679 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1680 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1681 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1683 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1686 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1687 the last two operands are CONST_INT_P before using them as such.
1689 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1691 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1692 to get the integer element types.
1694 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1696 * config/aarch64/aarch64-sve-builtins.h
1697 (function_expander::convert_to_pmode): Declare.
1698 * config/aarch64/aarch64-sve-builtins.cc
1699 (function_expander::convert_to_pmode): New function.
1700 (function_expander::get_contiguous_base): Use it.
1701 (function_expander::prepare_gather_address_operands): Likewise.
1702 * config/aarch64/aarch64-sve-builtins-sve2.cc
1703 (svwhilerw_svwhilewr_impl::expand): Likewise.
1705 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1708 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1709 cfun->machine->label_is_assembled.
1710 (aarch64_print_patchable_function_entry): New.
1711 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1712 * config/aarch64/aarch64.h (struct machine_function): New field,
1715 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1718 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1721 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1724 * cgraph.c (cgraph_edge::resolve_speculation,
1725 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1726 call_stmt_site_hash.
1728 2020-01-21 Martin Liska <mliska@suse.cz>
1730 * config/rs6000/rs6000.c (common_mode_defined): Remove
1733 2020-01-21 Richard Biener <rguenther@suse.de>
1735 PR tree-optimization/92328
1736 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1737 type when value-numbering same-sized store by inserting a
1739 (eliminate_dom_walker::eliminate_stmt): When eliminating
1740 a redundant store handle bit-reinterpretation of the same value.
1742 2020-01-21 Andrew Pinski <apinski@marvel.com>
1745 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1747 (prepare_block_for_update): This. Use a worklist instead of
1750 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1752 * gcc/config/arm/arm.c (clear_operation_p):
1753 Initialise last_regno, skip first iteration
1754 based on the first_set value and use ints instead
1755 of the unnecessary HOST_WIDE_INTs.
1757 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1760 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1761 compare_mode other than SFmode or DFmode.
1763 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1766 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1767 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1768 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1770 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1772 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1774 2020-01-20 Andrew Pinski <apinski@marvell.com>
1777 * targhooks.c (default_print_patchable_function_entry): Use
1778 output_asm_insn to emit the nop instruction.
1780 2020-01-20 Fangrui Song <maskray@google.com>
1783 * targhooks.c (default_print_patchable_function_entry): Align to
1786 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
1789 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1790 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
1791 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1792 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
1793 (*tls_dynamic_gnu2_lea_64): Renamed to ...
1794 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
1795 Remove the {q} suffix from lea.
1796 (*tls_dynamic_gnu2_call_64): Renamed to ...
1797 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
1798 (*tls_dynamic_gnu2_combine_64): Renamed to ...
1799 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
1800 Pass Pmode to gen_tls_dynamic_gnu2_64.
1802 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1804 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1806 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
1808 * config/aarch64/aarch64-sve-builtins-base.cc
1809 (svld1ro_impl::memory_vector_mode): Remove parameter name.
1811 2020-01-20 Richard Biener <rguenther@suse.de>
1814 * dwarf2out.c (prune_unused_types): Unconditionally mark
1815 called function DIEs.
1817 2020-01-20 Martin Liska <mliska@suse.cz>
1819 PR tree-optimization/93199
1820 * tree-eh.c (struct leh_state): Add
1821 new field outer_non_cleanup.
1822 (cleanup_is_dead_in): Pass leh_state instead
1823 of eh_region. Add a checking that state->outer_non_cleanup
1824 points to outer non-clean up region.
1825 (lower_try_finally): Record outer_non_cleanup
1827 (lower_catch): Likewise.
1828 (lower_eh_filter): Likewise.
1829 (lower_eh_must_not_throw): Likewise.
1830 (lower_cleanup): Likewise.
1832 2020-01-20 Richard Biener <rguenther@suse.de>
1834 PR tree-optimization/93094
1835 * tree-vectorizer.h (vect_loop_versioning): Adjust.
1836 (vect_transform_loop): Likewise.
1837 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1838 loop_vectorized_call to vect_transform_loop.
1839 * tree-vect-loop.c (vect_transform_loop): Pass down
1840 loop_vectorized_call to vect_loop_versioning.
1841 * tree-vect-loop-manip.c (vect_loop_versioning): Use
1842 the earlier discovered loop_vectorized_call.
1844 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
1846 * doc/contribute.texi: Update for SVN -> Git transition.
1847 * doc/install.texi: Likewise.
1849 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1852 * cgraph.c (cgraph_edge::make_speculative): Increase number of
1853 speculative targets.
1854 (verify_speculative_call): New function
1855 (cgraph_node::verify_node): Use it.
1856 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
1859 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1862 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
1863 (cgraph_edge::make_direct): Remove all indirect targets.
1864 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
1865 (cgraph_node::verify_node): Verify that only one call_stmt or
1866 lto_stmt_uid is set.
1867 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
1869 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
1870 (lto_output_ref): Simplify streaming of stmt.
1871 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
1873 2020-01-18 Tamar Christina <tamar.christina@arm.com>
1875 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
1876 Mark parameter unused.
1878 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
1880 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
1882 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
1884 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
1886 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
1888 * Makefile.in: Add coroutine-passes.o.
1889 * builtin-types.def (BT_CONST_SIZE): New.
1890 (BT_FN_BOOL_PTR): New.
1891 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
1892 * builtins.def (DEF_COROUTINE_BUILTIN): New.
1893 * coroutine-builtins.def: New file.
1894 * coroutine-passes.cc: New file.
1895 * function.h (struct GTY function): Add a bit to indicate that the
1896 function is a coroutine component.
1897 * internal-fn.c (expand_CO_FRAME): New.
1898 (expand_CO_YIELD): New.
1899 (expand_CO_SUSPN): New.
1900 (expand_CO_ACTOR): New.
1901 * internal-fn.def (CO_ACTOR): New.
1905 * passes.def: Add pass_coroutine_lower_builtins,
1906 pass_coroutine_early_expand_ifns.
1907 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
1908 (make_pass_coroutine_early_expand_ifns): New.
1909 * doc/invoke.texi: Document the fcoroutines command line
1912 2020-01-18 Jakub Jelinek <jakub@redhat.com>
1914 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
1917 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
1918 after checking the argument is a REG. Don't use REGNO (reg)
1919 again to set last_regno, reuse regno variable instead.
1921 2020-01-17 David Malcolm <dmalcolm@redhat.com>
1923 * doc/analyzer.texi (Limitations): Add note about NaN.
1925 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1926 Sudakshina Das <sudi.das@arm.com>
1928 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
1929 and valid immediate.
1930 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
1931 (lshrdi3): Generate thumb2_lsrl for valid immediates.
1932 * config/arm/constraints.md (Pg): New.
1933 * config/arm/predicates.md (long_shift_imm): New.
1934 (arm_reg_or_long_shift_imm): Likewise.
1935 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
1936 (thumb2_lsll): Likewise.
1939 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1940 Sudakshina Das <sudi.das@arm.com>
1942 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
1943 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
1944 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
1945 register pairs for doubleword quantities for ARMv8.1M-Mainline.
1946 * config/arm/thumb2.md (thumb2_asrl): New.
1947 (thumb2_lsll): Likewise.
1949 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1951 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
1954 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
1956 * gdbinit.in (help-gcc-hooks): New command.
1957 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
1958 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
1961 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1963 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
1964 correct target macro.
1966 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1968 * config/aarch64/aarch64-protos.h
1969 (aarch64_sve_ld1ro_operand_p): New.
1970 * config/aarch64/aarch64-sve-builtins-base.cc
1971 (class load_replicate): New.
1972 (class svld1ro_impl): New.
1973 (class svld1rq_impl): Change to inherit from load_replicate.
1974 (svld1ro): New sve intrinsic function base.
1975 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
1976 New DEF_SVE_FUNCTION.
1977 * config/aarch64/aarch64-sve-builtins-base.h
1978 (svld1ro): New decl.
1979 * config/aarch64/aarch64-sve-builtins.cc
1980 (function_expander::add_mem_operand): Modify assert to allow
1982 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
1984 * config/aarch64/aarch64.c
1985 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
1986 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
1987 (aarch64_sve_ld1ro_operand_p): New.
1988 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
1989 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
1990 * config/aarch64/predicates.md
1991 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
1993 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1995 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
1996 Introduce this ACLE specified predefined macro.
1997 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
1998 (fp): Disabling this disables f64mm.
1999 (simd): Disabling this disables f64mm.
2000 (fp16): Disabling this disables f64mm.
2001 (sve): Disabling this disables f64mm.
2002 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2003 (AARCH64_ISA_F64MM): New.
2004 (TARGET_F64MM): New.
2005 * doc/invoke.texi (f64mm): Document new option.
2007 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2009 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2010 (neoversen1_tunings): Likewise.
2012 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2015 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2016 Add assert to ensure prolog has been emitted.
2017 (aarch64_split_atomic_op): Likewise.
2018 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2019 Use epilogue_completed rather than reload_completed.
2020 (aarch64_atomic_exchange<mode>): Likewise.
2021 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2022 (atomic_nand<mode>): Likewise.
2023 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2024 (atomic_fetch_nand<mode>): Likewise.
2025 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2026 (atomic_nand_fetch<mode>): Likewise.
2028 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2031 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2033 (REVERSE_CONDITION): Delete.
2034 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2035 (CCFP_CCFPE): Likewise.
2036 (e): New mode attribute.
2037 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2038 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2039 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2040 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2041 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2042 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2043 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2044 name of generator from gen_ccmpdi to gen_ccmpccdi.
2045 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2046 the previous comparison but aren't able to, use the new ccmp_rev
2049 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2051 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2052 than testing directly for INTEGER_CST.
2053 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2055 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2057 PR tree-optimization/93292
2058 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2059 get_vectype_for_scalar_type returns NULL.
2061 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2063 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2064 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2066 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2068 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2070 * params.opt: (max-predicted-iterations): Set bounds.
2071 * predict.c (real_almost_one, real_br_prob_base,
2072 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2073 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2074 probabilities; do not truncate to reg_br_prob_bases.
2075 (estimate_loops_at_level): Pass max_cyclic_prob.
2076 (estimate_loops): Compute max_cyclic_prob.
2077 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2079 * profile-count.c (profile_probability::to_sreal): New.
2080 * profile-count.h (class sreal): Move up in file.
2081 (profile_probability::to_sreal): Declare.
2083 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2086 (arm_invalid_conversion): New function for target hook.
2087 (arm_invalid_unary_op): New function for target hook.
2088 (arm_invalid_binary_op): New function for target hook.
2090 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2092 * config.gcc: Add arm_bf16.h.
2093 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2094 (arm_simd_builtin_std_type): Add BFmode.
2095 (arm_init_simd_builtin_types): Define element types for vector types.
2096 (arm_init_bf16_types): New function.
2097 (arm_init_builtins): Add arm_init_bf16_types function call.
2098 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2099 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2100 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2101 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2102 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2103 (arm_mangle_type): Add __bf16.
2104 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2105 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2106 arm_bf16_ptr_type_node.
2107 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2108 define_split between ARM registers.
2109 * config/arm/arm_bf16.h: New file.
2110 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2111 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2113 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2114 * config/arm/vfp.md: Add BFmode to movhf patterns.
2116 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2117 Andre Vieira <andre.simoesdiasvieira@arm.com>
2119 * config/arm/arm-cpus.in (mve, mve_float): New features.
2120 (dsp, mve, mve.fp): New options.
2121 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2122 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2123 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2125 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2126 Thomas Preud'homme <thomas.preudhomme@arm.com>
2128 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2130 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2131 error for using -mcmse when targeting Armv8.1-M Mainline.
2133 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2134 Thomas Preud'homme <thomas.preudhomme@arm.com>
2136 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2137 address in r4 when targeting Armv8.1-M Mainline.
2138 (nonsecure_call_value_internal): Likewise.
2139 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2140 a register match_operand again. Emit BLXNS when targeting
2142 (nonsecure_call_value_reg_thumb2): Likewise.
2144 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2145 Thomas Preud'homme <thomas.preudhomme@arm.com>
2147 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2148 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2149 variable as true when floating-point ABI is not hard. Replace
2150 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2151 Generate VLSTM and VLLDM instruction respectively before and
2152 after a function call to cmse_nonsecure_call function.
2153 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2154 (VUNSPEC_VLLDM): Likewise.
2155 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2156 (lazy_load_multiple_insn): Likewise.
2158 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2159 Thomas Preud'homme <thomas.preudhomme@arm.com>
2161 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2162 (arm_emit_vfp_multi_reg_pop): Likewise.
2163 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2164 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2165 restore callee-saved VFP registers.
2167 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2168 Thomas Preud'homme <thomas.preudhomme@arm.com>
2170 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2171 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2172 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2173 callee-saved GPRs as well as clear ip register before doing a nonsecure
2174 call then restore callee-saved GPRs after it when targeting
2176 (arm_reorg): Adapt to function rename.
2178 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2179 Thomas Preud'homme <thomas.preudhomme@arm.com>
2181 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2182 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2183 clear_vfp_multiple pattern based on a new vfp parameter.
2184 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2185 targeting Armv8.1-M Mainline.
2186 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2187 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2188 whether VFP registers are available before looking call_used_regs for a
2190 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2191 of prototype of clear_operation_p.
2192 (clear_vfp_multiple_operation): New predicate.
2193 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2194 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2196 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2197 Thomas Preud'homme <thomas.preudhomme@arm.com>
2199 * config/arm/arm-protos.h (clear_operation_p): Declare.
2200 * config/arm/arm.c (clear_operation_p): New function.
2201 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2202 targeting Armv8.1-M Mainline or successor.
2203 (output_return_instruction): Only output APSR register clearing if
2204 Armv8.1-M Mainline instructions not available.
2205 (thumb_exit): Likewise.
2206 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2207 * config/arm/thumb2.md (clear_apsr): New define_insn.
2208 (clear_multiple): Likewise.
2209 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2211 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2212 Thomas Preud'homme <thomas.preudhomme@arm.com>
2214 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2215 (use_return_insn): Also return false for Armv8.1-M Mainline.
2216 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2217 Mainline instructions are available.
2218 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2219 when targeting Armv8.1-M Mainline Security Extensions.
2220 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2221 Mainline entry function.
2222 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2223 targeting Armv8.1-M Mainline or successor.
2224 (arm_expand_epilogue): Fix indentation of caller-saved register
2225 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2227 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2228 (FP_SYSREGS): Likewise.
2229 (enum vfp_sysregs_encoding): Define enum.
2230 (fp_sysreg_names): Declare.
2231 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2232 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2233 (pop_fpsysreg_insn): Likewise.
2235 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2236 Thomas Preud'homme <thomas.preudhomme@arm.com>
2238 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2239 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2240 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2241 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2242 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2243 (ARMv8_1m_main): New feature group.
2244 (armv8.1-m.main): New architecture.
2245 * config/arm/arm-tables.opt: Regenerate.
2246 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2247 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2248 (arm_options_perform_arch_sanity_checks): Error out when targeting
2249 Armv8.1-M Mainline Security Extensions.
2250 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2252 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2254 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2255 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2256 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2257 aarch64_bfdot_laneq): New.
2258 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2259 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2260 vbfdotq_laneq_f32): New.
2261 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2262 VBFMLA_W, VBF): New.
2263 (isquadop): Add V4BF, V8BF.
2265 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2267 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2268 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2269 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2270 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2271 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2272 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2273 usdot_laneq, sudot_lane,sudot_laneq): New.
2274 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2275 (aarch64_<sur>dot_lane): New.
2276 * config/aarch64/arm_neon.h (vusdot_s32): New.
2278 (vusdot_lane_s32): New.
2279 (vsudot_lane_s32): New.
2280 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2281 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2283 2020-01-16 Martin Liska <mliska@suse.cz>
2285 * value-prof.c (dump_histogram_value): Fix
2286 obvious spacing issue.
2288 2020-01-16 Andrew Pinski <apinski@marvell.com>
2290 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2291 !storage_order_barrier_p.
2293 2020-01-16 Andrew Pinski <apinski@marvell.com>
2295 * sched-int.h (_dep): Add unused bit-field field for the padding.
2296 * sched-deps.c (init_dep_1): Init unused field.
2298 2020-01-16 Andrew Pinski <apinski@marvell.com>
2300 * optabs.h (create_expand_operand): Initialize target field also.
2302 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2304 PR tree-optimization/92429
2305 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2306 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2308 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2311 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2313 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2314 aarch64_sve_int_mode to each mode.
2316 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2318 * doc/analyzer.texi (Overview): Add note about
2319 -fdump-ipa-analyzer.
2321 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2323 PR tree-optimization/93231
2324 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2325 input_type is unsigned. Use tree_to_shwi for shift constant.
2326 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2327 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2328 inputs more efficiently.
2330 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2332 * config/i386/i386.md (*movsf_internal): Do not require
2333 SSE2 ISA for alternatives 14 and 15.
2335 2020-01-15 Richard Biener <rguenther@suse.de>
2338 * tree-eh.c (sink_clobbers): If we already visited the destination
2339 block do not defer insertion.
2340 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2341 the purpose of defered insertion.
2343 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2345 * BASE-VER: Bump to 10.0.1.
2347 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2349 PR tree-optimization/93247
2350 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2351 type of the stmt that we're going to vectorize.
2353 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2355 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2356 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2359 2020-01-15 Martin Liska <mliska@suse.cz>
2361 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2362 2 calls of streamer_read_hwi in a function call.
2364 2020-01-15 Richard Biener <rguenther@suse.de>
2366 * alias.c (record_alias_subset): Avoid redundant work when
2367 subset is already recorded.
2369 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2371 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2372 the analyzer options provide CWE identifiers.
2374 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2376 * tree-diagnostic-path.cc (path_summary::event_range::print):
2377 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2378 using get_pure_location.
2380 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2382 PR tree-optimization/93262
2383 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2384 perform head trimming only if the last argument is constant,
2385 either all ones, or larger or equal to head trim, in the latter
2386 case decrease the last argument by head_trim.
2388 PR tree-optimization/93249
2389 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2390 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2391 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2392 perform head trim unless we can prove there are no '\0' chars
2393 from the source among the first head_trim chars.
2395 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2397 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2399 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2402 * config/i386/sse.md
2403 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2404 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2405 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2406 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2407 just a single alternative instead of two, make operands 1 and 2
2410 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2413 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2416 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2418 * Makefile.in (lang_opt_files): Add analyzer.opt.
2419 (ANALYZER_OBJS): New.
2420 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2421 tristate.o and ANALYZER_OBJS.
2422 (TEXI_GCCINT_FILES): Add analyzer.texi.
2423 * common.opt (-fanalyzer): New driver option.
2424 * config.in: Regenerate.
2425 * configure: Regenerate.
2426 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2427 (gccdepdir): Also create depdir for "analyzer" subdir.
2428 * digraph.cc: New file.
2429 * digraph.h: New file.
2430 * doc/analyzer.texi: New file.
2431 * doc/gccint.texi ("Static Analyzer") New menu item.
2432 (analyzer.texi): Include it.
2433 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2434 ("Warning Options"): Add static analysis warnings to the list.
2435 (-Wno-analyzer-double-fclose): New option.
2436 (-Wno-analyzer-double-free): New option.
2437 (-Wno-analyzer-exposure-through-output-file): New option.
2438 (-Wno-analyzer-file-leak): New option.
2439 (-Wno-analyzer-free-of-non-heap): New option.
2440 (-Wno-analyzer-malloc-leak): New option.
2441 (-Wno-analyzer-possible-null-argument): New option.
2442 (-Wno-analyzer-possible-null-dereference): New option.
2443 (-Wno-analyzer-null-argument): New option.
2444 (-Wno-analyzer-null-dereference): New option.
2445 (-Wno-analyzer-stale-setjmp-buffer): New option.
2446 (-Wno-analyzer-tainted-array-index): New option.
2447 (-Wno-analyzer-use-after-free): New option.
2448 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2449 (-Wno-analyzer-use-of-uninitialized-value): New option.
2450 (-Wanalyzer-too-complex): New option.
2451 (-fanalyzer-call-summaries): New warning.
2452 (-fanalyzer-checker=): New warning.
2453 (-fanalyzer-fine-grained): New warning.
2454 (-fno-analyzer-state-merge): New warning.
2455 (-fno-analyzer-state-purge): New warning.
2456 (-fanalyzer-transitivity): New warning.
2457 (-fanalyzer-verbose-edges): New warning.
2458 (-fanalyzer-verbose-state-changes): New warning.
2459 (-fanalyzer-verbosity=): New warning.
2460 (-fdump-analyzer): New warning.
2461 (-fdump-analyzer-callgraph): New warning.
2462 (-fdump-analyzer-exploded-graph): New warning.
2463 (-fdump-analyzer-exploded-nodes): New warning.
2464 (-fdump-analyzer-exploded-nodes-2): New warning.
2465 (-fdump-analyzer-exploded-nodes-3): New warning.
2466 (-fdump-analyzer-supergraph): New warning.
2467 * doc/sourcebuild.texi (dg-require-dot): New.
2468 (dg-check-dot): New.
2469 * gdbinit.in (break-on-saved-diagnostic): New command.
2470 * graphviz.cc: New file.
2471 * graphviz.h: New file.
2472 * ordered-hash-map-tests.cc: New file.
2473 * ordered-hash-map.h: New file.
2474 * passes.def (pass_analyzer): Add before
2475 pass_ipa_whole_program_visibility.
2476 * selftest-run-tests.c (selftest::run_tests): Call
2477 selftest::ordered_hash_map_tests_cc_tests.
2478 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2480 * shortest-paths.h: New file.
2481 * timevar.def (TV_ANALYZER): New timevar.
2482 (TV_ANALYZER_SUPERGRAPH): Likewise.
2483 (TV_ANALYZER_STATE_PURGE): Likewise.
2484 (TV_ANALYZER_PLAN): Likewise.
2485 (TV_ANALYZER_SCC): Likewise.
2486 (TV_ANALYZER_WORKLIST): Likewise.
2487 (TV_ANALYZER_DUMP): Likewise.
2488 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2489 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2490 * tree-pass.h (make_pass_analyzer): New decl.
2491 * tristate.cc: New file.
2492 * tristate.h: New file.
2494 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2497 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2498 alternatives 9 and 10.
2500 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2502 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2503 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2504 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2505 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2506 (selftest::hash_map_tests_c_tests): Call it.
2507 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2508 New static constant, using the value of = H::empty_zero_p.
2509 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2510 from default_hash_traits <Value>.
2511 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2513 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2514 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2515 calls to mark_empty with !Descriptor::empty_zero_p.
2516 (hash_table::empty_slow): Conditionalize the memset call with a
2517 check that Descriptor::empty_zero_p; otherwise, loop through the
2518 entries calling mark_empty on them.
2519 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2520 (pointer_hash::empty_zero_p): Likewise.
2521 (pair_hash::empty_zero_p): Likewise.
2522 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2524 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2525 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2526 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2527 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2528 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2529 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2530 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2531 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2533 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2536 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2538 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2539 fix typo on return value.
2541 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2544 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2546 (cgraph_edge::make_speculative): Add param for setting speculative_id
2548 (cgraph_edge::speculative_call_info): Update comments and find reference
2549 by speculative_id for multiple indirect targets.
2550 (cgraph_edge::resolve_speculation): Decrease the speculations
2551 for indirect edge, drop it's speculative if not direct target
2552 left. Update comments.
2553 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2554 (cgraph_node::dump): Print num_speculative_call_targets.
2555 (cgraph_node::verify_node): Don't report error if speculative
2556 edge not include statement.
2557 (cgraph_edge::num_speculative_call_targets_p): New function.
2558 * cgraph.h (int common_target_id): Remove.
2559 (int common_target_probability): Remove.
2560 (num_speculative_call_targets): New variable.
2561 (make_speculative): Add param for setting speculative_id.
2562 (cgraph_edge::num_speculative_call_targets_p): New declare.
2563 (target_prob): New variable.
2564 (speculative_id): New variable.
2565 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2566 call summaries for multiple speculative call targets.
2567 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2568 * ipa-profile.c (struct speculative_call_target): New struct.
2569 (class speculative_call_summary): New class.
2570 (class speculative_call_summaries): New class.
2571 (call_sums): New variable.
2572 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2573 (ipa_profile_write_edge_summary): New function.
2574 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2575 (ipa_profile_dump_all_summaries): New function.
2576 (ipa_profile_read_edge_summary): New function.
2577 (ipa_profile_read_summary_section): New function.
2578 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2579 (ipa_profile): Generate num_speculative_call_targets from
2581 * ipa-ref.h (speculative_id): New variable.
2582 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2583 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2584 common_target_probability. Stream out speculative_id and
2585 num_speculative_call_targets.
2586 (input_edge): Likewise.
2587 * predict.c (dump_prediction): Remove edges count assert to be
2589 * symtab.c (symtab_node::create_reference): Init speculative_id.
2590 (symtab_node::clone_references): Clone speculative_id.
2591 (symtab_node::clone_referring): Clone speculative_id.
2592 (symtab_node::clone_reference): Clone speculative_id.
2593 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2594 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2595 if indirect call contains multiple speculative targets.
2596 * value-prof.h (check_ic_target): Remove.
2597 * value-prof.c (gimple_value_profile_transformations):
2598 Use void function gimple_ic_transform.
2599 * value-prof.c (gimple_ic_transform): Handle topn case.
2600 Fix comment typos. Change it to a void function.
2602 2020-01-13 Andrew Pinski <apinski@marvell.com>
2604 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2605 (octeontx2t98): New define.
2606 (octeontx2t96): New define.
2607 (octeontx2t93): New define.
2608 (octeontx2f95): New define.
2609 (octeontx2f95n): New define.
2610 (octeontx2f95mm): New define.
2611 * config/aarch64/aarch64-tune.md: Regenerate.
2612 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2614 2020-01-13 Jason Merrill <jason@redhat.com>
2616 PR c++/33799 - destroy return value if local cleanup throws.
2617 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2619 2020-01-13 Martin Liska <mliska@suse.cz>
2621 * ipa-cp.c (get_max_overall_size): Use newly
2622 renamed param param_ipa_cp_unit_growth.
2623 * params.opt: Remove legacy param name.
2625 2020-01-13 Martin Sebor <msebor@redhat.com>
2627 PR tree-optimization/93213
2628 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2629 stores to be eliminated.
2631 2020-01-13 Martin Liska <mliska@suse.cz>
2633 * opts.c (print_help): Do not print CL_PARAM
2634 and CL_WARNING for CL_OPTIMIZATION.
2636 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2639 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2640 depending on optimization settings.
2642 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2644 PR tree-optimization/90838
2645 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2646 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2647 argument rather than to initialize temporary for targets that
2648 don't use the mode argument at all. Initialize ctzval to avoid
2651 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2653 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2654 * tree-core.h: Document it.
2655 * gimplify.c (gimplify_omp_workshare): Set it.
2656 * omp-low.c (lower_omp_target): Use it.
2657 * tree-pretty-print.c (dump_omp_clause): Print it.
2659 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2660 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2662 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2664 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2665 * common.opt (fdiagnostics-path-format=): New option.
2666 (diagnostic_path_format): New enum.
2667 (fdiagnostics-show-path-depths): New option.
2668 * coretypes.h (diagnostic_event_id_t): New forward decl.
2669 * diagnostic-color.c (color_dict): Add "path".
2670 * diagnostic-event-id.h: New file.
2671 * diagnostic-format-json.cc (json_from_expanded_location): Make
2673 (json_end_diagnostic): Call context->make_json_for_path if it
2674 exists and the diagnostic has a path.
2675 (diagnostic_output_format_init): Clear context->print_path.
2676 * diagnostic-path.h: New file.
2677 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2678 when printing a run of events in a diagnostic_path so that they
2679 all get the same color.
2680 (layout::m_diagnostic_path_p): New field.
2681 (layout::layout): Initialize it.
2682 (layout::print_any_labels): Don't colorize the label text for an
2683 event in a diagnostic_path.
2684 (gcc_rich_location::add_location_if_nearby): Add
2685 "restrict_to_current_line_spans" and "label" params. Pass the
2686 former to layout.maybe_add_location_range; pass the latter
2687 when calling add_range.
2688 * diagnostic.c: Include "diagnostic-path.h".
2689 (diagnostic_initialize): Initialize context->path_format and
2690 context->show_path_depths.
2691 (diagnostic_show_any_path): New function.
2692 (diagnostic_path::interprocedural_p): New function.
2693 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2694 (simple_diagnostic_path::num_events): New function.
2695 (simple_diagnostic_path::get_event): New function.
2696 (simple_diagnostic_path::add_event): New function.
2697 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2698 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2699 (debug): New overload taking a diagnostic_path *.
2700 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2701 * diagnostic.h (enum diagnostic_path_format): New enum.
2702 (json::value): New forward decl.
2703 (diagnostic_context::path_format): New field.
2704 (diagnostic_context::show_path_depths): New field.
2705 (diagnostic_context::print_path): New callback field.
2706 (diagnostic_context::make_json_for_path): New callback field.
2707 (diagnostic_show_any_path): New decl.
2708 (json_from_expanded_location): New decl.
2709 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2710 (-fdiagnostics-show-path-depths): New option.
2711 (-fdiagnostics-color): Add "path" to description of default
2712 GCC_COLORS; describe it.
2713 (-fdiagnostics-format=json): Document how diagnostic paths are
2714 represented in the JSON output format.
2715 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2716 Add optional params "restrict_to_current_line_spans" and "label".
2717 * opts.c (common_handle_option): Handle
2718 OPT_fdiagnostics_path_format_ and
2719 OPT_fdiagnostics_show_path_depths.
2720 * pretty-print.c: Include "diagnostic-event-id.h".
2721 (pp_format): Implement "%@" format code for printing
2722 diagnostic_event_id_t *.
2723 (selftest::test_pp_format): Add tests for "%@".
2724 * selftest-run-tests.c (selftest::run_tests): Call
2725 selftest::tree_diagnostic_path_cc_tests.
2726 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2727 * toplev.c (general_init): Initialize global_dc->path_format and
2728 global_dc->show_path_depths.
2729 * tree-diagnostic-path.cc: New file.
2730 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2731 non-static. Drop "diagnostic" param in favor of storing the
2732 original value of "where" and re-using it.
2733 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2734 maybe_unwind_expanded_macro_loc.
2735 (tree_diagnostics_defaults): Initialize context->print_path and
2736 context->make_json_for_path.
2737 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2739 (default_tree_make_json_for_path): New decl.
2740 (maybe_unwind_expanded_macro_loc): New decl.
2742 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2744 PR tree-optimization/93210
2745 * fold-const.h (native_encode_initializer,
2746 can_native_interpret_type_p): Declare.
2747 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2749 (native_encode_initializer): New function, moved from dwarf2out.c.
2750 Adjust to native_encode_expr compatible arguments, including dry-run
2751 and partial extraction modes. Don't handle STRING_CST.
2752 (can_native_interpret_type_p): No longer static.
2753 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2754 offset / BITS_PER_UNIT fits into int and don't call it if
2755 can_native_interpret_type_p fails. If suboff is NULL and for
2756 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2757 native_encode_initializer.
2758 (fold_const_aggregate_ref_1): Formatting fix.
2759 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2760 (tree_add_const_value_attribute): Adjust caller.
2762 PR tree-optimization/90838
2763 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2764 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2765 CTZ_DEFINED_VALUE_AT_ZERO.
2767 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2770 * lra-constraints.c (match_reload): Permit input operands have the
2771 same mode as output while other input operands have a different
2774 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
2776 PR tree-optimization/90838
2777 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2778 (check_ctz_string): Likewise.
2779 (optimize_count_trailing_zeroes): Likewise.
2780 (simplify_count_trailing_zeroes): Likewise.
2781 (pass_forwprop::execute): Try ctz simplification.
2782 * match.pd: Add matching for ctz idioms.
2784 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2786 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2788 (aarch64_invalid_unary_op): New function for target hook.
2789 (aarch64_invalid_binary_op): New function for target hook.
2791 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2793 * config.gcc: Add arm_bf16.h.
2794 * config/aarch64/aarch64-builtins.c
2795 (aarch64_simd_builtin_std_type): Add BFmode.
2796 (aarch64_init_simd_builtin_types): Define element types for vector
2798 (aarch64_init_bf16_types): New function.
2799 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2800 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2802 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2803 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2805 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2806 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2807 * config/aarch64/aarch64.c
2808 (aarch64_classify_vector_mode): Add support for BF types.
2809 (aarch64_gimplify_va_arg_expr): Add support for BF types.
2810 (aarch64_vq_mode): Add support for BF types.
2811 (aarch64_simd_container_mode): Add support for BF types.
2812 (aarch64_mangle_type): Add support for BF scalar type.
2813 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2814 * config/aarch64/arm_bf16.h: New file.
2815 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2816 * config/aarch64/iterators.md: Add BF types to mode attributes.
2817 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2819 2020-01-10 Jason Merrill <jason@redhat.com>
2821 PR c++/93173 - incorrect tree sharing.
2822 * gimplify.c (copy_if_shared): No longer static.
2823 * gimplify.h: Declare it.
2825 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2827 * doc/invoke.texi (-msve-vector-bits=): Document that
2828 -msve-vector-bits=128 now generates VL-specific code for
2829 little-endian targets.
2830 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2831 build_vector_type_for_mode to construct the data vector types.
2832 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2833 VL-specific code for -msve-vector-bits=128 on little-endian targets.
2834 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2835 for 128-bit vectors.
2837 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2839 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2842 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2844 * config/aarch64/aarch64-builtins.c
2845 (aarch64_builtin_vectorized_function): Check for specific vector modes,
2846 rather than checking the number of elements and the element mode.
2848 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2850 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2851 get_related_vectype_for_scalar_type rather than build_vector_type
2852 to create the index type for a conditional reduction.
2854 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2856 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
2857 for any type of gather or scatter, including strided accesses.
2859 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2861 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
2864 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2866 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
2868 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
2869 parameter and its use to reset DR_OFFSET's.
2870 (vect_transform_loop): Remove orig_drs_init argument.
2871 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
2872 member of dr_vec_info rather than the offset of the associated
2873 data_reference's innermost_loop_behavior.
2874 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
2875 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
2876 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
2877 get_dr_vinfo_offset.
2878 (vectorizable_store): Likewise.
2879 (vectorizable_load): Likewise.
2881 2020-01-10 Richard Biener <rguenther@suse.de>
2883 * gimple-ssa-store-merging
2884 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
2886 2020-01-10 Martin Liska <mliska@suse.cz>
2889 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
2890 encapsulation that was there before r280040.
2892 2020-01-10 Richard Biener <rguenther@suse.de>
2895 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
2896 sequences to avoid walking them again for secondary opportunities.
2897 (pass_lower_eh_dispatch::execute): Instead actually insert
2900 2020-01-10 Richard Biener <rguenther@suse.de>
2903 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
2904 (cleanup_all_empty_eh): Walk landing pads in reverse order to
2905 avoid quadraticness.
2907 2020-01-10 Martin Jambor <mjambor@suse.cz>
2909 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
2910 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
2911 to get param_ipa_sra_max_replacements.
2912 (param_splitting_across_edge): Pass the caller to
2913 pull_accesses_from_callee.
2915 2020-01-10 Martin Jambor <mjambor@suse.cz>
2917 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
2918 * ipa-cp.c (max_new_size): Removed.
2919 (orig_overall_size): New variable.
2920 (get_max_overall_size): New function.
2921 (estimate_local_effects): Use it. Adjust dump.
2922 (decide_about_value): Likewise.
2923 (ipcp_propagate_stage): Do not calculate max_new_size, just store
2924 orig_overall_size. Adjust dump.
2925 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
2927 2020-01-10 Martin Jambor <mjambor@suse.cz>
2929 * params.opt (param_ipa_max_agg_items): Mark as Optimization
2930 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
2931 instead of param_ipa_max_agg_items.
2932 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
2933 optimization info for the callee.
2935 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
2937 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
2938 markers if debug_inline_points is false.
2940 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2942 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
2944 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
2945 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
2946 aarch64-sve-builtins-sve2.h.
2947 (aarch64-sve-builtins-sve2.o): New rule.
2948 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
2949 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
2950 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
2951 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
2952 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
2953 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
2955 * config/aarch64/aarch64-sve.md: Update comments with SVE2
2956 instructions that are handled here.
2957 (@cond_asrd<mode>): Generalize to...
2958 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
2959 (*cond_asrd<mode>_2): Generalize to...
2960 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
2961 (*cond_asrd<mode>_z): Generalize to...
2962 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
2963 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
2964 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
2965 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
2966 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
2968 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2969 (@aarch64_scatter_stnt<mode>): Likewise.
2970 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2971 (@aarch64_mul_lane_<mode>): Likewise.
2972 (@aarch64_sve_suqadd<mode>_const): Likewise.
2973 (*<sur>h<addsub><mode>): Generalize to...
2974 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
2976 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
2977 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
2978 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
2979 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
2980 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
2981 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
2982 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
2983 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
2984 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
2985 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
2986 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
2987 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
2988 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
2989 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
2990 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
2991 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
2992 (@aarch64_sve2_xar<mode>): Likewise.
2993 (@aarch64_sve2_bcax<mode>): Likewise.
2994 (*aarch64_sve2_eor3<mode>): Rename to...
2995 (@aarch64_sve2_eor3<mode>): ...this.
2996 (@aarch64_sve2_bsl<mode>): New expander.
2997 (@aarch64_sve2_nbsl<mode>): Likewise.
2998 (@aarch64_sve2_bsl1n<mode>): Likewise.
2999 (@aarch64_sve2_bsl2n<mode>): Likewise.
3000 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3001 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3002 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3003 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3004 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3005 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3006 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3007 (<su>mull<bt><Vwide>): Generalize to...
3008 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3010 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3011 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3012 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3013 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3014 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3015 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3016 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3017 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3018 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3019 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3020 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3021 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3022 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3023 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3024 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3025 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3026 (<SHRNB:r>shrnb<mode>): Generalize to...
3027 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3029 (<SHRNT:r>shrnt<mode>): Generalize to...
3030 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3032 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3033 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3034 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3035 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3036 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3037 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3038 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3039 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3040 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3041 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3042 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3043 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3044 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3045 (@aarch64_sve2_cvtnt<mode>): Likewise.
3046 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3047 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3048 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3049 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3050 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3051 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3052 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3053 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3054 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3055 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3056 (@aarch64_sve2_pmul<mode>): Likewise.
3057 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3058 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3059 (@aarch64_sve2_tbl2<mode>): Likewise.
3060 (@aarch64_sve2_tbx<mode>): Likewise.
3061 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3062 (@aarch64_sve2_histcnt<mode>): Likewise.
3063 (@aarch64_sve2_histseg<mode>): Likewise.
3064 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3065 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3066 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3067 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3068 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3069 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3070 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3071 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3072 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3073 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3074 (SVE2_PMULL_PAIR_I): New mode iterators.
3075 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3076 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3077 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3078 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3079 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3080 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3081 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3082 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3083 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3084 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3085 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3086 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3087 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3088 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3089 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3090 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3091 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3092 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3093 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3094 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3095 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3096 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3097 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3098 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3099 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3100 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3101 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3102 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3103 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3104 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3105 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3107 (VNARROW, Ventype): New mode attributes.
3108 (Vewtype): Handle VNx2DI. Fix typo in comment.
3109 (VDOUBLE): New mode attribute.
3110 (sve_lane_con): Handle VNx8HI.
3111 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3112 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3113 (sve_int_op, sve_int_op_rev): Handle the above codes.
3114 (sve_pred_int_rhs2_operand): Likewise.
3115 (MULLBT, SHRNB, SHRNT): Delete.
3116 (SVE_INT_SHIFT_IMM): New int iterator.
3117 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3118 and UNSPEC_WHILEHS for TARGET_SVE2.
3119 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3120 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3121 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3122 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3123 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3124 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3125 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3126 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3127 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3128 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3129 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3130 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3131 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3132 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3133 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3134 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3135 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3136 (optab): Handle the new unspecs.
3137 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3139 (lr): Handle the new unspecs.
3141 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3142 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3143 (sve_int_qsub_op): New int attributes.
3144 (sve_fp_op, rot): Handle the new unspecs.
3145 * config/aarch64/aarch64-sve-builtins.h
3146 (function_resolver::require_matching_pointer_type): Declare.
3147 (function_resolver::resolve_unary): Add an optional boolean argument.
3148 (function_resolver::finish_opt_n_resolution): Add an optional
3149 type_suffix_index argument.
3150 (gimple_folder::redirect_call): Declare.
3151 (gimple_expander::prepare_gather_address_operands): Add an optional
3153 * config/aarch64/aarch64-sve-builtins.cc: Include
3154 aarch64-sve-builtins-sve2.h.
3155 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3156 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3157 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3158 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3159 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3160 (TYPES_s_unsigned): Likewise.
3161 (TYPES_s_integer): Use TYPES_s_unsigned.
3162 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3163 (TYPES_sd_integer): Use them.
3164 (TYPES_d_unsigned): New macro.
3165 (TYPES_d_integer): Use it.
3166 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3167 (TYPES_cvt_narrow): Likewise.
3168 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3169 (preds_mx): New variable.
3170 (function_builder::add_overloaded_function): Allow the new feature
3171 set to be more restrictive than the original one.
3172 (function_resolver::infer_pointer_type): Remove qualifiers from
3173 the pointer type before printing it.
3174 (function_resolver::require_matching_pointer_type): New function.
3175 (function_resolver::resolve_sv_displacement): Handle functions
3176 that don't support 32-bit vector indices or svint32_t vector offsets.
3177 (function_resolver::finish_opt_n_resolution): Take the inferred type
3178 as a separate argument.
3179 (function_resolver::resolve_unary): Optionally treat all forms in
3180 the same way as normal merging functions.
3181 (gimple_folder::redirect_call): New function.
3182 (function_expander::prepare_gather_address_operands): Add an argument
3183 that says whether scaled forms are available. If they aren't,
3184 handle scaling of vector indices and don't add the extension and
3186 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3187 fall back to using cond_* instead.
3188 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3189 Split out the member variables into...
3190 (rtx_code_function_base): ...this new base class.
3191 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3192 (unspec_based_function): Split out the member variables into...
3193 (unspec_based_function_base): ...this new base class.
3194 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3195 (unspec_based_function_exact_insn): New class.
3196 (unspec_based_add_function, unspec_based_add_lane_function)
3197 (unspec_based_lane_function, unspec_based_pred_function)
3198 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3199 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3200 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3202 (unspec_based_fused_function): New class.
3203 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3204 (unspec_based_fused_lane_function): New class.
3205 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3207 (CODE_FOR_MODE1): New macro.
3208 (fixed_insn_function): New class.
3209 (while_comparison): Likewise.
3210 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3211 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3212 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3213 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3214 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3215 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3216 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3217 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3218 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3219 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3220 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3221 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3222 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3223 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3224 Also add an initial argument for unary_convert_narrowt, regardless
3225 of the predication type.
3226 (build_32_64): Allow loads and stores to specify MODE_none.
3227 (build_sv_index64, build_sv_uint_offset): New functions.
3228 (long_type_suffix): New function.
3229 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3230 (binary_imm_long_base, load_gather_sv_base): Likewise.
3231 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3232 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3233 (unary_narrowb_base, unary_narrowt_base): Likewise.
3234 (binary_long_lane_def, binary_long_lane): New shape.
3235 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3236 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3237 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3238 (binary_to_uint_def, binary_to_uint): Likewise.
3239 (binary_wide_def, binary_wide): Likewise.
3240 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3241 (compare_def, compare): Likewise.
3242 (compare_ptr_def, compare_ptr): Likewise.
3243 (load_ext_gather_index_restricted_def,
3244 load_ext_gather_index_restricted): Likewise.
3245 (load_ext_gather_offset_restricted_def,
3246 load_ext_gather_offset_restricted): Likewise.
3247 (load_gather_sv_def): Inherit from load_gather_sv_base.
3248 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3249 (shift_left_imm_def, shift_left_imm): Likewise.
3250 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3251 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3252 (store_scatter_index_restricted_def,
3253 store_scatter_index_restricted): Likewise.
3254 (store_scatter_offset_restricted_def,
3255 store_scatter_offset_restricted): Likewise.
3256 (tbl_tuple_def, tbl_tuple): Likewise.
3257 (ternary_long_lane_def, ternary_long_lane): Likewise.
3258 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3259 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3260 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3261 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3262 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3263 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3264 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3265 (ternary_uint_def, ternary_uint): Likewise.
3266 (unary_convert): Fix typo in comment.
3267 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3268 (unary_long_def, unary_long): Likewise.
3269 (unary_narrowb_def, unary_narrowb): Likewise.
3270 (unary_narrowt_def, unary_narrowt): Likewise.
3271 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3272 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3273 (unary_to_int_def, unary_to_int): Likewise.
3274 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3275 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3276 (svasrd_impl): Delete.
3277 (svcadd_impl::expand): Handle integer operations too.
3278 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3279 new functions to derive the unspec numbers.
3280 (svmla_svmls_lane_impl): Replace with...
3281 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3282 integer operations too.
3283 (svwhile_impl): Rename to...
3284 (svwhilelx_impl): ...this and inherit from while_comparison.
3285 (svasrd): Use unspec_based_function.
3286 (svmla_lane): Use svmla_lane_impl.
3287 (svmls_lane): Use svmls_lane_impl.
3288 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3289 (svwhilele, svwhilelt): Use svwhilelx_impl.
3290 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3291 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3292 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3293 * config/aarch64/aarch64-sve-builtins.def: Include
3294 aarch64-sve-builtins-sve2.def.
3296 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3298 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3299 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3300 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3301 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3302 immediates as well as vector ones.
3303 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3304 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3305 (aarch64_sve_qsub_immediate): Update calls accordingly.
3307 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3309 * config/aarch64/aarch64-sve2.md: Add banner comments.
3310 (<su>mulh<r>s<mode>3): Move further up file.
3311 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3312 (*aarch64_sve2_sra<mode>): Move further down file.
3313 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3315 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3317 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3319 (while_optab_cmp): Handle them.
3320 * config/aarch64/aarch64-sve.md
3321 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3322 and add a "@" marker.
3323 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3324 instead of gen_aarch64_sve2_while_ptest.
3325 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3327 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3329 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3330 (UNSPEC_WHILELE): ...this.
3331 (UNSPEC_WHILE_LO): Rename to...
3332 (UNSPEC_WHILELO): ...this.
3333 (UNSPEC_WHILE_LS): Rename to...
3334 (UNSPEC_WHILELS): ...this.
3335 (UNSPEC_WHILE_LT): Rename to...
3336 (UNSPEC_WHILELT): ...this.
3337 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3338 (cmp_op, while_optab_cmp): Likewise.
3339 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3340 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3341 (svwhilelt): Likewise.
3343 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3345 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3346 (unary_to_uint): Define.
3347 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3348 (unary_count): Rename to...
3349 (unary_to_uint_def, unary_to_uint): ...this.
3350 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3352 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3354 * config/aarch64/aarch64-sve-builtins-functions.h
3355 (code_for_mode_function): New class.
3356 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3357 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3358 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3359 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3360 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3362 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3364 * config/aarch64/iterators.md (addsub): New code attribute.
3365 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3367 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3368 in the asm string and attributes. Fix indentation.
3369 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3371 (@aarch64_sve_<optab><mode>): ...this.
3372 * config/aarch64/aarch64-sve-builtins.h
3373 (function_expander::expand_signed_unpred_op): Delete.
3374 * config/aarch64/aarch64-sve-builtins.cc
3375 (function_expander::expand_signed_unpred_op): Likewise.
3376 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3377 try using code_for_aarch64_sve instead.
3378 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3379 (svqsub_impl): Likewise.
3380 (svqadd, svqsub): Use rtx_code_function instead.
3382 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3384 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3385 (HADDSUB, sur, addsub): Remove them.
3387 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3389 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3390 functions the same way, rather than singling out those that
3391 aren't mapped directly to optabs.
3393 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3395 * target.def (compatible_vector_types_p): New target hook.
3396 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3397 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3398 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3399 * doc/tm.texi: Regenerate.
3400 * gimple-expr.c: Include target.h.
3401 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3402 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3404 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3405 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3406 Use the original predicate if it already has a suitable type.
3408 2020-01-09 Martin Jambor <mjambor@suse.cz>
3410 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3411 resolve_speculation and redirect_call_stmt_to_callee static. Change
3412 return type of set_call_stmt to cgraph_edge *.
3413 * auto-profile.c (afdo_indirect_call): Adjust call to
3414 redirect_call_stmt_to_callee.
3415 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3416 make the this pointer explicit, adjust self-recursive calls and the
3417 call top make_direct. Return the resulting edge.
3418 (cgraph_edge::remove): Make this pointer explicit.
3419 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3420 (cgraph_edge::make_direct): Likewise, adjust call to
3421 resolve_speculation.
3422 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3423 call to set_call_stmt.
3424 (cgraph_update_edges_for_call_stmt_node): Update call to
3425 set_call_stmt and remove.
3426 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3427 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3428 (cgraph_node::create_edge_including_clones): Moved "first" definition
3429 of edge to the block where it was used. Adjusted calls to
3431 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3432 cgraph_edge::remove.
3433 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3434 make_direct and redirect_call_stmt_to_callee.
3435 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3436 resolve_speculation and make_direct.
3437 * ipa-inline-transform.c (inline_transform): Adjust call to
3438 redirect_call_stmt_to_callee.
3439 (check_speculations_1):: Adjust call to resolve_speculation.
3440 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3441 resolve-speculation.
3442 (inline_small_functions): Adjust call to resolve_speculation.
3443 (ipa_inline): Likewise.
3444 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3446 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3447 safe with regards to edge removal, adjust calls to
3448 redirect_call_stmt_to_callee.
3449 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3450 and redirect_call_stmt_to_callee.
3451 * multiple_target.c (create_dispatcher_calls): Adjust call to
3452 redirect_call_stmt_to_callee
3453 (redirect_to_specific_clone): Likewise.
3454 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3455 Adjust calls to cgraph_edge::remove.
3456 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3457 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3458 (expand_call_inline): Adjust call to cgraph_edge::remove.
3460 2020-01-09 Martin Liska <mliska@suse.cz>
3462 * params.opt: Set Optimization for
3463 param_max_speculative_devirt_maydefs.
3465 2020-01-09 Martin Sebor <msebor@redhat.com>
3469 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3471 2020-01-09 Martin Liska <mliska@suse.cz>
3473 * auto-profile.c (auto_profile): Use opt_for_fn
3475 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3476 (propagate_vals_across_arith_jfunc): Likewise.
3477 (hint_time_bonus): Likewise.
3478 (incorporate_penalties): Likewise.
3479 (good_cloning_opportunity_p): Likewise.
3480 (perform_estimation_of_a_value): Likewise.
3481 (estimate_local_effects): Likewise.
3482 (ipcp_propagate_stage): Likewise.
3483 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3484 (set_switch_stmt_execution_predicate): Likewise.
3485 (analyze_function_body): Likewise.
3486 * ipa-inline-analysis.c (offline_size): Likewise.
3487 * ipa-inline.c (early_inliner): Likewise.
3488 * ipa-prop.c (ipa_analyze_node): Likewise.
3489 (ipcp_transform_function): Likewise.
3490 * ipa-sra.c (process_scan_results): Likewise.
3491 (ipa_sra_summarize_function): Likewise.
3492 * params.opt: Rename ipcp-unit-growth to
3493 ipa-cp-unit-growth. Add Optimization for various
3494 IPA-related parameters.
3496 2020-01-09 Richard Biener <rguenther@suse.de>
3499 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3501 2020-01-09 Richard Biener <rguenther@suse.de>
3503 PR tree-optimization/93040
3504 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3506 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3508 * common/config/avr/avr-common.c (avr_option_optimization_table)
3509 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3511 2020-01-09 Martin Liska <mliska@suse.cz>
3513 * cgraphclones.c (symbol_table::materialize_all_clones):
3514 Use cgraph_node::dump_name.
3516 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3519 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3520 output_operand_lossage instead of gcc_unreachable.
3521 * doc/md.texi (riscv f constraint): Fix typo.
3524 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3525 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3526 CONST_SCALAR_INT_P instead of CONST_INT_P.
3527 (*subv<mode>4_1): Rename to ...
3528 (subv<mode>4_1): ... this.
3529 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3530 define_insn_and_split patterns.
3531 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3534 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3536 * vec.c (class selftest::count_dtor): New class.
3537 (selftest::test_auto_delete_vec): New test.
3538 (selftest::vec_c_tests): Call it.
3539 * vec.h (class auto_delete_vec): New class template.
3540 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3542 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3544 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3546 2020-01-08 Jim Wilson <jimw@sifive.com>
3548 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3549 use of TLS_MODEL_LOCAL_EXEC when not pic.
3551 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3553 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3556 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3559 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3560 *stack_protect_set_3 peephole2): Also check that the second
3561 insns source is general_operand.
3564 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3565 predicate for output operand instead of register_operand.
3566 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3567 memory destination and non-memory operands[2].
3569 2020-01-08 Martin Liska <mliska@suse.cz>
3571 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3572 ::dump_asm_name instead of (::name or ::asm_name).
3573 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3574 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3575 (analyze_functions): Likewise.
3576 (expand_all_functions): Likewise.
3577 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3578 (propagate_bits_across_jump_function): Likewise.
3579 (dump_profile_updates): Likewise.
3580 (ipcp_store_bits_results): Likewise.
3581 (ipcp_store_vr_results): Likewise.
3582 * ipa-devirt.c (dump_targets): Likewise.
3583 * ipa-fnsummary.c (analyze_function_body): Likewise.
3584 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3585 (process_hsa_functions): Likewise.
3586 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3587 (set_alias_uids): Likewise.
3588 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3589 * ipa-inline.c (recursive_inlining): Likewise.
3590 (inline_to_all_callers_1): Likewise.
3591 (ipa_inline): Likewise.
3592 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3593 (ipa_propagate_frequency): Likewise.
3594 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3595 (remove_described_reference): Likewise.
3596 * ipa-pure-const.c (worse_state): Likewise.
3597 (check_retval_uses): Likewise.
3598 (analyze_function): Likewise.
3599 (propagate_pure_const): Likewise.
3600 (propagate_nothrow): Likewise.
3601 (dump_malloc_lattice): Likewise.
3602 (propagate_malloc): Likewise.
3603 (pass_local_pure_const::execute): Likewise.
3604 * ipa-visibility.c (optimize_weakref): Likewise.
3605 (function_and_variable_visibility): Likewise.
3606 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3607 (ipa_discover_variable_flags): Likewise.
3608 * lto-streamer-out.c (output_function): Likewise.
3609 (output_constructor): Likewise.
3610 * tree-inline.c (copy_bb): Likewise.
3611 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3612 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3614 2020-01-08 Richard Biener <rguenther@suse.de>
3617 * tree-eh.c (sink_clobbers): Update virtual operands for
3618 the first and last stmt only. Add a dry-run capability.
3619 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3620 after CFG manipulations and in RPO order to catch all
3621 secondary opportunities reliably.
3623 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3626 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3628 2019-01-08 Richard Biener <rguenther@suse.de>
3631 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3632 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3633 virtual operand, also updating SSA use.
3634 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3635 Update stmt after resetting virtual operand.
3636 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3637 * gimple-iterator.c (gsi_remove): When not removing the stmt
3638 permanently do not delink immediate uses or mark the stmt modified.
3640 2020-01-08 Martin Liska <mliska@suse.cz>
3642 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3643 (ipa_call_context::estimate_size_and_time): Likewise.
3644 (inline_analyze_function): Likewise.
3646 2020-01-08 Martin Liska <mliska@suse.cz>
3648 * cgraph.c (cgraph_node::dump): Use systematically
3651 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3653 Add -nodevicespecs option for avr.
3656 * config/avr/avr.opt (-nodevicespecs): New driver option.
3657 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3658 "-specs=device-specs/..." if that option is not set.
3659 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3661 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3663 Implement 64-bit double functions for avr.
3666 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3667 --with-double-comparison.
3668 * doc/install.texi: Document them.
3669 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3670 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3671 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3672 * doc/invoke.texi (AVR Built-in Macros): Document them.
3673 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3674 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3675 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3677 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3680 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3681 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3682 when only building rm-profile multilibs.
3684 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3687 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3688 lattice for a value to check.
3689 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3690 finite propagation in self-recursive scc.
3692 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3694 * ipa-inline.c (caller_growth_limits): Restore the AND.
3696 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3698 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3699 (VEC_ALLREG_ALT): New iterator.
3700 (VEC_ALLREG_INT_MODE): New iterator.
3701 (VCMP_MODE): New iterator.
3702 (VCMP_MODE_INT): New iterator.
3703 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3704 (vec_cmp<u>v64qidi): New define_expand.
3705 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3706 (vec_cmpu<mode>di_exec): New define_expand.
3707 (vec_cmp<u>v64qidi_exec): New define_expand.
3708 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3709 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3710 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3711 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3712 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3713 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3714 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3715 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3716 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3717 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3719 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3720 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3722 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3724 * config/gcn/constraints.md (DA): Update description and match.
3726 (Db): New constraint.
3727 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3729 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3730 Implement 'Db' mixed immediate type.
3731 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3732 (addcv64si3_dup<exec_vcc>): Delete.
3733 (subcv64si3<exec_vcc>): Rework constraints.
3734 (addv64di3): Rework constraints.
3735 (addv64di3_exec): Rework constraints.
3736 (subv64di3): Rework constraints.
3737 (addv64di3_dup): Delete.
3738 (addv64di3_dup_exec): Delete.
3739 (addv64di3_zext): Rework constraints.
3740 (addv64di3_zext_exec): Rework constraints.
3741 (addv64di3_zext_dup): Rework constraints.
3742 (addv64di3_zext_dup_exec): Rework constraints.
3743 (addv64di3_zext_dup2): Rework constraints.
3744 (addv64di3_zext_dup2_exec): Rework constraints.
3745 (addv64di3_sext_dup2): Rework constraints.
3746 (addv64di3_sext_dup2_exec): Rework constraints.
3748 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3750 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3751 existing target checks.
3753 2020-01-07 Richard Biener <rguenther@suse.de>
3755 * doc/install.texi: Bump minimal supported MPC version.
3757 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3759 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3760 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3761 * langhooks.c: Include stor-layout.h.
3762 (lhd_simulate_enum_decl): New function.
3763 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3764 handle_arm_sve_h for the LTO frontend.
3765 (register_vector_type): Cope with null returns from pushdecl.
3767 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3769 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3770 (aarch64_sve::nvectors_if_data_type): Replace with...
3771 (aarch64_sve::builtin_type_p): ...this.
3772 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3773 (find_vector_type): Delete.
3774 (add_sve_type_attribute): New function.
3775 (lookup_sve_type_attribute): Likewise.
3776 (register_builtin_types): Add an "SVE type" attribute to each type.
3777 (register_tuple_type): Likewise.
3778 (svbool_type_p, nvectors_if_data_type): Delete.
3779 (mangle_builtin_type): Use lookup_sve_type_attribute.
3780 (builtin_type_p): Likewise. Add an overload that returns the
3781 number of constituent vector and predicate registers.
3782 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3783 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3784 instead of aarch64_sve_argument_p.
3785 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3786 (aarch64_pass_by_reference): Likewise.
3787 (aarch64_function_value_1): Likewise.
3788 (aarch64_return_in_memory): Likewise.
3789 (aarch64_layout_arg): Likewise.
3791 2020-01-07 Jakub Jelinek <jakub@redhat.com>
3793 PR tree-optimization/93156
3794 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3795 least significant bit is always clear.
3797 PR tree-optimization/93118
3798 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
3799 simplifier with two intermediate conversions.
3801 2020-01-07 Martin Liska <mliska@suse.cz>
3803 * params.opt: Add Optimization for various parameters.
3805 2020-01-07 Martin Liska <mliska@suse.cz>
3808 * doc/extend.texi: Explain cloning for target_clone
3811 2020-01-07 Martin Liska <mliska@suse.cz>
3813 PR tree-optimization/92860
3814 * common.opt: Make in Optimization option
3815 as it is affected by -O0, which is an Optimization
3817 * tree-inline.c (tree_inlinable_function_p):
3818 Use opt_for_fn for warn_inline.
3819 (expand_call_inline): Likewise.
3821 2020-01-07 Martin Liska <mliska@suse.cz>
3823 PR tree-optimization/92860
3824 * common.opt: Make flag_ree as optimization
3827 2020-01-07 Martin Liska <mliska@suse.cz>
3829 PR optimization/92860
3830 * params.opt: Mark param_min_crossjump_insns with Optimization
3833 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
3835 * ipa-inline-analysis.c (estimate_growth): Fix typo.
3836 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3838 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
3840 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3841 helper function to return the valid addressing formats for a given
3842 hard register and mode.
3843 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3845 * config/rs6000/constraints.md (Q constraint): Update
3847 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3850 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3851 Use 'Q' for doing vector extract from memory.
3852 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
3854 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
3855 doing vector extract from memory.
3856 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
3857 extract from memory.
3859 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
3860 for the offset being 34-bits when -mcpu=future is used.
3862 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
3864 * config/pa/pa.md: Revert change to use ordered_comparison_operator
3865 instead of cmpib_comparison_operator in cmpib patterns.
3866 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
3867 of cmpib_comparison_operator. Revise comment.
3869 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3871 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
3872 in an IFN_DIV_POW2 node to be equal.
3874 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3876 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
3877 (vect_check_scalar_mask): ...this.
3878 (vectorizable_store, vectorizable_load): Update call accordingly.
3879 (vectorizable_call): Use vect_check_scalar_mask to check the mask
3880 argument in calls to conditional internal functions.
3882 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3884 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
3885 '0' matching inputs.
3886 (subv64di3_exec): Likewise.
3888 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
3890 * config/mips/mips.c (vr4130_align_insns): Fix typo.
3891 * doc/md.texi (movstr): Likewise.
3893 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3895 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
3898 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3900 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
3902 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
3903 to a temporary file and use move-if-change to update the real
3904 file where necessary.
3906 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3908 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
3909 rather than Upa for CPY /M.
3911 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3913 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
3916 2020-01-06 Martin Liska <mliska@suse.cz>
3918 PR tree-optimization/92860
3919 * params.opt: Mark param_max_combine_insns with Optimization
3922 2020-01-05 Jakub Jelinek <jakub@redhat.com>
3925 * config/i386/i386.md (SWIDWI): New mode iterator.
3926 (DWI, dwi): Add TImode variants.
3927 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
3928 <general_hilo_operand> instead of <general_operand>. Use
3929 CONST_SCALAR_INT_P instead of CONST_INT_P.
3930 (*addv<mode>4_1): Rename to ...
3931 (addv<mode>4_1): ... this.
3932 (QWI): New mode attribute.
3933 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3934 define_insn_and_split patterns.
3935 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3937 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
3938 <general_hilo_operand> instead of <general_operand>.
3939 (*addcarry<mode>_1): New define_insn.
3940 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
3942 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
3944 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
3945 Use "call" instead of "set".
3947 2020-01-03 Martin Jambor <mjambor@suse.cz>
3950 * ipa-cp.c (print_all_lattices): Skip functions without info.
3952 2020-01-03 Jakub Jelinek <jakub@redhat.com>
3955 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
3956 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
3957 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
3958 for 'e' simd clones.
3961 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
3963 (mprefer-vector-width=): Add Save.
3964 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
3965 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
3966 (ix86_debug_options, ix86_function_specific_print): Adjust
3967 ix86_target_string callers.
3968 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
3969 (ix86_valid_target_attribute_tree): Likewise.
3970 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
3971 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
3972 ix86_target_string caller.
3975 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
3976 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
3977 instead of gen_int_shift_amount + convert_modes.
3979 PR rtl-optimization/93088
3980 * loop-iv.c (find_single_def_src): Punt after looking through
3981 128 reg copies for regs with single definitions. Move definitions
3984 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
3986 * config/arm/arm-c.c (arm_cpu_builtins): Define
3987 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
3988 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
3989 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
3990 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
3991 * config/arm/arm-tables.opt: Regenerated.
3992 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
3993 arm_arch_i8mm and arm_arch_bf16 when enabled.
3994 * config/arm/arm.h (TARGET_I8MM): New macro.
3995 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
3996 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
3997 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
3998 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
3999 (v8_6_a_simd_variants): New.
4000 (v8_*_a_simd_variants): Add i8mm and bf16.
4001 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4003 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4006 * predict.c (compute_function_frequency): Don't call
4007 warn_function_cold on functions that already have cold attribute.
4009 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4012 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4013 COMDAT group function labels in .data.rel.ro.local section.
4014 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4017 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4018 comparison_operator in B and S integer comparisons. Likewise, use
4019 ordered_comparison_operator instead of cmpib_comparison_operator in
4021 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4023 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4025 Update copyright years.
4027 * gcc.c (process_command): Update copyright notice dates.
4028 * gcov-dump.c (print_version): Ditto.
4029 * gcov.c (print_version): Ditto.
4030 * gcov-tool.c (print_version): Ditto.
4031 * gengtype.c (create_file): Ditto.
4032 * doc/cpp.texi: Bump @copying's copyright year.
4033 * doc/cppinternals.texi: Ditto.
4034 * doc/gcc.texi: Ditto.
4035 * doc/gccint.texi: Ditto.
4036 * doc/gcov.texi: Ditto.
4037 * doc/install.texi: Ditto.
4038 * doc/invoke.texi: Ditto.
4040 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4042 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4045 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4047 PR tree-optimization/93098
4048 * match.pd (popcount): For shift amounts, use integer_onep
4049 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4050 tests. Make sure that precision is power of two larger than or equal
4051 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4052 instead of ULL suffixed constants. Formatting fixes.
4054 Copyright (C) 2020 Free Software Foundation, Inc.
4056 Copying and distribution of this file, with or without modification,
4057 are permitted in any medium without royalty provided the copyright
4058 notice and this notice are preserved.