1 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
4 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
5 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
7 2020-02-15 Jason Merrill <jason@redhat.com>
9 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
11 2020-02-15 Jakub Jelinek <jakub@redhat.com>
13 PR tree-optimization/93744
14 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
15 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
16 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
17 sure @2 in the first and @1 in the other patterns has no side-effects.
19 2020-02-15 David Malcolm <dmalcolm@redhat.com>
20 Bernd Edlinger <bernd.edlinger@hotmail.de>
24 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
25 * configure.ac (--with-diagnostics-urls): New configuration
26 option, based on --with-diagnostics-color.
27 (DIAGNOSTICS_URLS_DEFAULT): New define.
28 * config.h: Regenerate.
29 * configure: Regenerate.
30 * diagnostic.c (diagnostic_urls_init): Handle -1 for
31 DIAGNOSTICS_URLS_DEFAULT from configure-time
32 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
33 and TERM_URLS environment variable.
34 * diagnostic-url.h (diagnostic_url_format): New enum type.
35 (diagnostic_urls_enabled_p): rename to...
36 (determine_url_format): ... this, and change return type.
37 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
38 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
39 the linux console, and mingw.
40 (diagnostic_urls_enabled_p): rename to...
41 (determine_url_format): ... this, and adjust.
42 * pretty-print.h (pretty_printer::show_urls): rename to...
43 (pretty_printer::url_format): ... this, and change to enum.
44 * pretty-print.c (pretty_printer::pretty_printer,
45 pp_begin_url, pp_end_url, test_urls): Adjust.
46 * doc/install.texi (--with-diagnostics-urls): Document the new
48 (--with-diagnostics-color): Document the existing interaction
49 with GCC_COLORS better.
50 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
51 vindex reference. Update description of defaults based on the above.
52 (-fdiagnostics-color): Update description of how -fdiagnostics-color
53 interacts with GCC_COLORS.
55 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
58 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
59 conjunction with TARGET_GNU_TLS in early return.
61 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
63 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
64 the mode is not wider than UNITS_PER_WORD.
66 2020-02-14 Martin Jambor <mjambor@suse.cz>
68 PR tree-optimization/93516
69 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
70 access of the same type as the parent.
71 (propagate_subaccesses_from_lhs): Likewise.
73 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
76 * config/i386/avx512vbmi2intrin.h
77 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
78 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
79 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
80 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
81 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
82 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
83 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
84 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
85 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
86 of lacking a closing parenthesis.
87 * config/i386/avx512vbmi2vlintrin.h
88 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
89 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
90 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
91 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
92 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
93 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
94 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
95 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
96 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
97 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
98 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
99 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
100 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
101 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
102 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
103 _mm_shldi_epi32, _mm_mask_shldi_epi32,
104 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
105 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
107 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
110 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
111 the target function entry.
113 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
115 * common/config/arc/arc-common.c (arc_option_optimization_table):
116 Disable if-conversion step when optimized for size.
118 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
120 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
121 R12-R15 are always in ARCOMPACT16_REGS register class.
122 * config/arc/arc.opt (mq-class): Deprecate.
123 * config/arc/constraint.md ("q"): Remove dependency on mq-class
125 * doc/invoke.texi (mq-class): Update text.
126 * common/config/arc/arc-common.c (arc_option_optimization_table):
129 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
131 * config/arc/arc.c (arc_insn_cost): New function.
132 (TARGET_INSN_COST): Define.
133 * config/arc/arc.md (cost): New attribute.
134 (add_n): Use arc_nonmemory_operand.
135 (ashlsi3_insn): Likewise, also update constraints.
136 (ashrsi3_insn): Likewise.
138 (add_shift): Likewise.
139 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
141 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
143 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
145 (umulsidi_600): Likewise.
147 2020-02-13 Jakub Jelinek <jakub@redhat.com>
150 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
151 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
152 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
153 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
154 pass __A to the builtin followed by __W instead of __A followed by
156 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
157 _mm512_mask_popcnt_epi64): Likewise.
158 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
159 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
160 _mm256_mask_popcnt_epi64): Likewise.
162 PR tree-optimization/93582
163 * fold-const.h (shift_bytes_in_array_left,
164 shift_bytes_in_array_right): Declare.
165 * fold-const.c (shift_bytes_in_array_left,
166 shift_bytes_in_array_right): New function, moved from
167 gimple-ssa-store-merging.c, no longer static.
168 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
169 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
170 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
171 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
172 shift_bytes_in_array.
173 (verify_shift_bytes_in_array): Rename to ...
174 (verify_shift_bytes_in_array_left): ... this. Use
175 shift_bytes_in_array_left instead of shift_bytes_in_array.
176 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
177 instead of verify_shift_bytes_in_array.
178 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
179 / native_interpret_expr where the store covers all needed bits,
180 punt on PDP-endian, otherwise allow all involved offsets and sizes
181 not to be byte-aligned.
184 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
185 use const_0_to_255_operand predicate instead of immediate_operand.
186 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
187 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
188 vgf2p8affineinvqb_<mode><mask_name>,
189 vgf2p8affineqb_<mode><mask_name>): Drop mode from
190 const_0_to_255_operand predicated operands.
192 2020-02-12 Jeff Law <law@redhat.com>
194 * config/h8300/h8300.md (comparison shortening peepholes): Use
195 a mode iterator to merge the HImode and SImode peepholes.
197 2020-02-12 Jakub Jelinek <jakub@redhat.com>
200 * real.c (is_even): Make static. Function comment fix.
201 (is_halfway_below): Make static, don't assert R is not inf/nan,
202 instead return false for those. Small formatting fixes.
204 2020-02-12 Martin Sebor <msebor@redhat.com>
207 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
208 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
209 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
210 (strlen_check_and_optimize_call): Adjust callee name.
212 2020-02-12 Jeff Law <law@redhat.com>
214 * config/h8300/h8300.md (comparison shortening peepholes): Drop
215 (and (xor)) variant. Combine other two into single peephole.
217 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
219 PR rtl-optimization/93565
220 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
222 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
224 * config/aarch64/aarch64-simd.md
225 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
226 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
227 generating separate ADDV and zero_extend patterns.
228 * config/aarch64/iterators.md (VDQV_E): New iterator.
230 2020-02-12 Jeff Law <law@redhat.com>
232 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
233 expanders, splits, etc.
234 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
235 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
236 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
237 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
238 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
239 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
241 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
243 2020-02-12 Jakub Jelinek <jakub@redhat.com>
246 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
247 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
248 TARGET_AVX512DQ from condition.
249 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
250 instead of <mask_mode512bit_condition> in condition. If
251 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
253 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
256 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
259 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
261 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
263 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
264 where strlen is more legible.
265 (rs6000_builtin_vectorized_libmass): Ditto.
266 (rs6000_print_options_internal): Ditto.
268 2020-02-11 Martin Sebor <msebor@redhat.com>
270 PR tree-optimization/93683
271 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
273 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
275 * config/rs6000/predicates.md (cint34_operand): Rename the
276 -mprefixed-addr option to be -mprefixed.
277 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
278 the -mprefixed-addr option to be -mprefixed.
279 (OTHER_FUTURE_MASKS): Likewise.
280 (POWERPC_MASKS): Likewise.
281 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
282 the -mprefixed-addr option to be -mprefixed. Change error
283 messages to refer to -mprefixed.
284 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
286 (rs6000_legitimate_offset_address_p): Likewise.
287 (rs6000_mode_dependent_address): Likewise.
288 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
289 "-mprefixed" for target attributes and pragmas.
290 (address_to_insn_form): Rename the -mprefixed-addr option to be
292 (rs6000_adjust_insn_length): Likewise.
293 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
294 -mprefixed-addr option to be -mprefixed.
295 (ASM_OUTPUT_OPCODE): Likewise.
296 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
297 -mprefixed-addr option to be -mprefixed.
298 * config/rs6000/rs6000.opt (-mprefixed): Rename the
299 -mprefixed-addr option to be prefixed. Change the option from
300 being undocumented to being documented.
301 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
302 -mprefixed option. Update the -mpcrel documentation to mention
305 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
307 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
308 including FIRST_PSEUDO_REGISTER - 1.
309 * ira-color.c (print_hard_reg_set): Ditto.
311 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
313 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
314 (USTERNOP_QUALIFIERS): New define.
315 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
316 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
317 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
318 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
319 * config/arm/arm_neon.h (vusdot_s32): New.
320 (vusdot_lane_s32): New.
321 (vusdotq_lane_s32): New.
322 (vsudot_lane_s32): New.
323 (vsudotq_lane_s32): New.
324 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
325 * config/arm/iterators.md (DOTPROD_I8MM): New.
326 (sup, opsuffix): Add <us/su>.
327 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
328 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
330 2020-02-11 Richard Biener <rguenther@suse.de>
332 PR tree-optimization/93661
333 PR tree-optimization/93662
334 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
336 * tree-sra.c (get_access_for_expr): Likewise.
338 2020-02-10 Jakub Jelinek <jakub@redhat.com>
341 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
342 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
343 Change condition from TARGET_AVX2 to TARGET_AVX.
345 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
348 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
351 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
353 Try to generate zero-based comparisons.
354 * config/cris/cris.c (cris_reduce_compare): New function.
355 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
356 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
357 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
359 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
362 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
363 in Thumb state and also as a destination in Arm state. Add T16
366 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
368 * md.texi (Define Subst): Match closing paren in example.
370 2020-02-10 Jakub Jelinek <jakub@redhat.com>
374 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
375 arguments of strncmp.
377 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
380 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
381 but different source value.
382 (adjust_callers_for_value_intersection): New function.
383 (gather_edges_for_value): Adjust order of callers to let a
384 non-self-recursive caller be the first element.
385 (self_recursive_pass_through_p): Add a new parameter "simple", and
386 check generalized self-recursive pass-through jump function.
387 (self_recursive_agg_pass_through_p): Likewise.
388 (find_more_scalar_values_for_callers_subset): Compute value from
389 pass-through jump function for self-recursive.
390 (intersect_with_plats): Cleanup previous implementation code for value
391 itersection with self-recursive call edge.
392 (intersect_with_agg_replacements): Likewise.
393 (intersect_aggregates_with_edge): Deduce value from pass-through jump
394 function for self-recursive call edge. Cleanup previous implementation
395 code for value intersection with self-recursive call edge.
396 (decide_whether_version_node): Remove dead callers and adjust order
397 to let a non-self-recursive caller be the first element.
399 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
401 * recog.c: Move pass_split_before_sched2 code in front of
402 pass_split_before_regstack.
403 (pass_data_split_before_sched2): Rename pass to split3 from split4.
404 (pass_data_split_before_regstack): Rename pass to split4 from split3.
405 (rest_of_handle_split_before_sched2): Remove.
406 (pass_split_before_sched2::execute): Unconditionally call
408 (enable_split_before_sched2): New function.
409 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
410 (pass_split_before_regstack::gate): Ditto.
411 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
412 Update name check for renamed split4 pass.
413 * config/sh/sh.c (register_sh_passes): Update pass insertion
414 point for renamed split4 pass.
416 2020-02-09 Jakub Jelinek <jakub@redhat.com>
418 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
419 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
420 copying them around between host and target.
422 2020-02-08 Andrew Pinski <apinski@marvell.com>
425 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
426 STRICT_ALIGNMENT also.
428 2020-02-08 Jim Wilson <jimw@sifive.com>
431 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
433 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
434 Jakub Jelinek <jakub@redhat.com>
437 * config/i386/i386.h (CALL_USED_REGISTERS): Make
438 xmm16-xmm31 call-used even in 64-bit ms-abi.
440 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
442 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
443 (simd_ummla, simd_usmmla): Likewise.
444 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
445 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
448 2020-02-07 Richard Biener <rguenther@suse.de>
451 * tree-inline.c (fold_marked_statements): Do a PRE walk,
452 skipping unreachable regions.
453 (optimize_inline_calls): Skip folding stmts when we didn't
456 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
459 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
460 Don't return aggregates with only SFmode and DFmode in SSE
462 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
464 2020-02-07 Jakub Jelinek <jakub@redhat.com>
467 * config/rs6000/rs6000-logue.c
468 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
469 if it fails, move rs into end_addr and retry. Add
470 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
471 the insn pattern doesn't describe well what exactly happens to
475 * config/i386/predicates.md (avx_identity_operand): Remove.
476 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
477 (avx_<castmode><avxsizesuffix>_<castmode>,
478 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
479 a VEC_CONCAT of the operand and UNSPEC_CAST.
480 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
481 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
485 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
486 recog_data.insn if distance_non_agu_define changed it.
488 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
491 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
492 we only had X-FORM (reg+reg) addressing for vectors. Also before
493 ISA 3.0, we only had X-FORM addressing for scalars in the
494 traditional Altivec registers.
496 2020-02-06 <zhongyunde@huawei.com>
497 Vladimir Makarov <vmakarov@redhat.com>
499 PR rtl-optimization/93561
500 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
503 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
505 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
508 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
510 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
511 where the low and the high 32 bits are equal to each other specially,
512 with an rldimi instruction.
514 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
516 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
518 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
520 * config/arm/arm-tables.opt: Regenerate.
522 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
525 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
526 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
527 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
529 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
531 PR rtl-optimization/87763
532 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
534 2020-02-06 Delia Burduv <delia.burduv@arm.com>
536 * config/aarch64/aarch64-simd-builtins.def
537 (bfmlaq): New built-in function.
538 (bfmlalb): New built-in function.
539 (bfmlalt): New built-in function.
540 (bfmlalb_lane): New built-in function.
541 (bfmlalt_lane): New built-in function.
542 * config/aarch64/aarch64-simd.md
543 (aarch64_bfmmlaqv4sf): New pattern.
544 (aarch64_bfmlal<bt>v4sf): New pattern.
545 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
546 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
547 (vbfmlalbq_f32): New intrinsic.
548 (vbfmlaltq_f32): New intrinsic.
549 (vbfmlalbq_lane_f32): New intrinsic.
550 (vbfmlaltq_lane_f32): New intrinsic.
551 (vbfmlalbq_laneq_f32): New intrinsic.
552 (vbfmlaltq_laneq_f32): New intrinsic.
553 * config/aarch64/iterators.md (BF_MLA): New int iterator.
554 (bt): New int attribute.
556 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
558 * config/i386/i386.md (*pushtf): Emit "#" instead of
559 calling gcc_unreachable in insn output.
562 (*pushsf_rex64): Ditto for alternatives other than 1.
563 (*pushsf): Ditto for alternatives other than 1.
565 2020-02-06 Martin Liska <mliska@suse.cz>
567 PR gcov-profile/91971
568 PR gcov-profile/93466
569 * coverage.c (coverage_init): Revert mangling of
570 path into filename. It can lead to huge filename length.
571 Creation of subfolders seem more natural.
573 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
576 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
577 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
578 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
580 2020-02-06 Jakub Jelinek <jakub@redhat.com>
583 * config/i386/predicates.md (avx_identity_operand): New predicate.
584 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
585 define_insn_and_split.
588 * omp-low.c (use_pointer_for_field): For nested constructs, also
589 look for map clauses on target construct.
590 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
591 taskreg_nesting_level.
594 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
595 shared clause, call omp_notice_variable on outer context if any.
597 2020-02-05 Jason Merrill <jason@redhat.com>
600 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
601 non-zero address even if weak and not yet defined.
603 2020-02-05 Martin Sebor <msebor@redhat.com>
605 PR tree-optimization/92765
606 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
607 * tree-ssa-strlen.c (compute_string_length): Remove.
608 (determine_min_objsize): Remove.
609 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
610 Avoid using type size as the upper bound on string length.
611 (handle_builtin_string_cmp): Add an argument. Adjust.
612 (strlen_check_and_optimize_call): Pass additional argument to
613 handle_builtin_string_cmp.
615 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
617 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
618 (*pushdi2_rex64 peephole2): Unconditionally split after
620 (*ashl<mode>3_doubleword): Ditto.
621 (*<shift_insn><mode>3_doubleword): Ditto.
623 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
626 * config/rs6000/rs6000.c (get_vector_offset): Fix
628 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
630 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
632 2020-02-05 David Malcolm <dmalcolm@redhat.com>
635 (Special Functions for Debugging the Analyzer): Update description
636 of __analyzer_dump_exploded_nodes.
638 2020-02-05 Jakub Jelinek <jakub@redhat.com>
641 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
642 include sets and not clobbers in the vzeroupper pattern.
643 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
644 the parallel has 17 (64-bit) or 9 (32-bit) elts.
645 (*avx_vzeroupper_1): New define_insn_and_split.
648 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
649 don't run when !optimize.
650 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
653 2020-02-05 Richard Biener <rguenther@suse.de>
656 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
657 checks before matching calls.
659 2020-02-05 Jakub Jelinek <jakub@redhat.com>
661 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
662 function comment typo.
665 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
666 simd_clone_create failed when i == 0, adjust clone->nargs by
669 2020-02-05 Martin Liska <mliska@suse.cz>
672 * doc/invoke.texi: Document that one should
673 not combine ASLR and -fpch.
675 2020-02-04 Richard Biener <rguenther@suse.de>
677 PR tree-optimization/93538
678 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
680 2020-02-04 Richard Biener <rguenther@suse.de>
682 PR tree-optimization/91123
683 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
684 (vn_walk_cb_data::last_vuse): New member.
685 (vn_walk_cb_data::saved_operands): Likewsie.
686 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
687 (vn_walk_cb_data::push_partial_def): Use finish.
688 (vn_reference_lookup_2): Update last_vuse and use finish if
689 we've saved operands.
690 (vn_reference_lookup_3): Use finish and update calls to
691 push_partial_defs everywhere. When translating through
692 memcpy or aggregate copies save off operands and alias-set.
693 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
694 operation for redundant store removal.
696 2020-02-04 Richard Biener <rguenther@suse.de>
698 PR tree-optimization/92819
699 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
700 generating more stmts than before.
702 2020-02-04 Martin Liska <mliska@suse.cz>
704 * config/arm/arm.c (arm_gen_far_branch): Move the function
705 outside of selftests.
707 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
709 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
710 function to adjust PC-relative vector addresses.
711 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
712 handle vectors with PC-relative addresses.
714 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
716 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
718 (hard_reg_and_mode_to_addr_mask): Delete.
719 (rs6000_adjust_vec_address): If the original vector address
720 was REG+REG or REG+OFFSET and the element is not zero, do the add
721 of the elements in the original address before adding the offset
722 for the vector element. Use address_to_insn_form to validate the
723 address using the register being loaded, rather than guessing
724 whether the address is a DS-FORM or DQ-FORM address.
726 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
728 * config/rs6000/rs6000.c (get_vector_offset): New helper function
729 to calculate the offset in memory from the start of a vector of a
730 particular element. Add code to keep the element number in
731 bounds if the element number is variable.
732 (rs6000_adjust_vec_address): Move calculation of offset of the
733 vector element to get_vector_offset.
734 (rs6000_split_vec_extract_var): Do not do the initial AND of
735 element here, move the code to get_vector_offset.
737 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
739 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
742 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
744 * config/rs6000/constraints.md: Improve documentation.
746 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
749 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
750 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
752 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
754 * config.gcc: Remove "carrizo" support.
755 * config/gcn/gcn-opts.h (processor_type): Likewise.
756 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
757 * config/gcn/gcn.opt (gpu_type): Likewise.
758 * config/gcn/t-omp-device: Likewise.
760 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
763 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
764 * config/arm/arm.c (arm_gen_far_branch): New function
766 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
768 2020-02-03 Julian Brown <julian@codesourcery.com>
769 Tobias Burnus <tobias@codesourcery.com>
771 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
773 2020-02-03 Jakub Jelinek <jakub@redhat.com>
776 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
777 valid RTL to sum up the lowest and second lowest bytes of the popcnt
780 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
782 PR rtl-optimization/91333
783 * ira-color.c (struct allocno_color_data): Add member
785 (init_allocno_threads): Set the member up.
786 (bucket_allocno_compare_func): Add compare hard reg
789 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
791 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
793 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
794 * config.in: Regenerated.
795 * configure: Regenerated.
796 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
797 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
798 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
800 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
802 * configure: Regenerate.
804 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
806 PR rtl-optimization/91333
807 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
808 reg preferences comparison up.
810 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
812 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
813 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
814 aarch64-sve-builtins-base.h.
815 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
816 aarch64-sve-builtins-base.cc.
817 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
818 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
820 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
821 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
822 (svcvtnt): New functions.
823 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
824 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
825 (svcvtnt): New functions.
826 (svcvt): Add a form that converts f32 to bf16.
827 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
828 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
830 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
831 Treat B as bfloat16_t.
832 (ternary_bfloat_lane_base): New class.
833 (ternary_bfloat_def): Likewise.
834 (ternary_bfloat): New shape.
835 (ternary_bfloat_lane_def): New class.
836 (ternary_bfloat_lane): New shape.
837 (ternary_bfloat_lanex2_def): New class.
838 (ternary_bfloat_lanex2): New shape.
839 (ternary_bfloat_opt_n_def): New class.
840 (ternary_bfloat_opt_n): New shape.
841 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
842 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
843 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
844 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
845 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
846 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
847 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
848 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
849 the pattern off the narrow mode instead of the wider one.
850 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
851 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
852 (sve_fp_op): Handle them.
853 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
854 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
856 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
858 * config/aarch64/arm_sve.h: Include arm_bf16.h.
859 * config/aarch64/aarch64-modes.def (BF): Move definition before
860 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
861 (SVE_MODES): Handle BF modes.
862 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
864 (aarch64_full_sve_mode): Likewise.
865 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
867 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
868 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
869 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
870 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
872 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
874 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
875 (TYPES_all_data): Add bf16.
876 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
877 (register_tuple_type): Increase buffer size.
878 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
879 (bf16): New type suffix.
880 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
881 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
882 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
883 Change type from all_data to all_arith.
884 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
887 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
888 Matthew Malcomson <matthew.malcomson@arm.com>
889 Richard Sandiford <richard.sandiford@arm.com>
891 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
892 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
893 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
894 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
895 __ARM_FEATURE_MATMUL_FP64.
896 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
897 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
898 be disabled at the same time.
899 (f32mm): New extension.
900 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
901 (AARCH64_FL_F64MM): Bump to the next bit up.
902 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
903 (TARGET_SVE_F64MM): New macros.
904 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
905 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
906 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
907 (UNSPEC_ZIP2Q): New unspeccs.
908 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
909 (optab, sur, perm_insn): Handle the new unspecs.
910 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
911 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
912 TARGET_SVE_F64MM instead of separate tests.
913 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
914 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
915 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
916 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
917 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
918 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
919 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
920 (TYPES_s_signed): New macro.
921 (TYPES_s_integer): Use it.
922 (TYPES_d_float): New macro.
923 (TYPES_d_data): Use it.
924 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
925 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
926 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
927 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
929 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
931 (ternary_resize2_lane_base): Likewise.
932 (ternary_resize2_base): New class.
933 (ternary_qq_lane_base): Likewise.
934 (ternary_intq_uintq_lane_def): Likewise.
935 (ternary_intq_uintq_lane): New shape.
936 (ternary_intq_uintq_opt_n_def): New class
937 (ternary_intq_uintq_opt_n): New shape.
938 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
939 (ternary_uintq_intq_def): New class.
940 (ternary_uintq_intq): New shape.
941 (ternary_uintq_intq_lane_def): New class.
942 (ternary_uintq_intq_lane): New shape.
943 (ternary_uintq_intq_opt_n_def): New class.
944 (ternary_uintq_intq_opt_n): New shape.
945 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
946 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
947 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
948 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
950 (svdotprod_lane_impl): ...this new class.
951 (svmmla_impl, svusdot_impl): New classes.
952 (svdot_lane): Update to use svdotprod_lane_impl.
953 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
954 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
956 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
957 function, with no types defined.
958 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
959 AARCH64_FL_I8MM functions.
960 (svmmla): New AARCH64_FL_F32MM function.
961 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
962 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
963 AARCH64_FL_F64MM function.
964 (REQUIRED_EXTENSIONS):
966 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
968 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
971 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
973 * config/i386/i386.md (*movoi_internal_avx): Do not check for
974 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
975 (*movti_internal): Do not check for
976 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
977 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
978 just after check for TARGET_AVX.
979 (*movdf_internal): Ditto.
980 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
981 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
982 * config/i386/sse.md (mov<mode>_internal): Only check
983 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
984 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
985 (<sse>_andnot<mode>3<mask_name>): Move check for
986 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
987 (<code><mode>3<mask_name>): Ditto.
988 (*andnot<mode>3): Ditto.
990 (*<code><mode>3): Ditto.
992 (*andnot<VI:mode>3): Remove
993 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
994 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
995 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
996 (sse4_1_blendv<ssemodesuffix>): Ditto.
997 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
998 Explain that tune applies to 128bit instructions only.
1000 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1002 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1003 to definition of hsa_kernel_description. Parse assembly to find SGPR
1004 and VGPR count of kernel and store in hsa_kernel_description.
1006 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1008 PR rtl-optimization/91838
1009 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1010 to truncate if allowed or reject combination.
1012 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1014 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1015 (find_inv_vars_cb): Likewise.
1017 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1019 * calls.c (special_function_p): Split out the check for DECL_NAME
1020 being non-NULL and fndecl being extern at file scope into a
1021 new maybe_special_function_p and call it. Drop check for fndecl
1022 being non-NULL that was after a usage of DECL_NAME (fndecl).
1023 * tree.h (maybe_special_function_p): New inline function.
1025 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1027 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1028 (mask_gather_load<mode>): ... here, and zero-initialize the
1030 (maskload<mode>di): Zero-initialize the destination.
1033 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1036 * doc/analyzer.texi (Limitations): Note that constraints on
1037 floating-point values are currently ignored.
1039 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1042 * symtab.c (symtab_node::noninterposable_alias): If localalias
1043 already exists, but is not usable, append numbers after it until
1044 a unique name is found. Formatting fix.
1047 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1050 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1052 * config/gcn/gcn.c (print_operand): Handle LTGT.
1053 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1055 2020-01-30 Richard Biener <rguenther@suse.de>
1057 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1058 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1060 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1062 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1063 without a DECL in .data.rel.ro.local.
1065 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1068 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1072 * config/i386/sse.md
1073 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1074 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1075 any_extend code iterator instead of always zero_extend.
1076 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1077 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1078 Use any_extend code iterator instead of always zero_extend.
1079 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1080 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1081 Use any_extend code iterator instead of always zero_extend.
1082 (*sse2_pmovmskb_ext): New define_insn.
1083 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1086 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1087 (*popcountsi2_zext_falsedep): New define_insn.
1089 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1091 * config.in: Regenerated.
1092 * configure: Regenerated.
1094 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1097 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1098 LLVM's assembler changed the default in version 9.
1100 2020-01-24 Jeff Law <law@redhat.com>
1102 PR tree-optimization/89689
1103 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1105 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1109 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1111 PR rtl-optimization/87763
1112 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1113 simplification to handle subregs as well as bare regs.
1114 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1116 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1119 * ira.c (ira): Revert use of simplified LRA algorithm.
1121 2020-01-29 Martin Jambor <mjambor@suse.cz>
1123 PR tree-optimization/92706
1124 * tree-sra.c (struct access): Fields first_link, last_link,
1125 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1126 next_rhs_queued and grp_rhs_queued respectively, new fields
1127 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1128 (struct assign_link): Field next renamed to next_rhs, new field
1129 next_lhs. Updated comment.
1130 (work_queue_head): Renamed to rhs_work_queue_head.
1131 (lhs_work_queue_head): New variable.
1132 (add_link_to_lhs): New function.
1133 (relink_to_new_repr): Also relink LHS lists.
1134 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1135 (add_access_to_lhs_work_queue): New function.
1136 (pop_access_from_work_queue): Renamed to
1137 pop_access_from_rhs_work_queue.
1138 (pop_access_from_lhs_work_queue): New function.
1139 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1141 (child_would_conflict_in_lacc): Renamed to
1142 child_would_conflict_in_acc. Adjusted parameter names.
1143 (create_artificial_child_access): New parameter set_grp_read, use it.
1144 (subtree_mark_written_and_enqueue): Renamed to
1145 subtree_mark_written_and_rhs_enqueue.
1146 (propagate_subaccesses_across_link): Renamed to
1147 propagate_subaccesses_from_rhs.
1148 (propagate_subaccesses_from_lhs): New function.
1149 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1152 2020-01-29 Martin Jambor <mjambor@suse.cz>
1154 PR tree-optimization/92706
1155 * tree-sra.c (struct access): Adjust comment of
1156 grp_total_scalarization.
1157 (find_access_in_subtree): Look for single children spanning an entire
1159 (scalarizable_type_p): Allow register accesses, adjust callers.
1160 (completely_scalarize): Remove function.
1161 (scalarize_elem): Likewise.
1162 (create_total_scalarization_access): Likewise.
1163 (sort_and_splice_var_accesses): Do not track total scalarization
1165 (analyze_access_subtree): New parameter totally, adjust to new meaning
1166 of grp_total_scalarization.
1167 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1168 (can_totally_scalarize_forest_p): New function.
1169 (create_total_scalarization_access): Likewise.
1170 (create_total_access_and_reshape): Likewise.
1171 (total_should_skip_creating_access): Likewise.
1172 (totally_scalarize_subtree): Likewise.
1173 (analyze_all_variable_accesses): Perform total scalarization after
1174 subaccess propagation using the new functions above.
1175 (initialize_constant_pool_replacements): Output initializers by
1176 traversing the access tree.
1178 2020-01-29 Martin Jambor <mjambor@suse.cz>
1180 * tree-sra.c (verify_sra_access_forest): New function.
1181 (verify_all_sra_access_forests): Likewise.
1182 (create_artificial_child_access): Set parent.
1183 (analyze_all_variable_accesses): Call the verifier.
1185 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1187 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1188 if called on indirect edge.
1189 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1190 speculative call if needed.
1192 2020-01-29 Richard Biener <rguenther@suse.de>
1194 PR tree-optimization/93428
1195 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1196 permutation when the load node is created.
1197 (vect_analyze_slp_instance): Re-use it here.
1199 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1201 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1203 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1205 PR rtl-optimization/93272
1206 * ira-lives.c (process_out_of_region_eh_regs): New function.
1207 (process_bb_node_lives): Call it.
1209 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1211 * coverage.c (read_counts_file): Make error message lowercase.
1213 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1215 * profile-count.c (profile_quality_display_names): Fix ordering.
1217 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1220 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1221 hash only when edge is first within the sequence.
1222 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1223 (symbol_table::create_edge): Do not set target_prob.
1224 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1226 (cgraph_edge::make_speculative): Drop target_prob parameter.
1227 (cgraph_edge::speculative_call_info): Remove.
1228 (cgraph_edge::first_speculative_call_target): New member function.
1229 (update_call_stmt_hash_for_removing_direct_edge): New function.
1230 (cgraph_edge::resolve_speculation): Rewrite to new API.
1231 (cgraph_edge::speculative_call_for_target): New member function.
1232 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1233 multiple speculation targets.
1234 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1236 (verify_speculative_call): Verify that targets form an interval.
1237 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1238 (cgraph_edge::first_speculative_call_target): New member function.
1239 (cgraph_edge::next_speculative_call_target): New member function.
1240 (cgraph_edge::speculative_call_target_ref): New member function.
1241 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1242 (cgraph_edge): Remove target_prob.
1243 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1244 Fix handling of speculative calls.
1245 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1246 * ipa-fnsummary.c (analyze_function_body): Likewise.
1247 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1248 * ipa-profile.c (dump_histogram): Fix formating.
1249 (ipa_profile_generate_summary): Watch for overflows.
1250 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1251 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1252 (update_indirect_edges_after_inlining): Update to new API.
1253 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1255 * profile-count.h: (profile_probability::adjusted): New.
1256 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1257 updating of profile.
1258 * value-prof.c (gimple_ic_transform): Rename to ...
1259 (dump_ic_profile): ... this one; update dumping.
1260 (stream_in_histogram_value): Fix formating.
1261 (gimple_value_profile_transformations): Update.
1263 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1266 * config/i386/i386.md (*movoi_internal_avx): Remove
1267 TARGET_SSE_TYPELESS_STORES check.
1268 (*movti_internal): Prefer TARGET_AVX over
1269 TARGET_SSE_TYPELESS_STORES.
1270 (*movtf_internal): Likewise.
1271 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1272 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1273 from TARGET_SSE_TYPELESS_STORES.
1275 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1277 * diagnostic-core.h (warning_at): Rename overload to...
1278 (warning_meta): ...this.
1279 (emit_diagnostic_valist): Delete decl of overload taking
1280 diagnostic_metadata.
1281 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1282 (warning_at): Rename overload taking diagnostic_metadata to...
1283 (warning_meta): ...this.
1285 2020-01-28 Richard Biener <rguenther@suse.de>
1287 PR tree-optimization/93439
1288 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1289 * tree-cfg.c (move_sese_region_to_fn): ... here.
1290 (verify_types_in_gimple_reference): Verify used cliques are
1293 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1296 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1297 argument of a pointer to struct gcc_options and pass it to
1298 parse_mtune_ctrl_str.
1299 (ix86_function_specific_restore): Pass opts to
1300 set_ix86_tune_features.
1301 (ix86_option_override_internal): Likewise.
1302 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1303 gcc_options and use it for x_ix86_tune_ctrl_string.
1305 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1307 PR rtl-optimization/87763
1308 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1309 simplification to handle subregs as well as bare regs.
1310 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1312 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1314 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1315 for reduction chains that (now) include a call.
1317 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1319 PR tree-optimization/92822
1320 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1321 out the don't-care elements of a vector whose significant elements
1322 are duplicates, make the don't-care elements duplicates too.
1324 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1326 PR tree-optimization/93434
1327 * tree-predcom.c (split_data_refs_to_components): Record which
1328 components have had aliasing loads removed. Prevent store-store
1329 commoning for all such components.
1331 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1334 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1335 -1 or is_vshift is true, use new_vector with number of elts npatterns
1336 rather than new_unary_operation.
1338 PR tree-optimization/93454
1339 * gimple-fold.c (fold_array_ctor_reference): Perform
1340 elt_size.to_uhwi () just once, instead of calling it in every
1341 iteration. Punt if that value is above size of the temporary
1342 buffer. Decrease third native_encode_expr argument when
1343 bufoff + elt_sz is above size of buf.
1345 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1347 * config/mips/mips.c (mips_declare_object_name)
1348 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1350 2020-01-27 Martin Liska <mliska@suse.cz>
1352 PR gcov-profile/93403
1353 * tree-profile.c (gimple_init_gcov_profiler): Generate
1354 both __gcov_indirect_call_profiler_v4 and
1355 __gcov_indirect_call_profiler_v4_atomic.
1357 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1360 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1362 (@aarch64_split_simd_mov<mode>): Use it.
1363 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1364 Leave the vec_extract patterns to handle 2-element vectors.
1365 (aarch64_simd_mov_from_<mode>high): Likewise.
1366 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1367 (vec_extractv2dfv1df): Likewise.
1369 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1371 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1372 jump conditions for *compare_condjump<GPI:mode>.
1374 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1377 * digraph.cc (test_edge::test_edge): Specify template for base
1380 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1382 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1384 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1386 * config/arc/arc-protos.h (gen_mlo): Remove.
1387 (gen_mhi): Likewise.
1388 * config/arc/arc.c (AUX_MULHI): Define.
1389 (arc_must_save_reister): Special handling for r58/59.
1390 (arc_compute_frame_size): Consider mlo/mhi registers.
1391 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1393 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1394 mlo/mhi name selection.
1395 (arc_restore_callee_saves): Don't early restore blink when ISR.
1396 (arc_expand_prologue): Add mlo/mhi saving.
1397 (arc_expand_epilogue): Add mlo/mhi restoring.
1400 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1401 numbering when MUL64 option is used.
1402 (DWARF2_FRAME_REG_OUT): Define.
1403 * config/arc/arc.md (arc600_stall): New pattern.
1404 (VUNSPEC_ARC_ARC600_STALL): Define.
1405 (mulsi64): Use correct mlo/mhi registers.
1406 (mulsi_600): Clean it up.
1407 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1409 (mhi_operand): Likewise.
1411 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1412 Petro Karashchenko <petro.karashchenko@ring.com>
1414 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1415 attributes if needed.
1416 (prepare_move_operands): Generate special unspec instruction for
1418 (arc_isuncached_mem_p): Propagate uncached attribute to each
1420 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1421 (VUNSPEC_ARC_STDI): Likewise.
1422 (ALLI): New mode iterator.
1423 (mALLI): New mode attribute.
1424 (lddi): New instruction pattern.
1426 (stdidi_split): Split instruction for architectures which are not
1427 supporting ll64 option.
1428 (lddidi_split): Likewise.
1430 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1432 PR rtl-optimization/92989
1433 * lra-lives.c (process_bb_lives): Update the live-in set before
1434 processing additional clobbers.
1436 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1438 PR rtl-optimization/93170
1439 * cselib.c (cselib_invalidate_regno_val): New function, split out
1441 (cselib_invalidate_regno): ...here.
1442 (cselib_invalidated_by_call_p): New function.
1443 (cselib_process_insn): Iterate over all the hard-register entries in
1444 REG_VALUES and invalidate any that cross call-clobbered registers.
1446 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1448 * dojump.c (split_comparison): Use HONOR_NANS rather than
1449 HONOR_SNANS when splitting LTGT.
1451 2020-01-27 Martin Liska <mliska@suse.cz>
1454 * opts.c (print_filtered_help): Exclude language-specific
1455 options from --help=common unless enabled in all FEs.
1457 2020-01-27 Martin Liska <mliska@suse.cz>
1459 * opts.c (print_help): Exclude params from
1460 all except --help=param.
1462 2020-01-27 Martin Liska <mliska@suse.cz>
1465 * config/i386/i386-features.c (make_resolver_func):
1466 Align the code with ppc64 target implementation.
1467 Do not generate a unique name for resolver function.
1469 2020-01-27 Richard Biener <rguenther@suse.de>
1471 PR tree-optimization/93397
1472 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1473 converted reduction chain SLP graph adjustment.
1475 2020-01-26 Marek Polacek <polacek@redhat.com>
1478 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1481 2020-01-26 Jason Merrill <jason@redhat.com>
1484 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1487 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1489 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1490 (rx_setmem): Likewise.
1492 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1495 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1496 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1497 drop <di> from constraint of last operand.
1500 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1501 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1502 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1504 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1507 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1510 2020-01-24 Jeff Law <law@redhat.com>
1512 PR tree-optimization/92788
1513 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1516 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1519 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1520 *avx_vperm_broadcast_<mode>,
1521 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1522 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1523 Move before avx2_perm<mode>/avx512f_perm<mode>.
1526 * simplify-rtx.c (simplify_const_unary_operation,
1527 simplify_const_binary_operation): Punt for mode precision above
1528 MAX_BITSIZE_MODE_ANY_INT.
1530 2020-01-24 Andrew Pinski <apinski@marvell.com>
1532 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1535 2020-01-24 Jeff Law <law@redhat.com>
1538 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1539 for REGs. Call output_operand_lossage to get more reasonable
1542 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1544 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1545 gcn_fp_compare_operator.
1546 (vec_cmpu<mode>di): Use gcn_compare_operator.
1547 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1548 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1549 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1550 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1551 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1552 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1553 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1554 gcn_fp_compare_operator.
1555 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1556 gcn_fp_compare_operator.
1557 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1558 gcn_fp_compare_operator.
1559 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1560 gcn_fp_compare_operator.
1562 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1564 * doc/install.texi (Cross-Compiler-Specific Options): Document
1565 `--with-toolexeclibdir' option.
1567 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1569 * target.def (flags_regnum): Also mention effect on delay slot filling.
1570 * doc/tm.texi: Regenerate.
1572 2020-01-23 Jeff Law <law@redhat.com>
1574 PR translation/90162
1575 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1577 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1580 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1583 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1585 PR rtl-optimization/93402
1586 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1589 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1591 * config.in: Regenerated.
1592 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1593 for TARGET_LIBC_GNUSTACK.
1594 * configure: Regenerated.
1595 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1596 found to be 2.31 or greater.
1598 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1600 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1602 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1603 (mips_asm_file_end): New function. Delegate to
1604 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1605 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1607 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1610 * config/i386/i386-modes.def (POImode): New mode.
1611 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1612 * config/i386/i386.md (DPWI): New mode attribute.
1613 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1615 (QPWI): ... this. Use POI instead of OI for TImode.
1616 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1617 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1620 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1623 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1625 (speculation_tracker_rev): New pattern.
1626 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1627 Use speculation_tracker_rev to track the inverse condition.
1629 2020-01-23 Richard Biener <rguenther@suse.de>
1631 PR tree-optimization/93381
1632 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1633 alias-set of the def as argument and record the first one.
1634 (vn_walk_cb_data::first_set): New member.
1635 (vn_reference_lookup_3): Pass the alias-set of the current def
1636 to push_partial_def. Fix alias-set used in the aggregate copy
1638 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1639 * real.c (clear_significand_below): Fix out-of-bound access.
1641 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1644 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1645 New define_insn patterns.
1647 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1649 * doc/sourcebuild.texi (check-function-bodies): Add an
1650 optional target/xfail selector.
1652 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1654 PR rtl-optimization/93124
1655 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1656 bare USE and CLOBBER insns.
1658 2020-01-22 Andrew Pinski <apinski@marvell.com>
1660 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1662 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1665 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1666 diagnostic_manager into "ana" namespace.
1667 * selftest-run-tests.c (selftest::run_tests): Update for move of
1668 selftest::run_analyzer_selftests to
1669 ana::selftest::run_analyzer_selftests.
1671 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1673 * cfgexpand.c (union_stack_vars): Update the size.
1675 2020-01-22 Richard Biener <rguenther@suse.de>
1677 PR tree-optimization/93381
1678 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1679 throughout, handle all conversions the same.
1681 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1684 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1685 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1686 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1687 Call force_reg on high_in2 unconditionally.
1689 2020-01-22 Martin Liska <mliska@suse.cz>
1691 PR tree-optimization/92924
1692 * profile.c (compute_value_histograms): Divide
1695 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1698 * output.h (assemble_name_resolve): Declare.
1699 * varasm.c (assemble_name_resolve): New function.
1700 (assemble_name): Use it.
1701 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1703 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1705 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1706 update_web_docs_git instead of update_web_docs_svn.
1708 2020-01-21 Andrew Pinski <apinski@marvell.com>
1711 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1712 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1713 (*tlsgd_small_<mode>): Likewise.
1714 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1715 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1716 register. Convert that register back to dest using convert_mode.
1718 2020-01-21 Jim Wilson <jimw@sifive.com>
1720 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1723 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1724 Uros Bizjak <ubizjak@gmail.com>
1727 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1729 (legitimize_tls_address): Do GNU2 TLS address computation in
1730 ptr_mode and zero-extend result to Pmode.
1731 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1732 :P with :PTR and Pmode with ptr_mode.
1733 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1734 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1735 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1737 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1740 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1741 the last two operands are CONST_INT_P before using them as such.
1743 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1745 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1746 to get the integer element types.
1748 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1750 * config/aarch64/aarch64-sve-builtins.h
1751 (function_expander::convert_to_pmode): Declare.
1752 * config/aarch64/aarch64-sve-builtins.cc
1753 (function_expander::convert_to_pmode): New function.
1754 (function_expander::get_contiguous_base): Use it.
1755 (function_expander::prepare_gather_address_operands): Likewise.
1756 * config/aarch64/aarch64-sve-builtins-sve2.cc
1757 (svwhilerw_svwhilewr_impl::expand): Likewise.
1759 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1762 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1763 cfun->machine->label_is_assembled.
1764 (aarch64_print_patchable_function_entry): New.
1765 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1766 * config/aarch64/aarch64.h (struct machine_function): New field,
1769 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1772 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1775 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1778 * cgraph.c (cgraph_edge::resolve_speculation,
1779 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1780 call_stmt_site_hash.
1782 2020-01-21 Martin Liska <mliska@suse.cz>
1784 * config/rs6000/rs6000.c (common_mode_defined): Remove
1787 2020-01-21 Richard Biener <rguenther@suse.de>
1789 PR tree-optimization/92328
1790 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1791 type when value-numbering same-sized store by inserting a
1793 (eliminate_dom_walker::eliminate_stmt): When eliminating
1794 a redundant store handle bit-reinterpretation of the same value.
1796 2020-01-21 Andrew Pinski <apinski@marvel.com>
1799 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1801 (prepare_block_for_update): This. Use a worklist instead of
1804 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1806 * gcc/config/arm/arm.c (clear_operation_p):
1807 Initialise last_regno, skip first iteration
1808 based on the first_set value and use ints instead
1809 of the unnecessary HOST_WIDE_INTs.
1811 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1814 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1815 compare_mode other than SFmode or DFmode.
1817 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1820 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1821 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1822 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1824 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1826 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1828 2020-01-20 Andrew Pinski <apinski@marvell.com>
1831 * targhooks.c (default_print_patchable_function_entry): Use
1832 output_asm_insn to emit the nop instruction.
1834 2020-01-20 Fangrui Song <maskray@google.com>
1837 * targhooks.c (default_print_patchable_function_entry): Align to
1840 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
1843 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1844 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
1845 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1846 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
1847 (*tls_dynamic_gnu2_lea_64): Renamed to ...
1848 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
1849 Remove the {q} suffix from lea.
1850 (*tls_dynamic_gnu2_call_64): Renamed to ...
1851 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
1852 (*tls_dynamic_gnu2_combine_64): Renamed to ...
1853 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
1854 Pass Pmode to gen_tls_dynamic_gnu2_64.
1856 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1858 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1860 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
1862 * config/aarch64/aarch64-sve-builtins-base.cc
1863 (svld1ro_impl::memory_vector_mode): Remove parameter name.
1865 2020-01-20 Richard Biener <rguenther@suse.de>
1868 * dwarf2out.c (prune_unused_types): Unconditionally mark
1869 called function DIEs.
1871 2020-01-20 Martin Liska <mliska@suse.cz>
1873 PR tree-optimization/93199
1874 * tree-eh.c (struct leh_state): Add
1875 new field outer_non_cleanup.
1876 (cleanup_is_dead_in): Pass leh_state instead
1877 of eh_region. Add a checking that state->outer_non_cleanup
1878 points to outer non-clean up region.
1879 (lower_try_finally): Record outer_non_cleanup
1881 (lower_catch): Likewise.
1882 (lower_eh_filter): Likewise.
1883 (lower_eh_must_not_throw): Likewise.
1884 (lower_cleanup): Likewise.
1886 2020-01-20 Richard Biener <rguenther@suse.de>
1888 PR tree-optimization/93094
1889 * tree-vectorizer.h (vect_loop_versioning): Adjust.
1890 (vect_transform_loop): Likewise.
1891 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1892 loop_vectorized_call to vect_transform_loop.
1893 * tree-vect-loop.c (vect_transform_loop): Pass down
1894 loop_vectorized_call to vect_loop_versioning.
1895 * tree-vect-loop-manip.c (vect_loop_versioning): Use
1896 the earlier discovered loop_vectorized_call.
1898 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
1900 * doc/contribute.texi: Update for SVN -> Git transition.
1901 * doc/install.texi: Likewise.
1903 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1906 * cgraph.c (cgraph_edge::make_speculative): Increase number of
1907 speculative targets.
1908 (verify_speculative_call): New function
1909 (cgraph_node::verify_node): Use it.
1910 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
1913 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1916 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
1917 (cgraph_edge::make_direct): Remove all indirect targets.
1918 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
1919 (cgraph_node::verify_node): Verify that only one call_stmt or
1920 lto_stmt_uid is set.
1921 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
1923 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
1924 (lto_output_ref): Simplify streaming of stmt.
1925 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
1927 2020-01-18 Tamar Christina <tamar.christina@arm.com>
1929 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
1930 Mark parameter unused.
1932 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
1934 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
1936 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
1938 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
1940 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
1942 * Makefile.in: Add coroutine-passes.o.
1943 * builtin-types.def (BT_CONST_SIZE): New.
1944 (BT_FN_BOOL_PTR): New.
1945 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
1946 * builtins.def (DEF_COROUTINE_BUILTIN): New.
1947 * coroutine-builtins.def: New file.
1948 * coroutine-passes.cc: New file.
1949 * function.h (struct GTY function): Add a bit to indicate that the
1950 function is a coroutine component.
1951 * internal-fn.c (expand_CO_FRAME): New.
1952 (expand_CO_YIELD): New.
1953 (expand_CO_SUSPN): New.
1954 (expand_CO_ACTOR): New.
1955 * internal-fn.def (CO_ACTOR): New.
1959 * passes.def: Add pass_coroutine_lower_builtins,
1960 pass_coroutine_early_expand_ifns.
1961 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
1962 (make_pass_coroutine_early_expand_ifns): New.
1963 * doc/invoke.texi: Document the fcoroutines command line
1966 2020-01-18 Jakub Jelinek <jakub@redhat.com>
1968 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
1971 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
1972 after checking the argument is a REG. Don't use REGNO (reg)
1973 again to set last_regno, reuse regno variable instead.
1975 2020-01-17 David Malcolm <dmalcolm@redhat.com>
1977 * doc/analyzer.texi (Limitations): Add note about NaN.
1979 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1980 Sudakshina Das <sudi.das@arm.com>
1982 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
1983 and valid immediate.
1984 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
1985 (lshrdi3): Generate thumb2_lsrl for valid immediates.
1986 * config/arm/constraints.md (Pg): New.
1987 * config/arm/predicates.md (long_shift_imm): New.
1988 (arm_reg_or_long_shift_imm): Likewise.
1989 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
1990 (thumb2_lsll): Likewise.
1993 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1994 Sudakshina Das <sudi.das@arm.com>
1996 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
1997 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
1998 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
1999 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2000 * config/arm/thumb2.md (thumb2_asrl): New.
2001 (thumb2_lsll): Likewise.
2003 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2005 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2008 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2010 * gdbinit.in (help-gcc-hooks): New command.
2011 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2012 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2015 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2017 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2018 correct target macro.
2020 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2022 * config/aarch64/aarch64-protos.h
2023 (aarch64_sve_ld1ro_operand_p): New.
2024 * config/aarch64/aarch64-sve-builtins-base.cc
2025 (class load_replicate): New.
2026 (class svld1ro_impl): New.
2027 (class svld1rq_impl): Change to inherit from load_replicate.
2028 (svld1ro): New sve intrinsic function base.
2029 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2030 New DEF_SVE_FUNCTION.
2031 * config/aarch64/aarch64-sve-builtins-base.h
2032 (svld1ro): New decl.
2033 * config/aarch64/aarch64-sve-builtins.cc
2034 (function_expander::add_mem_operand): Modify assert to allow
2036 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2038 * config/aarch64/aarch64.c
2039 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2040 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2041 (aarch64_sve_ld1ro_operand_p): New.
2042 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2043 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2044 * config/aarch64/predicates.md
2045 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2047 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2049 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2050 Introduce this ACLE specified predefined macro.
2051 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2052 (fp): Disabling this disables f64mm.
2053 (simd): Disabling this disables f64mm.
2054 (fp16): Disabling this disables f64mm.
2055 (sve): Disabling this disables f64mm.
2056 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2057 (AARCH64_ISA_F64MM): New.
2058 (TARGET_F64MM): New.
2059 * doc/invoke.texi (f64mm): Document new option.
2061 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2063 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2064 (neoversen1_tunings): Likewise.
2066 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2069 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2070 Add assert to ensure prolog has been emitted.
2071 (aarch64_split_atomic_op): Likewise.
2072 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2073 Use epilogue_completed rather than reload_completed.
2074 (aarch64_atomic_exchange<mode>): Likewise.
2075 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2076 (atomic_nand<mode>): Likewise.
2077 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2078 (atomic_fetch_nand<mode>): Likewise.
2079 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2080 (atomic_nand_fetch<mode>): Likewise.
2082 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2085 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2087 (REVERSE_CONDITION): Delete.
2088 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2089 (CCFP_CCFPE): Likewise.
2090 (e): New mode attribute.
2091 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2092 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2093 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2094 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2095 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2096 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2097 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2098 name of generator from gen_ccmpdi to gen_ccmpccdi.
2099 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2100 the previous comparison but aren't able to, use the new ccmp_rev
2103 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2105 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2106 than testing directly for INTEGER_CST.
2107 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2109 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2111 PR tree-optimization/93292
2112 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2113 get_vectype_for_scalar_type returns NULL.
2115 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2117 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2118 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2120 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2122 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2124 * params.opt: (max-predicted-iterations): Set bounds.
2125 * predict.c (real_almost_one, real_br_prob_base,
2126 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2127 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2128 probabilities; do not truncate to reg_br_prob_bases.
2129 (estimate_loops_at_level): Pass max_cyclic_prob.
2130 (estimate_loops): Compute max_cyclic_prob.
2131 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2133 * profile-count.c (profile_probability::to_sreal): New.
2134 * profile-count.h (class sreal): Move up in file.
2135 (profile_probability::to_sreal): Declare.
2137 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2140 (arm_invalid_conversion): New function for target hook.
2141 (arm_invalid_unary_op): New function for target hook.
2142 (arm_invalid_binary_op): New function for target hook.
2144 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2146 * config.gcc: Add arm_bf16.h.
2147 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2148 (arm_simd_builtin_std_type): Add BFmode.
2149 (arm_init_simd_builtin_types): Define element types for vector types.
2150 (arm_init_bf16_types): New function.
2151 (arm_init_builtins): Add arm_init_bf16_types function call.
2152 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2153 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2154 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2155 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2156 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2157 (arm_mangle_type): Add __bf16.
2158 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2159 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2160 arm_bf16_ptr_type_node.
2161 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2162 define_split between ARM registers.
2163 * config/arm/arm_bf16.h: New file.
2164 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2165 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2167 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2168 * config/arm/vfp.md: Add BFmode to movhf patterns.
2170 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2171 Andre Vieira <andre.simoesdiasvieira@arm.com>
2173 * config/arm/arm-cpus.in (mve, mve_float): New features.
2174 (dsp, mve, mve.fp): New options.
2175 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2176 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2177 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2179 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2180 Thomas Preud'homme <thomas.preudhomme@arm.com>
2182 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2184 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2185 error for using -mcmse when targeting Armv8.1-M Mainline.
2187 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2188 Thomas Preud'homme <thomas.preudhomme@arm.com>
2190 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2191 address in r4 when targeting Armv8.1-M Mainline.
2192 (nonsecure_call_value_internal): Likewise.
2193 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2194 a register match_operand again. Emit BLXNS when targeting
2196 (nonsecure_call_value_reg_thumb2): Likewise.
2198 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2199 Thomas Preud'homme <thomas.preudhomme@arm.com>
2201 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2202 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2203 variable as true when floating-point ABI is not hard. Replace
2204 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2205 Generate VLSTM and VLLDM instruction respectively before and
2206 after a function call to cmse_nonsecure_call function.
2207 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2208 (VUNSPEC_VLLDM): Likewise.
2209 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2210 (lazy_load_multiple_insn): Likewise.
2212 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2213 Thomas Preud'homme <thomas.preudhomme@arm.com>
2215 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2216 (arm_emit_vfp_multi_reg_pop): Likewise.
2217 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2218 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2219 restore callee-saved VFP registers.
2221 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2222 Thomas Preud'homme <thomas.preudhomme@arm.com>
2224 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2225 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2226 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2227 callee-saved GPRs as well as clear ip register before doing a nonsecure
2228 call then restore callee-saved GPRs after it when targeting
2230 (arm_reorg): Adapt to function rename.
2232 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2233 Thomas Preud'homme <thomas.preudhomme@arm.com>
2235 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2236 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2237 clear_vfp_multiple pattern based on a new vfp parameter.
2238 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2239 targeting Armv8.1-M Mainline.
2240 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2241 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2242 whether VFP registers are available before looking call_used_regs for a
2244 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2245 of prototype of clear_operation_p.
2246 (clear_vfp_multiple_operation): New predicate.
2247 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2248 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2250 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2251 Thomas Preud'homme <thomas.preudhomme@arm.com>
2253 * config/arm/arm-protos.h (clear_operation_p): Declare.
2254 * config/arm/arm.c (clear_operation_p): New function.
2255 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2256 targeting Armv8.1-M Mainline or successor.
2257 (output_return_instruction): Only output APSR register clearing if
2258 Armv8.1-M Mainline instructions not available.
2259 (thumb_exit): Likewise.
2260 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2261 * config/arm/thumb2.md (clear_apsr): New define_insn.
2262 (clear_multiple): Likewise.
2263 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2265 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2266 Thomas Preud'homme <thomas.preudhomme@arm.com>
2268 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2269 (use_return_insn): Also return false for Armv8.1-M Mainline.
2270 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2271 Mainline instructions are available.
2272 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2273 when targeting Armv8.1-M Mainline Security Extensions.
2274 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2275 Mainline entry function.
2276 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2277 targeting Armv8.1-M Mainline or successor.
2278 (arm_expand_epilogue): Fix indentation of caller-saved register
2279 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2281 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2282 (FP_SYSREGS): Likewise.
2283 (enum vfp_sysregs_encoding): Define enum.
2284 (fp_sysreg_names): Declare.
2285 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2286 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2287 (pop_fpsysreg_insn): Likewise.
2289 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2290 Thomas Preud'homme <thomas.preudhomme@arm.com>
2292 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2293 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2294 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2295 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2296 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2297 (ARMv8_1m_main): New feature group.
2298 (armv8.1-m.main): New architecture.
2299 * config/arm/arm-tables.opt: Regenerate.
2300 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2301 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2302 (arm_options_perform_arch_sanity_checks): Error out when targeting
2303 Armv8.1-M Mainline Security Extensions.
2304 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2306 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2308 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2309 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2310 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2311 aarch64_bfdot_laneq): New.
2312 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2313 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2314 vbfdotq_laneq_f32): New.
2315 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2316 VBFMLA_W, VBF): New.
2317 (isquadop): Add V4BF, V8BF.
2319 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2321 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2322 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2323 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2324 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2325 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2326 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2327 usdot_laneq, sudot_lane,sudot_laneq): New.
2328 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2329 (aarch64_<sur>dot_lane): New.
2330 * config/aarch64/arm_neon.h (vusdot_s32): New.
2332 (vusdot_lane_s32): New.
2333 (vsudot_lane_s32): New.
2334 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2335 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2337 2020-01-16 Martin Liska <mliska@suse.cz>
2339 * value-prof.c (dump_histogram_value): Fix
2340 obvious spacing issue.
2342 2020-01-16 Andrew Pinski <apinski@marvell.com>
2344 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2345 !storage_order_barrier_p.
2347 2020-01-16 Andrew Pinski <apinski@marvell.com>
2349 * sched-int.h (_dep): Add unused bit-field field for the padding.
2350 * sched-deps.c (init_dep_1): Init unused field.
2352 2020-01-16 Andrew Pinski <apinski@marvell.com>
2354 * optabs.h (create_expand_operand): Initialize target field also.
2356 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2358 PR tree-optimization/92429
2359 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2360 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2362 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2365 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2367 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2368 aarch64_sve_int_mode to each mode.
2370 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2372 * doc/analyzer.texi (Overview): Add note about
2373 -fdump-ipa-analyzer.
2375 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2377 PR tree-optimization/93231
2378 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2379 input_type is unsigned. Use tree_to_shwi for shift constant.
2380 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2381 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2382 inputs more efficiently.
2384 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2386 * config/i386/i386.md (*movsf_internal): Do not require
2387 SSE2 ISA for alternatives 14 and 15.
2389 2020-01-15 Richard Biener <rguenther@suse.de>
2392 * tree-eh.c (sink_clobbers): If we already visited the destination
2393 block do not defer insertion.
2394 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2395 the purpose of defered insertion.
2397 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2399 * BASE-VER: Bump to 10.0.1.
2401 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2403 PR tree-optimization/93247
2404 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2405 type of the stmt that we're going to vectorize.
2407 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2409 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2410 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2413 2020-01-15 Martin Liska <mliska@suse.cz>
2415 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2416 2 calls of streamer_read_hwi in a function call.
2418 2020-01-15 Richard Biener <rguenther@suse.de>
2420 * alias.c (record_alias_subset): Avoid redundant work when
2421 subset is already recorded.
2423 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2425 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2426 the analyzer options provide CWE identifiers.
2428 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2430 * tree-diagnostic-path.cc (path_summary::event_range::print):
2431 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2432 using get_pure_location.
2434 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2436 PR tree-optimization/93262
2437 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2438 perform head trimming only if the last argument is constant,
2439 either all ones, or larger or equal to head trim, in the latter
2440 case decrease the last argument by head_trim.
2442 PR tree-optimization/93249
2443 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2444 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2445 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2446 perform head trim unless we can prove there are no '\0' chars
2447 from the source among the first head_trim chars.
2449 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2451 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2453 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2456 * config/i386/sse.md
2457 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2458 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2459 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2460 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2461 just a single alternative instead of two, make operands 1 and 2
2464 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2467 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2470 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2472 * Makefile.in (lang_opt_files): Add analyzer.opt.
2473 (ANALYZER_OBJS): New.
2474 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2475 tristate.o and ANALYZER_OBJS.
2476 (TEXI_GCCINT_FILES): Add analyzer.texi.
2477 * common.opt (-fanalyzer): New driver option.
2478 * config.in: Regenerate.
2479 * configure: Regenerate.
2480 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2481 (gccdepdir): Also create depdir for "analyzer" subdir.
2482 * digraph.cc: New file.
2483 * digraph.h: New file.
2484 * doc/analyzer.texi: New file.
2485 * doc/gccint.texi ("Static Analyzer") New menu item.
2486 (analyzer.texi): Include it.
2487 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2488 ("Warning Options"): Add static analysis warnings to the list.
2489 (-Wno-analyzer-double-fclose): New option.
2490 (-Wno-analyzer-double-free): New option.
2491 (-Wno-analyzer-exposure-through-output-file): New option.
2492 (-Wno-analyzer-file-leak): New option.
2493 (-Wno-analyzer-free-of-non-heap): New option.
2494 (-Wno-analyzer-malloc-leak): New option.
2495 (-Wno-analyzer-possible-null-argument): New option.
2496 (-Wno-analyzer-possible-null-dereference): New option.
2497 (-Wno-analyzer-null-argument): New option.
2498 (-Wno-analyzer-null-dereference): New option.
2499 (-Wno-analyzer-stale-setjmp-buffer): New option.
2500 (-Wno-analyzer-tainted-array-index): New option.
2501 (-Wno-analyzer-use-after-free): New option.
2502 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2503 (-Wno-analyzer-use-of-uninitialized-value): New option.
2504 (-Wanalyzer-too-complex): New option.
2505 (-fanalyzer-call-summaries): New warning.
2506 (-fanalyzer-checker=): New warning.
2507 (-fanalyzer-fine-grained): New warning.
2508 (-fno-analyzer-state-merge): New warning.
2509 (-fno-analyzer-state-purge): New warning.
2510 (-fanalyzer-transitivity): New warning.
2511 (-fanalyzer-verbose-edges): New warning.
2512 (-fanalyzer-verbose-state-changes): New warning.
2513 (-fanalyzer-verbosity=): New warning.
2514 (-fdump-analyzer): New warning.
2515 (-fdump-analyzer-callgraph): New warning.
2516 (-fdump-analyzer-exploded-graph): New warning.
2517 (-fdump-analyzer-exploded-nodes): New warning.
2518 (-fdump-analyzer-exploded-nodes-2): New warning.
2519 (-fdump-analyzer-exploded-nodes-3): New warning.
2520 (-fdump-analyzer-supergraph): New warning.
2521 * doc/sourcebuild.texi (dg-require-dot): New.
2522 (dg-check-dot): New.
2523 * gdbinit.in (break-on-saved-diagnostic): New command.
2524 * graphviz.cc: New file.
2525 * graphviz.h: New file.
2526 * ordered-hash-map-tests.cc: New file.
2527 * ordered-hash-map.h: New file.
2528 * passes.def (pass_analyzer): Add before
2529 pass_ipa_whole_program_visibility.
2530 * selftest-run-tests.c (selftest::run_tests): Call
2531 selftest::ordered_hash_map_tests_cc_tests.
2532 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2534 * shortest-paths.h: New file.
2535 * timevar.def (TV_ANALYZER): New timevar.
2536 (TV_ANALYZER_SUPERGRAPH): Likewise.
2537 (TV_ANALYZER_STATE_PURGE): Likewise.
2538 (TV_ANALYZER_PLAN): Likewise.
2539 (TV_ANALYZER_SCC): Likewise.
2540 (TV_ANALYZER_WORKLIST): Likewise.
2541 (TV_ANALYZER_DUMP): Likewise.
2542 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2543 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2544 * tree-pass.h (make_pass_analyzer): New decl.
2545 * tristate.cc: New file.
2546 * tristate.h: New file.
2548 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2551 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2552 alternatives 9 and 10.
2554 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2556 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2557 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2558 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2559 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2560 (selftest::hash_map_tests_c_tests): Call it.
2561 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2562 New static constant, using the value of = H::empty_zero_p.
2563 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2564 from default_hash_traits <Value>.
2565 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2567 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2568 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2569 calls to mark_empty with !Descriptor::empty_zero_p.
2570 (hash_table::empty_slow): Conditionalize the memset call with a
2571 check that Descriptor::empty_zero_p; otherwise, loop through the
2572 entries calling mark_empty on them.
2573 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2574 (pointer_hash::empty_zero_p): Likewise.
2575 (pair_hash::empty_zero_p): Likewise.
2576 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2578 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2579 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2580 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2581 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2582 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2583 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2584 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2585 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2587 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2590 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2592 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2593 fix typo on return value.
2595 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2598 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2600 (cgraph_edge::make_speculative): Add param for setting speculative_id
2602 (cgraph_edge::speculative_call_info): Update comments and find reference
2603 by speculative_id for multiple indirect targets.
2604 (cgraph_edge::resolve_speculation): Decrease the speculations
2605 for indirect edge, drop it's speculative if not direct target
2606 left. Update comments.
2607 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2608 (cgraph_node::dump): Print num_speculative_call_targets.
2609 (cgraph_node::verify_node): Don't report error if speculative
2610 edge not include statement.
2611 (cgraph_edge::num_speculative_call_targets_p): New function.
2612 * cgraph.h (int common_target_id): Remove.
2613 (int common_target_probability): Remove.
2614 (num_speculative_call_targets): New variable.
2615 (make_speculative): Add param for setting speculative_id.
2616 (cgraph_edge::num_speculative_call_targets_p): New declare.
2617 (target_prob): New variable.
2618 (speculative_id): New variable.
2619 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2620 call summaries for multiple speculative call targets.
2621 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2622 * ipa-profile.c (struct speculative_call_target): New struct.
2623 (class speculative_call_summary): New class.
2624 (class speculative_call_summaries): New class.
2625 (call_sums): New variable.
2626 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2627 (ipa_profile_write_edge_summary): New function.
2628 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2629 (ipa_profile_dump_all_summaries): New function.
2630 (ipa_profile_read_edge_summary): New function.
2631 (ipa_profile_read_summary_section): New function.
2632 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2633 (ipa_profile): Generate num_speculative_call_targets from
2635 * ipa-ref.h (speculative_id): New variable.
2636 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2637 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2638 common_target_probability. Stream out speculative_id and
2639 num_speculative_call_targets.
2640 (input_edge): Likewise.
2641 * predict.c (dump_prediction): Remove edges count assert to be
2643 * symtab.c (symtab_node::create_reference): Init speculative_id.
2644 (symtab_node::clone_references): Clone speculative_id.
2645 (symtab_node::clone_referring): Clone speculative_id.
2646 (symtab_node::clone_reference): Clone speculative_id.
2647 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2648 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2649 if indirect call contains multiple speculative targets.
2650 * value-prof.h (check_ic_target): Remove.
2651 * value-prof.c (gimple_value_profile_transformations):
2652 Use void function gimple_ic_transform.
2653 * value-prof.c (gimple_ic_transform): Handle topn case.
2654 Fix comment typos. Change it to a void function.
2656 2020-01-13 Andrew Pinski <apinski@marvell.com>
2658 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2659 (octeontx2t98): New define.
2660 (octeontx2t96): New define.
2661 (octeontx2t93): New define.
2662 (octeontx2f95): New define.
2663 (octeontx2f95n): New define.
2664 (octeontx2f95mm): New define.
2665 * config/aarch64/aarch64-tune.md: Regenerate.
2666 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2668 2020-01-13 Jason Merrill <jason@redhat.com>
2670 PR c++/33799 - destroy return value if local cleanup throws.
2671 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2673 2020-01-13 Martin Liska <mliska@suse.cz>
2675 * ipa-cp.c (get_max_overall_size): Use newly
2676 renamed param param_ipa_cp_unit_growth.
2677 * params.opt: Remove legacy param name.
2679 2020-01-13 Martin Sebor <msebor@redhat.com>
2681 PR tree-optimization/93213
2682 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2683 stores to be eliminated.
2685 2020-01-13 Martin Liska <mliska@suse.cz>
2687 * opts.c (print_help): Do not print CL_PARAM
2688 and CL_WARNING for CL_OPTIMIZATION.
2690 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2693 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2694 depending on optimization settings.
2696 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2698 PR tree-optimization/90838
2699 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2700 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2701 argument rather than to initialize temporary for targets that
2702 don't use the mode argument at all. Initialize ctzval to avoid
2705 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2707 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2708 * tree-core.h: Document it.
2709 * gimplify.c (gimplify_omp_workshare): Set it.
2710 * omp-low.c (lower_omp_target): Use it.
2711 * tree-pretty-print.c (dump_omp_clause): Print it.
2713 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2714 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2716 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2718 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2719 * common.opt (fdiagnostics-path-format=): New option.
2720 (diagnostic_path_format): New enum.
2721 (fdiagnostics-show-path-depths): New option.
2722 * coretypes.h (diagnostic_event_id_t): New forward decl.
2723 * diagnostic-color.c (color_dict): Add "path".
2724 * diagnostic-event-id.h: New file.
2725 * diagnostic-format-json.cc (json_from_expanded_location): Make
2727 (json_end_diagnostic): Call context->make_json_for_path if it
2728 exists and the diagnostic has a path.
2729 (diagnostic_output_format_init): Clear context->print_path.
2730 * diagnostic-path.h: New file.
2731 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2732 when printing a run of events in a diagnostic_path so that they
2733 all get the same color.
2734 (layout::m_diagnostic_path_p): New field.
2735 (layout::layout): Initialize it.
2736 (layout::print_any_labels): Don't colorize the label text for an
2737 event in a diagnostic_path.
2738 (gcc_rich_location::add_location_if_nearby): Add
2739 "restrict_to_current_line_spans" and "label" params. Pass the
2740 former to layout.maybe_add_location_range; pass the latter
2741 when calling add_range.
2742 * diagnostic.c: Include "diagnostic-path.h".
2743 (diagnostic_initialize): Initialize context->path_format and
2744 context->show_path_depths.
2745 (diagnostic_show_any_path): New function.
2746 (diagnostic_path::interprocedural_p): New function.
2747 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2748 (simple_diagnostic_path::num_events): New function.
2749 (simple_diagnostic_path::get_event): New function.
2750 (simple_diagnostic_path::add_event): New function.
2751 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2752 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2753 (debug): New overload taking a diagnostic_path *.
2754 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2755 * diagnostic.h (enum diagnostic_path_format): New enum.
2756 (json::value): New forward decl.
2757 (diagnostic_context::path_format): New field.
2758 (diagnostic_context::show_path_depths): New field.
2759 (diagnostic_context::print_path): New callback field.
2760 (diagnostic_context::make_json_for_path): New callback field.
2761 (diagnostic_show_any_path): New decl.
2762 (json_from_expanded_location): New decl.
2763 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2764 (-fdiagnostics-show-path-depths): New option.
2765 (-fdiagnostics-color): Add "path" to description of default
2766 GCC_COLORS; describe it.
2767 (-fdiagnostics-format=json): Document how diagnostic paths are
2768 represented in the JSON output format.
2769 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2770 Add optional params "restrict_to_current_line_spans" and "label".
2771 * opts.c (common_handle_option): Handle
2772 OPT_fdiagnostics_path_format_ and
2773 OPT_fdiagnostics_show_path_depths.
2774 * pretty-print.c: Include "diagnostic-event-id.h".
2775 (pp_format): Implement "%@" format code for printing
2776 diagnostic_event_id_t *.
2777 (selftest::test_pp_format): Add tests for "%@".
2778 * selftest-run-tests.c (selftest::run_tests): Call
2779 selftest::tree_diagnostic_path_cc_tests.
2780 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2781 * toplev.c (general_init): Initialize global_dc->path_format and
2782 global_dc->show_path_depths.
2783 * tree-diagnostic-path.cc: New file.
2784 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2785 non-static. Drop "diagnostic" param in favor of storing the
2786 original value of "where" and re-using it.
2787 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2788 maybe_unwind_expanded_macro_loc.
2789 (tree_diagnostics_defaults): Initialize context->print_path and
2790 context->make_json_for_path.
2791 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2793 (default_tree_make_json_for_path): New decl.
2794 (maybe_unwind_expanded_macro_loc): New decl.
2796 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2798 PR tree-optimization/93210
2799 * fold-const.h (native_encode_initializer,
2800 can_native_interpret_type_p): Declare.
2801 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2803 (native_encode_initializer): New function, moved from dwarf2out.c.
2804 Adjust to native_encode_expr compatible arguments, including dry-run
2805 and partial extraction modes. Don't handle STRING_CST.
2806 (can_native_interpret_type_p): No longer static.
2807 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2808 offset / BITS_PER_UNIT fits into int and don't call it if
2809 can_native_interpret_type_p fails. If suboff is NULL and for
2810 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2811 native_encode_initializer.
2812 (fold_const_aggregate_ref_1): Formatting fix.
2813 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2814 (tree_add_const_value_attribute): Adjust caller.
2816 PR tree-optimization/90838
2817 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2818 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2819 CTZ_DEFINED_VALUE_AT_ZERO.
2821 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2824 * lra-constraints.c (match_reload): Permit input operands have the
2825 same mode as output while other input operands have a different
2828 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
2830 PR tree-optimization/90838
2831 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2832 (check_ctz_string): Likewise.
2833 (optimize_count_trailing_zeroes): Likewise.
2834 (simplify_count_trailing_zeroes): Likewise.
2835 (pass_forwprop::execute): Try ctz simplification.
2836 * match.pd: Add matching for ctz idioms.
2838 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2840 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2842 (aarch64_invalid_unary_op): New function for target hook.
2843 (aarch64_invalid_binary_op): New function for target hook.
2845 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2847 * config.gcc: Add arm_bf16.h.
2848 * config/aarch64/aarch64-builtins.c
2849 (aarch64_simd_builtin_std_type): Add BFmode.
2850 (aarch64_init_simd_builtin_types): Define element types for vector
2852 (aarch64_init_bf16_types): New function.
2853 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2854 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2856 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2857 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2859 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2860 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2861 * config/aarch64/aarch64.c
2862 (aarch64_classify_vector_mode): Add support for BF types.
2863 (aarch64_gimplify_va_arg_expr): Add support for BF types.
2864 (aarch64_vq_mode): Add support for BF types.
2865 (aarch64_simd_container_mode): Add support for BF types.
2866 (aarch64_mangle_type): Add support for BF scalar type.
2867 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2868 * config/aarch64/arm_bf16.h: New file.
2869 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2870 * config/aarch64/iterators.md: Add BF types to mode attributes.
2871 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2873 2020-01-10 Jason Merrill <jason@redhat.com>
2875 PR c++/93173 - incorrect tree sharing.
2876 * gimplify.c (copy_if_shared): No longer static.
2877 * gimplify.h: Declare it.
2879 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2881 * doc/invoke.texi (-msve-vector-bits=): Document that
2882 -msve-vector-bits=128 now generates VL-specific code for
2883 little-endian targets.
2884 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2885 build_vector_type_for_mode to construct the data vector types.
2886 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2887 VL-specific code for -msve-vector-bits=128 on little-endian targets.
2888 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2889 for 128-bit vectors.
2891 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2893 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2896 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2898 * config/aarch64/aarch64-builtins.c
2899 (aarch64_builtin_vectorized_function): Check for specific vector modes,
2900 rather than checking the number of elements and the element mode.
2902 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2904 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2905 get_related_vectype_for_scalar_type rather than build_vector_type
2906 to create the index type for a conditional reduction.
2908 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2910 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
2911 for any type of gather or scatter, including strided accesses.
2913 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2915 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
2918 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2920 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
2922 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
2923 parameter and its use to reset DR_OFFSET's.
2924 (vect_transform_loop): Remove orig_drs_init argument.
2925 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
2926 member of dr_vec_info rather than the offset of the associated
2927 data_reference's innermost_loop_behavior.
2928 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
2929 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
2930 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
2931 get_dr_vinfo_offset.
2932 (vectorizable_store): Likewise.
2933 (vectorizable_load): Likewise.
2935 2020-01-10 Richard Biener <rguenther@suse.de>
2937 * gimple-ssa-store-merging
2938 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
2940 2020-01-10 Martin Liska <mliska@suse.cz>
2943 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
2944 encapsulation that was there before r280040.
2946 2020-01-10 Richard Biener <rguenther@suse.de>
2949 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
2950 sequences to avoid walking them again for secondary opportunities.
2951 (pass_lower_eh_dispatch::execute): Instead actually insert
2954 2020-01-10 Richard Biener <rguenther@suse.de>
2957 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
2958 (cleanup_all_empty_eh): Walk landing pads in reverse order to
2959 avoid quadraticness.
2961 2020-01-10 Martin Jambor <mjambor@suse.cz>
2963 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
2964 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
2965 to get param_ipa_sra_max_replacements.
2966 (param_splitting_across_edge): Pass the caller to
2967 pull_accesses_from_callee.
2969 2020-01-10 Martin Jambor <mjambor@suse.cz>
2971 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
2972 * ipa-cp.c (max_new_size): Removed.
2973 (orig_overall_size): New variable.
2974 (get_max_overall_size): New function.
2975 (estimate_local_effects): Use it. Adjust dump.
2976 (decide_about_value): Likewise.
2977 (ipcp_propagate_stage): Do not calculate max_new_size, just store
2978 orig_overall_size. Adjust dump.
2979 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
2981 2020-01-10 Martin Jambor <mjambor@suse.cz>
2983 * params.opt (param_ipa_max_agg_items): Mark as Optimization
2984 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
2985 instead of param_ipa_max_agg_items.
2986 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
2987 optimization info for the callee.
2989 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
2991 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
2992 markers if debug_inline_points is false.
2994 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2996 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
2998 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
2999 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3000 aarch64-sve-builtins-sve2.h.
3001 (aarch64-sve-builtins-sve2.o): New rule.
3002 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3003 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3004 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3005 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3006 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3007 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3009 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3010 instructions that are handled here.
3011 (@cond_asrd<mode>): Generalize to...
3012 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3013 (*cond_asrd<mode>_2): Generalize to...
3014 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3015 (*cond_asrd<mode>_z): Generalize to...
3016 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3017 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3018 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3019 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3020 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3022 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3023 (@aarch64_scatter_stnt<mode>): Likewise.
3024 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3025 (@aarch64_mul_lane_<mode>): Likewise.
3026 (@aarch64_sve_suqadd<mode>_const): Likewise.
3027 (*<sur>h<addsub><mode>): Generalize to...
3028 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3030 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3031 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3032 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3033 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3034 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3035 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3036 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3037 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3038 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3039 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3040 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3041 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3042 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3043 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3044 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3045 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3046 (@aarch64_sve2_xar<mode>): Likewise.
3047 (@aarch64_sve2_bcax<mode>): Likewise.
3048 (*aarch64_sve2_eor3<mode>): Rename to...
3049 (@aarch64_sve2_eor3<mode>): ...this.
3050 (@aarch64_sve2_bsl<mode>): New expander.
3051 (@aarch64_sve2_nbsl<mode>): Likewise.
3052 (@aarch64_sve2_bsl1n<mode>): Likewise.
3053 (@aarch64_sve2_bsl2n<mode>): Likewise.
3054 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3055 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3056 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3057 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3058 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3059 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3060 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3061 (<su>mull<bt><Vwide>): Generalize to...
3062 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3064 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3065 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3066 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3067 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3068 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3069 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3070 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3071 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3072 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3073 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3074 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3075 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3076 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3077 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3078 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3079 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3080 (<SHRNB:r>shrnb<mode>): Generalize to...
3081 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3083 (<SHRNT:r>shrnt<mode>): Generalize to...
3084 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3086 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3087 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3088 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3089 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3090 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3091 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3092 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3093 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3094 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3095 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3096 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3097 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3098 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3099 (@aarch64_sve2_cvtnt<mode>): Likewise.
3100 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3101 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3102 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3103 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3104 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3105 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3106 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3107 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3108 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3109 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3110 (@aarch64_sve2_pmul<mode>): Likewise.
3111 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3112 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3113 (@aarch64_sve2_tbl2<mode>): Likewise.
3114 (@aarch64_sve2_tbx<mode>): Likewise.
3115 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3116 (@aarch64_sve2_histcnt<mode>): Likewise.
3117 (@aarch64_sve2_histseg<mode>): Likewise.
3118 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3119 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3120 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3121 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3122 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3123 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3124 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3125 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3126 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3127 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3128 (SVE2_PMULL_PAIR_I): New mode iterators.
3129 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3130 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3131 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3132 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3133 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3134 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3135 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3136 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3137 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3138 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3139 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3140 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3141 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3142 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3143 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3144 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3145 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3146 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3147 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3148 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3149 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3150 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3151 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3152 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3153 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3154 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3155 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3156 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3157 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3158 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3159 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3161 (VNARROW, Ventype): New mode attributes.
3162 (Vewtype): Handle VNx2DI. Fix typo in comment.
3163 (VDOUBLE): New mode attribute.
3164 (sve_lane_con): Handle VNx8HI.
3165 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3166 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3167 (sve_int_op, sve_int_op_rev): Handle the above codes.
3168 (sve_pred_int_rhs2_operand): Likewise.
3169 (MULLBT, SHRNB, SHRNT): Delete.
3170 (SVE_INT_SHIFT_IMM): New int iterator.
3171 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3172 and UNSPEC_WHILEHS for TARGET_SVE2.
3173 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3174 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3175 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3176 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3177 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3178 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3179 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3180 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3181 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3182 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3183 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3184 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3185 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3186 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3187 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3188 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3189 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3190 (optab): Handle the new unspecs.
3191 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3193 (lr): Handle the new unspecs.
3195 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3196 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3197 (sve_int_qsub_op): New int attributes.
3198 (sve_fp_op, rot): Handle the new unspecs.
3199 * config/aarch64/aarch64-sve-builtins.h
3200 (function_resolver::require_matching_pointer_type): Declare.
3201 (function_resolver::resolve_unary): Add an optional boolean argument.
3202 (function_resolver::finish_opt_n_resolution): Add an optional
3203 type_suffix_index argument.
3204 (gimple_folder::redirect_call): Declare.
3205 (gimple_expander::prepare_gather_address_operands): Add an optional
3207 * config/aarch64/aarch64-sve-builtins.cc: Include
3208 aarch64-sve-builtins-sve2.h.
3209 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3210 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3211 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3212 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3213 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3214 (TYPES_s_unsigned): Likewise.
3215 (TYPES_s_integer): Use TYPES_s_unsigned.
3216 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3217 (TYPES_sd_integer): Use them.
3218 (TYPES_d_unsigned): New macro.
3219 (TYPES_d_integer): Use it.
3220 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3221 (TYPES_cvt_narrow): Likewise.
3222 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3223 (preds_mx): New variable.
3224 (function_builder::add_overloaded_function): Allow the new feature
3225 set to be more restrictive than the original one.
3226 (function_resolver::infer_pointer_type): Remove qualifiers from
3227 the pointer type before printing it.
3228 (function_resolver::require_matching_pointer_type): New function.
3229 (function_resolver::resolve_sv_displacement): Handle functions
3230 that don't support 32-bit vector indices or svint32_t vector offsets.
3231 (function_resolver::finish_opt_n_resolution): Take the inferred type
3232 as a separate argument.
3233 (function_resolver::resolve_unary): Optionally treat all forms in
3234 the same way as normal merging functions.
3235 (gimple_folder::redirect_call): New function.
3236 (function_expander::prepare_gather_address_operands): Add an argument
3237 that says whether scaled forms are available. If they aren't,
3238 handle scaling of vector indices and don't add the extension and
3240 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3241 fall back to using cond_* instead.
3242 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3243 Split out the member variables into...
3244 (rtx_code_function_base): ...this new base class.
3245 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3246 (unspec_based_function): Split out the member variables into...
3247 (unspec_based_function_base): ...this new base class.
3248 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3249 (unspec_based_function_exact_insn): New class.
3250 (unspec_based_add_function, unspec_based_add_lane_function)
3251 (unspec_based_lane_function, unspec_based_pred_function)
3252 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3253 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3254 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3256 (unspec_based_fused_function): New class.
3257 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3258 (unspec_based_fused_lane_function): New class.
3259 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3261 (CODE_FOR_MODE1): New macro.
3262 (fixed_insn_function): New class.
3263 (while_comparison): Likewise.
3264 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3265 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3266 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3267 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3268 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3269 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3270 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3271 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3272 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3273 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3274 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3275 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3276 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3277 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3278 Also add an initial argument for unary_convert_narrowt, regardless
3279 of the predication type.
3280 (build_32_64): Allow loads and stores to specify MODE_none.
3281 (build_sv_index64, build_sv_uint_offset): New functions.
3282 (long_type_suffix): New function.
3283 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3284 (binary_imm_long_base, load_gather_sv_base): Likewise.
3285 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3286 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3287 (unary_narrowb_base, unary_narrowt_base): Likewise.
3288 (binary_long_lane_def, binary_long_lane): New shape.
3289 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3290 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3291 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3292 (binary_to_uint_def, binary_to_uint): Likewise.
3293 (binary_wide_def, binary_wide): Likewise.
3294 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3295 (compare_def, compare): Likewise.
3296 (compare_ptr_def, compare_ptr): Likewise.
3297 (load_ext_gather_index_restricted_def,
3298 load_ext_gather_index_restricted): Likewise.
3299 (load_ext_gather_offset_restricted_def,
3300 load_ext_gather_offset_restricted): Likewise.
3301 (load_gather_sv_def): Inherit from load_gather_sv_base.
3302 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3303 (shift_left_imm_def, shift_left_imm): Likewise.
3304 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3305 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3306 (store_scatter_index_restricted_def,
3307 store_scatter_index_restricted): Likewise.
3308 (store_scatter_offset_restricted_def,
3309 store_scatter_offset_restricted): Likewise.
3310 (tbl_tuple_def, tbl_tuple): Likewise.
3311 (ternary_long_lane_def, ternary_long_lane): Likewise.
3312 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3313 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3314 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3315 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3316 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3317 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3318 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3319 (ternary_uint_def, ternary_uint): Likewise.
3320 (unary_convert): Fix typo in comment.
3321 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3322 (unary_long_def, unary_long): Likewise.
3323 (unary_narrowb_def, unary_narrowb): Likewise.
3324 (unary_narrowt_def, unary_narrowt): Likewise.
3325 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3326 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3327 (unary_to_int_def, unary_to_int): Likewise.
3328 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3329 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3330 (svasrd_impl): Delete.
3331 (svcadd_impl::expand): Handle integer operations too.
3332 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3333 new functions to derive the unspec numbers.
3334 (svmla_svmls_lane_impl): Replace with...
3335 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3336 integer operations too.
3337 (svwhile_impl): Rename to...
3338 (svwhilelx_impl): ...this and inherit from while_comparison.
3339 (svasrd): Use unspec_based_function.
3340 (svmla_lane): Use svmla_lane_impl.
3341 (svmls_lane): Use svmls_lane_impl.
3342 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3343 (svwhilele, svwhilelt): Use svwhilelx_impl.
3344 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3345 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3346 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3347 * config/aarch64/aarch64-sve-builtins.def: Include
3348 aarch64-sve-builtins-sve2.def.
3350 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3352 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3353 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3354 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3355 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3356 immediates as well as vector ones.
3357 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3358 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3359 (aarch64_sve_qsub_immediate): Update calls accordingly.
3361 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3363 * config/aarch64/aarch64-sve2.md: Add banner comments.
3364 (<su>mulh<r>s<mode>3): Move further up file.
3365 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3366 (*aarch64_sve2_sra<mode>): Move further down file.
3367 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3369 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3371 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3373 (while_optab_cmp): Handle them.
3374 * config/aarch64/aarch64-sve.md
3375 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3376 and add a "@" marker.
3377 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3378 instead of gen_aarch64_sve2_while_ptest.
3379 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3381 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3383 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3384 (UNSPEC_WHILELE): ...this.
3385 (UNSPEC_WHILE_LO): Rename to...
3386 (UNSPEC_WHILELO): ...this.
3387 (UNSPEC_WHILE_LS): Rename to...
3388 (UNSPEC_WHILELS): ...this.
3389 (UNSPEC_WHILE_LT): Rename to...
3390 (UNSPEC_WHILELT): ...this.
3391 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3392 (cmp_op, while_optab_cmp): Likewise.
3393 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3394 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3395 (svwhilelt): Likewise.
3397 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3399 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3400 (unary_to_uint): Define.
3401 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3402 (unary_count): Rename to...
3403 (unary_to_uint_def, unary_to_uint): ...this.
3404 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3406 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3408 * config/aarch64/aarch64-sve-builtins-functions.h
3409 (code_for_mode_function): New class.
3410 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3411 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3412 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3413 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3414 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3416 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3418 * config/aarch64/iterators.md (addsub): New code attribute.
3419 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3421 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3422 in the asm string and attributes. Fix indentation.
3423 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3425 (@aarch64_sve_<optab><mode>): ...this.
3426 * config/aarch64/aarch64-sve-builtins.h
3427 (function_expander::expand_signed_unpred_op): Delete.
3428 * config/aarch64/aarch64-sve-builtins.cc
3429 (function_expander::expand_signed_unpred_op): Likewise.
3430 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3431 try using code_for_aarch64_sve instead.
3432 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3433 (svqsub_impl): Likewise.
3434 (svqadd, svqsub): Use rtx_code_function instead.
3436 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3438 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3439 (HADDSUB, sur, addsub): Remove them.
3441 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3443 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3444 functions the same way, rather than singling out those that
3445 aren't mapped directly to optabs.
3447 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3449 * target.def (compatible_vector_types_p): New target hook.
3450 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3451 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3452 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3453 * doc/tm.texi: Regenerate.
3454 * gimple-expr.c: Include target.h.
3455 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3456 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3458 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3459 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3460 Use the original predicate if it already has a suitable type.
3462 2020-01-09 Martin Jambor <mjambor@suse.cz>
3464 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3465 resolve_speculation and redirect_call_stmt_to_callee static. Change
3466 return type of set_call_stmt to cgraph_edge *.
3467 * auto-profile.c (afdo_indirect_call): Adjust call to
3468 redirect_call_stmt_to_callee.
3469 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3470 make the this pointer explicit, adjust self-recursive calls and the
3471 call top make_direct. Return the resulting edge.
3472 (cgraph_edge::remove): Make this pointer explicit.
3473 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3474 (cgraph_edge::make_direct): Likewise, adjust call to
3475 resolve_speculation.
3476 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3477 call to set_call_stmt.
3478 (cgraph_update_edges_for_call_stmt_node): Update call to
3479 set_call_stmt and remove.
3480 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3481 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3482 (cgraph_node::create_edge_including_clones): Moved "first" definition
3483 of edge to the block where it was used. Adjusted calls to
3485 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3486 cgraph_edge::remove.
3487 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3488 make_direct and redirect_call_stmt_to_callee.
3489 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3490 resolve_speculation and make_direct.
3491 * ipa-inline-transform.c (inline_transform): Adjust call to
3492 redirect_call_stmt_to_callee.
3493 (check_speculations_1):: Adjust call to resolve_speculation.
3494 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3495 resolve-speculation.
3496 (inline_small_functions): Adjust call to resolve_speculation.
3497 (ipa_inline): Likewise.
3498 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3500 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3501 safe with regards to edge removal, adjust calls to
3502 redirect_call_stmt_to_callee.
3503 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3504 and redirect_call_stmt_to_callee.
3505 * multiple_target.c (create_dispatcher_calls): Adjust call to
3506 redirect_call_stmt_to_callee
3507 (redirect_to_specific_clone): Likewise.
3508 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3509 Adjust calls to cgraph_edge::remove.
3510 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3511 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3512 (expand_call_inline): Adjust call to cgraph_edge::remove.
3514 2020-01-09 Martin Liska <mliska@suse.cz>
3516 * params.opt: Set Optimization for
3517 param_max_speculative_devirt_maydefs.
3519 2020-01-09 Martin Sebor <msebor@redhat.com>
3523 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3525 2020-01-09 Martin Liska <mliska@suse.cz>
3527 * auto-profile.c (auto_profile): Use opt_for_fn
3529 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3530 (propagate_vals_across_arith_jfunc): Likewise.
3531 (hint_time_bonus): Likewise.
3532 (incorporate_penalties): Likewise.
3533 (good_cloning_opportunity_p): Likewise.
3534 (perform_estimation_of_a_value): Likewise.
3535 (estimate_local_effects): Likewise.
3536 (ipcp_propagate_stage): Likewise.
3537 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3538 (set_switch_stmt_execution_predicate): Likewise.
3539 (analyze_function_body): Likewise.
3540 * ipa-inline-analysis.c (offline_size): Likewise.
3541 * ipa-inline.c (early_inliner): Likewise.
3542 * ipa-prop.c (ipa_analyze_node): Likewise.
3543 (ipcp_transform_function): Likewise.
3544 * ipa-sra.c (process_scan_results): Likewise.
3545 (ipa_sra_summarize_function): Likewise.
3546 * params.opt: Rename ipcp-unit-growth to
3547 ipa-cp-unit-growth. Add Optimization for various
3548 IPA-related parameters.
3550 2020-01-09 Richard Biener <rguenther@suse.de>
3553 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3555 2020-01-09 Richard Biener <rguenther@suse.de>
3557 PR tree-optimization/93040
3558 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3560 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3562 * common/config/avr/avr-common.c (avr_option_optimization_table)
3563 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3565 2020-01-09 Martin Liska <mliska@suse.cz>
3567 * cgraphclones.c (symbol_table::materialize_all_clones):
3568 Use cgraph_node::dump_name.
3570 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3573 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3574 output_operand_lossage instead of gcc_unreachable.
3575 * doc/md.texi (riscv f constraint): Fix typo.
3578 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3579 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3580 CONST_SCALAR_INT_P instead of CONST_INT_P.
3581 (*subv<mode>4_1): Rename to ...
3582 (subv<mode>4_1): ... this.
3583 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3584 define_insn_and_split patterns.
3585 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3588 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3590 * vec.c (class selftest::count_dtor): New class.
3591 (selftest::test_auto_delete_vec): New test.
3592 (selftest::vec_c_tests): Call it.
3593 * vec.h (class auto_delete_vec): New class template.
3594 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3596 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3598 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3600 2020-01-08 Jim Wilson <jimw@sifive.com>
3602 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3603 use of TLS_MODEL_LOCAL_EXEC when not pic.
3605 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3607 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3610 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3613 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3614 *stack_protect_set_3 peephole2): Also check that the second
3615 insns source is general_operand.
3618 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3619 predicate for output operand instead of register_operand.
3620 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3621 memory destination and non-memory operands[2].
3623 2020-01-08 Martin Liska <mliska@suse.cz>
3625 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3626 ::dump_asm_name instead of (::name or ::asm_name).
3627 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3628 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3629 (analyze_functions): Likewise.
3630 (expand_all_functions): Likewise.
3631 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3632 (propagate_bits_across_jump_function): Likewise.
3633 (dump_profile_updates): Likewise.
3634 (ipcp_store_bits_results): Likewise.
3635 (ipcp_store_vr_results): Likewise.
3636 * ipa-devirt.c (dump_targets): Likewise.
3637 * ipa-fnsummary.c (analyze_function_body): Likewise.
3638 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3639 (process_hsa_functions): Likewise.
3640 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3641 (set_alias_uids): Likewise.
3642 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3643 * ipa-inline.c (recursive_inlining): Likewise.
3644 (inline_to_all_callers_1): Likewise.
3645 (ipa_inline): Likewise.
3646 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3647 (ipa_propagate_frequency): Likewise.
3648 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3649 (remove_described_reference): Likewise.
3650 * ipa-pure-const.c (worse_state): Likewise.
3651 (check_retval_uses): Likewise.
3652 (analyze_function): Likewise.
3653 (propagate_pure_const): Likewise.
3654 (propagate_nothrow): Likewise.
3655 (dump_malloc_lattice): Likewise.
3656 (propagate_malloc): Likewise.
3657 (pass_local_pure_const::execute): Likewise.
3658 * ipa-visibility.c (optimize_weakref): Likewise.
3659 (function_and_variable_visibility): Likewise.
3660 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3661 (ipa_discover_variable_flags): Likewise.
3662 * lto-streamer-out.c (output_function): Likewise.
3663 (output_constructor): Likewise.
3664 * tree-inline.c (copy_bb): Likewise.
3665 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3666 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3668 2020-01-08 Richard Biener <rguenther@suse.de>
3671 * tree-eh.c (sink_clobbers): Update virtual operands for
3672 the first and last stmt only. Add a dry-run capability.
3673 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3674 after CFG manipulations and in RPO order to catch all
3675 secondary opportunities reliably.
3677 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3680 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3682 2019-01-08 Richard Biener <rguenther@suse.de>
3685 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3686 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3687 virtual operand, also updating SSA use.
3688 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3689 Update stmt after resetting virtual operand.
3690 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3691 * gimple-iterator.c (gsi_remove): When not removing the stmt
3692 permanently do not delink immediate uses or mark the stmt modified.
3694 2020-01-08 Martin Liska <mliska@suse.cz>
3696 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3697 (ipa_call_context::estimate_size_and_time): Likewise.
3698 (inline_analyze_function): Likewise.
3700 2020-01-08 Martin Liska <mliska@suse.cz>
3702 * cgraph.c (cgraph_node::dump): Use systematically
3705 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3707 Add -nodevicespecs option for avr.
3710 * config/avr/avr.opt (-nodevicespecs): New driver option.
3711 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3712 "-specs=device-specs/..." if that option is not set.
3713 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3715 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3717 Implement 64-bit double functions for avr.
3720 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3721 --with-double-comparison.
3722 * doc/install.texi: Document them.
3723 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3724 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3725 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3726 * doc/invoke.texi (AVR Built-in Macros): Document them.
3727 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3728 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3729 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3731 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3734 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3735 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3736 when only building rm-profile multilibs.
3738 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3741 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3742 lattice for a value to check.
3743 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3744 finite propagation in self-recursive scc.
3746 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3748 * ipa-inline.c (caller_growth_limits): Restore the AND.
3750 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3752 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3753 (VEC_ALLREG_ALT): New iterator.
3754 (VEC_ALLREG_INT_MODE): New iterator.
3755 (VCMP_MODE): New iterator.
3756 (VCMP_MODE_INT): New iterator.
3757 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3758 (vec_cmp<u>v64qidi): New define_expand.
3759 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3760 (vec_cmpu<mode>di_exec): New define_expand.
3761 (vec_cmp<u>v64qidi_exec): New define_expand.
3762 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3763 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3764 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3765 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3766 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3767 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3768 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3769 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3770 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3771 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3773 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3774 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3776 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3778 * config/gcn/constraints.md (DA): Update description and match.
3780 (Db): New constraint.
3781 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3783 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3784 Implement 'Db' mixed immediate type.
3785 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3786 (addcv64si3_dup<exec_vcc>): Delete.
3787 (subcv64si3<exec_vcc>): Rework constraints.
3788 (addv64di3): Rework constraints.
3789 (addv64di3_exec): Rework constraints.
3790 (subv64di3): Rework constraints.
3791 (addv64di3_dup): Delete.
3792 (addv64di3_dup_exec): Delete.
3793 (addv64di3_zext): Rework constraints.
3794 (addv64di3_zext_exec): Rework constraints.
3795 (addv64di3_zext_dup): Rework constraints.
3796 (addv64di3_zext_dup_exec): Rework constraints.
3797 (addv64di3_zext_dup2): Rework constraints.
3798 (addv64di3_zext_dup2_exec): Rework constraints.
3799 (addv64di3_sext_dup2): Rework constraints.
3800 (addv64di3_sext_dup2_exec): Rework constraints.
3802 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3804 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3805 existing target checks.
3807 2020-01-07 Richard Biener <rguenther@suse.de>
3809 * doc/install.texi: Bump minimal supported MPC version.
3811 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3813 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3814 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3815 * langhooks.c: Include stor-layout.h.
3816 (lhd_simulate_enum_decl): New function.
3817 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3818 handle_arm_sve_h for the LTO frontend.
3819 (register_vector_type): Cope with null returns from pushdecl.
3821 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3823 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3824 (aarch64_sve::nvectors_if_data_type): Replace with...
3825 (aarch64_sve::builtin_type_p): ...this.
3826 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3827 (find_vector_type): Delete.
3828 (add_sve_type_attribute): New function.
3829 (lookup_sve_type_attribute): Likewise.
3830 (register_builtin_types): Add an "SVE type" attribute to each type.
3831 (register_tuple_type): Likewise.
3832 (svbool_type_p, nvectors_if_data_type): Delete.
3833 (mangle_builtin_type): Use lookup_sve_type_attribute.
3834 (builtin_type_p): Likewise. Add an overload that returns the
3835 number of constituent vector and predicate registers.
3836 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3837 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3838 instead of aarch64_sve_argument_p.
3839 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3840 (aarch64_pass_by_reference): Likewise.
3841 (aarch64_function_value_1): Likewise.
3842 (aarch64_return_in_memory): Likewise.
3843 (aarch64_layout_arg): Likewise.
3845 2020-01-07 Jakub Jelinek <jakub@redhat.com>
3847 PR tree-optimization/93156
3848 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3849 least significant bit is always clear.
3851 PR tree-optimization/93118
3852 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
3853 simplifier with two intermediate conversions.
3855 2020-01-07 Martin Liska <mliska@suse.cz>
3857 * params.opt: Add Optimization for various parameters.
3859 2020-01-07 Martin Liska <mliska@suse.cz>
3862 * doc/extend.texi: Explain cloning for target_clone
3865 2020-01-07 Martin Liska <mliska@suse.cz>
3867 PR tree-optimization/92860
3868 * common.opt: Make in Optimization option
3869 as it is affected by -O0, which is an Optimization
3871 * tree-inline.c (tree_inlinable_function_p):
3872 Use opt_for_fn for warn_inline.
3873 (expand_call_inline): Likewise.
3875 2020-01-07 Martin Liska <mliska@suse.cz>
3877 PR tree-optimization/92860
3878 * common.opt: Make flag_ree as optimization
3881 2020-01-07 Martin Liska <mliska@suse.cz>
3883 PR optimization/92860
3884 * params.opt: Mark param_min_crossjump_insns with Optimization
3887 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
3889 * ipa-inline-analysis.c (estimate_growth): Fix typo.
3890 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3892 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
3894 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3895 helper function to return the valid addressing formats for a given
3896 hard register and mode.
3897 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3899 * config/rs6000/constraints.md (Q constraint): Update
3901 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3904 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3905 Use 'Q' for doing vector extract from memory.
3906 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
3908 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
3909 doing vector extract from memory.
3910 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
3911 extract from memory.
3913 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
3914 for the offset being 34-bits when -mcpu=future is used.
3916 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
3918 * config/pa/pa.md: Revert change to use ordered_comparison_operator
3919 instead of cmpib_comparison_operator in cmpib patterns.
3920 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
3921 of cmpib_comparison_operator. Revise comment.
3923 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3925 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
3926 in an IFN_DIV_POW2 node to be equal.
3928 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3930 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
3931 (vect_check_scalar_mask): ...this.
3932 (vectorizable_store, vectorizable_load): Update call accordingly.
3933 (vectorizable_call): Use vect_check_scalar_mask to check the mask
3934 argument in calls to conditional internal functions.
3936 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3938 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
3939 '0' matching inputs.
3940 (subv64di3_exec): Likewise.
3942 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
3944 * config/mips/mips.c (vr4130_align_insns): Fix typo.
3945 * doc/md.texi (movstr): Likewise.
3947 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3949 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
3952 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3954 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
3956 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
3957 to a temporary file and use move-if-change to update the real
3958 file where necessary.
3960 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3962 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
3963 rather than Upa for CPY /M.
3965 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3967 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
3970 2020-01-06 Martin Liska <mliska@suse.cz>
3972 PR tree-optimization/92860
3973 * params.opt: Mark param_max_combine_insns with Optimization
3976 2020-01-05 Jakub Jelinek <jakub@redhat.com>
3979 * config/i386/i386.md (SWIDWI): New mode iterator.
3980 (DWI, dwi): Add TImode variants.
3981 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
3982 <general_hilo_operand> instead of <general_operand>. Use
3983 CONST_SCALAR_INT_P instead of CONST_INT_P.
3984 (*addv<mode>4_1): Rename to ...
3985 (addv<mode>4_1): ... this.
3986 (QWI): New mode attribute.
3987 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3988 define_insn_and_split patterns.
3989 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3991 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
3992 <general_hilo_operand> instead of <general_operand>.
3993 (*addcarry<mode>_1): New define_insn.
3994 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
3996 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
3998 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
3999 Use "call" instead of "set".
4001 2020-01-03 Martin Jambor <mjambor@suse.cz>
4004 * ipa-cp.c (print_all_lattices): Skip functions without info.
4006 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4009 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4010 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4011 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4012 for 'e' simd clones.
4015 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4017 (mprefer-vector-width=): Add Save.
4018 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4019 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4020 (ix86_debug_options, ix86_function_specific_print): Adjust
4021 ix86_target_string callers.
4022 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4023 (ix86_valid_target_attribute_tree): Likewise.
4024 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4025 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4026 ix86_target_string caller.
4029 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4030 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4031 instead of gen_int_shift_amount + convert_modes.
4033 PR rtl-optimization/93088
4034 * loop-iv.c (find_single_def_src): Punt after looking through
4035 128 reg copies for regs with single definitions. Move definitions
4038 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4040 * config/arm/arm-c.c (arm_cpu_builtins): Define
4041 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4042 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4043 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4044 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4045 * config/arm/arm-tables.opt: Regenerated.
4046 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4047 arm_arch_i8mm and arm_arch_bf16 when enabled.
4048 * config/arm/arm.h (TARGET_I8MM): New macro.
4049 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4050 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4051 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4052 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4053 (v8_6_a_simd_variants): New.
4054 (v8_*_a_simd_variants): Add i8mm and bf16.
4055 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4057 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4060 * predict.c (compute_function_frequency): Don't call
4061 warn_function_cold on functions that already have cold attribute.
4063 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4066 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4067 COMDAT group function labels in .data.rel.ro.local section.
4068 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4071 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4072 comparison_operator in B and S integer comparisons. Likewise, use
4073 ordered_comparison_operator instead of cmpib_comparison_operator in
4075 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4077 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4079 Update copyright years.
4081 * gcc.c (process_command): Update copyright notice dates.
4082 * gcov-dump.c (print_version): Ditto.
4083 * gcov.c (print_version): Ditto.
4084 * gcov-tool.c (print_version): Ditto.
4085 * gengtype.c (create_file): Ditto.
4086 * doc/cpp.texi: Bump @copying's copyright year.
4087 * doc/cppinternals.texi: Ditto.
4088 * doc/gcc.texi: Ditto.
4089 * doc/gccint.texi: Ditto.
4090 * doc/gcov.texi: Ditto.
4091 * doc/install.texi: Ditto.
4092 * doc/invoke.texi: Ditto.
4094 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4096 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4099 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4101 PR tree-optimization/93098
4102 * match.pd (popcount): For shift amounts, use integer_onep
4103 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4104 tests. Make sure that precision is power of two larger than or equal
4105 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4106 instead of ULL suffixed constants. Formatting fixes.
4108 Copyright (C) 2020 Free Software Foundation, Inc.
4110 Copying and distribution of this file, with or without modification,
4111 are permitted in any medium without royalty provided the copyright
4112 notice and this notice are preserved.