1 2020-02-15 David Malcolm <dmalcolm@redhat.com>
2 Bernd Edlinger <bernd.edlinger@hotmail.de>
6 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
7 * configure.ac (--with-diagnostics-urls): New configuration
8 option, based on --with-diagnostics-color.
9 (DIAGNOSTICS_URLS_DEFAULT): New define.
10 * config.h: Regenerate.
11 * configure: Regenerate.
12 * diagnostic.c (diagnostic_urls_init): Handle -1 for
13 DIAGNOSTICS_URLS_DEFAULT from configure-time
14 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
15 and TERM_URLS environment variable.
16 * diagnostic-url.h (diagnostic_url_format): New enum type.
17 (diagnostic_urls_enabled_p): rename to...
18 (determine_url_format): ... this, and change return type.
19 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
20 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
21 the linux console, and mingw.
22 (diagnostic_urls_enabled_p): rename to...
23 (determine_url_format): ... this, and adjust.
24 * pretty-print.h (pretty_printer::show_urls): rename to...
25 (pretty_printer::url_format): ... this, and change to enum.
26 * pretty-print.c (pretty_printer::pretty_printer,
27 pp_begin_url, pp_end_url, test_urls): Adjust.
28 * doc/install.texi (--with-diagnostics-urls): Document the new
30 (--with-diagnostics-color): Document the existing interaction
31 with GCC_COLORS better.
32 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
33 vindex reference. Update description of defaults based on the above.
34 (-fdiagnostics-color): Update description of how -fdiagnostics-color
35 interacts with GCC_COLORS.
37 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
40 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
41 conjunction with TARGET_GNU_TLS in early return.
43 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
45 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
46 the mode is not wider than UNITS_PER_WORD.
48 2020-02-14 Martin Jambor <mjambor@suse.cz>
50 PR tree-optimization/93516
51 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
52 access of the same type as the parent.
53 (propagate_subaccesses_from_lhs): Likewise.
55 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
58 * config/i386/avx512vbmi2intrin.h
59 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
60 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
61 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
62 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
63 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
64 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
65 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
66 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
67 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
68 of lacking a closing parenthesis.
69 * config/i386/avx512vbmi2vlintrin.h
70 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
71 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
72 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
73 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
74 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
75 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
76 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
77 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
78 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
79 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
80 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
81 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
82 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
83 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
84 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
85 _mm_shldi_epi32, _mm_mask_shldi_epi32,
86 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
87 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
89 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
92 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
93 the target function entry.
95 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
97 * common/config/arc/arc-common.c (arc_option_optimization_table):
98 Disable if-conversion step when optimized for size.
100 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
102 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
103 R12-R15 are always in ARCOMPACT16_REGS register class.
104 * config/arc/arc.opt (mq-class): Deprecate.
105 * config/arc/constraint.md ("q"): Remove dependency on mq-class
107 * doc/invoke.texi (mq-class): Update text.
108 * common/config/arc/arc-common.c (arc_option_optimization_table):
111 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
113 * config/arc/arc.c (arc_insn_cost): New function.
114 (TARGET_INSN_COST): Define.
115 * config/arc/arc.md (cost): New attribute.
116 (add_n): Use arc_nonmemory_operand.
117 (ashlsi3_insn): Likewise, also update constraints.
118 (ashrsi3_insn): Likewise.
120 (add_shift): Likewise.
121 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
123 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
125 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
127 (umulsidi_600): Likewise.
129 2020-02-13 Jakub Jelinek <jakub@redhat.com>
132 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
133 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
134 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
135 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
136 pass __A to the builtin followed by __W instead of __A followed by
138 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
139 _mm512_mask_popcnt_epi64): Likewise.
140 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
141 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
142 _mm256_mask_popcnt_epi64): Likewise.
144 PR tree-optimization/93582
145 * fold-const.h (shift_bytes_in_array_left,
146 shift_bytes_in_array_right): Declare.
147 * fold-const.c (shift_bytes_in_array_left,
148 shift_bytes_in_array_right): New function, moved from
149 gimple-ssa-store-merging.c, no longer static.
150 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
151 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
152 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
153 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
154 shift_bytes_in_array.
155 (verify_shift_bytes_in_array): Rename to ...
156 (verify_shift_bytes_in_array_left): ... this. Use
157 shift_bytes_in_array_left instead of shift_bytes_in_array.
158 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
159 instead of verify_shift_bytes_in_array.
160 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
161 / native_interpret_expr where the store covers all needed bits,
162 punt on PDP-endian, otherwise allow all involved offsets and sizes
163 not to be byte-aligned.
166 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
167 use const_0_to_255_operand predicate instead of immediate_operand.
168 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
169 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
170 vgf2p8affineinvqb_<mode><mask_name>,
171 vgf2p8affineqb_<mode><mask_name>): Drop mode from
172 const_0_to_255_operand predicated operands.
174 2020-02-12 Jeff Law <law@redhat.com>
176 * config/h8300/h8300.md (comparison shortening peepholes): Use
177 a mode iterator to merge the HImode and SImode peepholes.
179 2020-02-12 Jakub Jelinek <jakub@redhat.com>
182 * real.c (is_even): Make static. Function comment fix.
183 (is_halfway_below): Make static, don't assert R is not inf/nan,
184 instead return false for those. Small formatting fixes.
186 2020-02-12 Martin Sebor <msebor@redhat.com>
189 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
190 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
191 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
192 (strlen_check_and_optimize_call): Adjust callee name.
194 2020-02-12 Jeff Law <law@redhat.com>
196 * config/h8300/h8300.md (comparison shortening peepholes): Drop
197 (and (xor)) variant. Combine other two into single peephole.
199 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
201 PR rtl-optimization/93565
202 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
204 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
206 * config/aarch64/aarch64-simd.md
207 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
208 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
209 generating separate ADDV and zero_extend patterns.
210 * config/aarch64/iterators.md (VDQV_E): New iterator.
212 2020-02-12 Jeff Law <law@redhat.com>
214 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
215 expanders, splits, etc.
216 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
217 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
218 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
219 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
220 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
221 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
223 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
225 2020-02-12 Jakub Jelinek <jakub@redhat.com>
228 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
229 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
230 TARGET_AVX512DQ from condition.
231 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
232 instead of <mask_mode512bit_condition> in condition. If
233 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
235 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
238 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
241 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
243 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
245 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
246 where strlen is more legible.
247 (rs6000_builtin_vectorized_libmass): Ditto.
248 (rs6000_print_options_internal): Ditto.
250 2020-02-11 Martin Sebor <msebor@redhat.com>
252 PR tree-optimization/93683
253 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
255 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
257 * config/rs6000/predicates.md (cint34_operand): Rename the
258 -mprefixed-addr option to be -mprefixed.
259 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
260 the -mprefixed-addr option to be -mprefixed.
261 (OTHER_FUTURE_MASKS): Likewise.
262 (POWERPC_MASKS): Likewise.
263 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
264 the -mprefixed-addr option to be -mprefixed. Change error
265 messages to refer to -mprefixed.
266 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
268 (rs6000_legitimate_offset_address_p): Likewise.
269 (rs6000_mode_dependent_address): Likewise.
270 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
271 "-mprefixed" for target attributes and pragmas.
272 (address_to_insn_form): Rename the -mprefixed-addr option to be
274 (rs6000_adjust_insn_length): Likewise.
275 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
276 -mprefixed-addr option to be -mprefixed.
277 (ASM_OUTPUT_OPCODE): Likewise.
278 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
279 -mprefixed-addr option to be -mprefixed.
280 * config/rs6000/rs6000.opt (-mprefixed): Rename the
281 -mprefixed-addr option to be prefixed. Change the option from
282 being undocumented to being documented.
283 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
284 -mprefixed option. Update the -mpcrel documentation to mention
287 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
289 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
290 including FIRST_PSEUDO_REGISTER - 1.
291 * ira-color.c (print_hard_reg_set): Ditto.
293 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
295 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
296 (USTERNOP_QUALIFIERS): New define.
297 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
298 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
299 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
300 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
301 * config/arm/arm_neon.h (vusdot_s32): New.
302 (vusdot_lane_s32): New.
303 (vusdotq_lane_s32): New.
304 (vsudot_lane_s32): New.
305 (vsudotq_lane_s32): New.
306 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
307 * config/arm/iterators.md (DOTPROD_I8MM): New.
308 (sup, opsuffix): Add <us/su>.
309 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
310 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
312 2020-02-11 Richard Biener <rguenther@suse.de>
314 PR tree-optimization/93661
315 PR tree-optimization/93662
316 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
318 * tree-sra.c (get_access_for_expr): Likewise.
320 2020-02-10 Jakub Jelinek <jakub@redhat.com>
323 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
324 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
325 Change condition from TARGET_AVX2 to TARGET_AVX.
327 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
330 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
333 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
335 Try to generate zero-based comparisons.
336 * config/cris/cris.c (cris_reduce_compare): New function.
337 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
338 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
339 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
341 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
344 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
345 in Thumb state and also as a destination in Arm state. Add T16
348 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
350 * md.texi (Define Subst): Match closing paren in example.
352 2020-02-10 Jakub Jelinek <jakub@redhat.com>
356 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
357 arguments of strncmp.
359 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
362 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
363 but different source value.
364 (adjust_callers_for_value_intersection): New function.
365 (gather_edges_for_value): Adjust order of callers to let a
366 non-self-recursive caller be the first element.
367 (self_recursive_pass_through_p): Add a new parameter "simple", and
368 check generalized self-recursive pass-through jump function.
369 (self_recursive_agg_pass_through_p): Likewise.
370 (find_more_scalar_values_for_callers_subset): Compute value from
371 pass-through jump function for self-recursive.
372 (intersect_with_plats): Cleanup previous implementation code for value
373 itersection with self-recursive call edge.
374 (intersect_with_agg_replacements): Likewise.
375 (intersect_aggregates_with_edge): Deduce value from pass-through jump
376 function for self-recursive call edge. Cleanup previous implementation
377 code for value intersection with self-recursive call edge.
378 (decide_whether_version_node): Remove dead callers and adjust order
379 to let a non-self-recursive caller be the first element.
381 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
383 * recog.c: Move pass_split_before_sched2 code in front of
384 pass_split_before_regstack.
385 (pass_data_split_before_sched2): Rename pass to split3 from split4.
386 (pass_data_split_before_regstack): Rename pass to split4 from split3.
387 (rest_of_handle_split_before_sched2): Remove.
388 (pass_split_before_sched2::execute): Unconditionally call
390 (enable_split_before_sched2): New function.
391 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
392 (pass_split_before_regstack::gate): Ditto.
393 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
394 Update name check for renamed split4 pass.
395 * config/sh/sh.c (register_sh_passes): Update pass insertion
396 point for renamed split4 pass.
398 2020-02-09 Jakub Jelinek <jakub@redhat.com>
400 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
401 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
402 copying them around between host and target.
404 2020-02-08 Andrew Pinski <apinski@marvell.com>
407 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
408 STRICT_ALIGNMENT also.
410 2020-02-08 Jim Wilson <jimw@sifive.com>
413 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
415 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
416 Jakub Jelinek <jakub@redhat.com>
419 * config/i386/i386.h (CALL_USED_REGISTERS): Make
420 xmm16-xmm31 call-used even in 64-bit ms-abi.
422 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
424 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
425 (simd_ummla, simd_usmmla): Likewise.
426 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
427 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
430 2020-02-07 Richard Biener <rguenther@suse.de>
433 * tree-inline.c (fold_marked_statements): Do a PRE walk,
434 skipping unreachable regions.
435 (optimize_inline_calls): Skip folding stmts when we didn't
438 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
441 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
442 Don't return aggregates with only SFmode and DFmode in SSE
444 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
446 2020-02-07 Jakub Jelinek <jakub@redhat.com>
449 * config/rs6000/rs6000-logue.c
450 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
451 if it fails, move rs into end_addr and retry. Add
452 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
453 the insn pattern doesn't describe well what exactly happens to
457 * config/i386/predicates.md (avx_identity_operand): Remove.
458 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
459 (avx_<castmode><avxsizesuffix>_<castmode>,
460 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
461 a VEC_CONCAT of the operand and UNSPEC_CAST.
462 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
463 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
467 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
468 recog_data.insn if distance_non_agu_define changed it.
470 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
473 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
474 we only had X-FORM (reg+reg) addressing for vectors. Also before
475 ISA 3.0, we only had X-FORM addressing for scalars in the
476 traditional Altivec registers.
478 2020-02-06 <zhongyunde@huawei.com>
479 Vladimir Makarov <vmakarov@redhat.com>
481 PR rtl-optimization/93561
482 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
485 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
487 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
490 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
492 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
493 where the low and the high 32 bits are equal to each other specially,
494 with an rldimi instruction.
496 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
498 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
500 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
502 * config/arm/arm-tables.opt: Regenerate.
504 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
507 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
508 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
509 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
511 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
513 PR rtl-optimization/87763
514 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
516 2020-02-06 Delia Burduv <delia.burduv@arm.com>
518 * config/aarch64/aarch64-simd-builtins.def
519 (bfmlaq): New built-in function.
520 (bfmlalb): New built-in function.
521 (bfmlalt): New built-in function.
522 (bfmlalb_lane): New built-in function.
523 (bfmlalt_lane): New built-in function.
524 * config/aarch64/aarch64-simd.md
525 (aarch64_bfmmlaqv4sf): New pattern.
526 (aarch64_bfmlal<bt>v4sf): New pattern.
527 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
528 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
529 (vbfmlalbq_f32): New intrinsic.
530 (vbfmlaltq_f32): New intrinsic.
531 (vbfmlalbq_lane_f32): New intrinsic.
532 (vbfmlaltq_lane_f32): New intrinsic.
533 (vbfmlalbq_laneq_f32): New intrinsic.
534 (vbfmlaltq_laneq_f32): New intrinsic.
535 * config/aarch64/iterators.md (BF_MLA): New int iterator.
536 (bt): New int attribute.
538 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
540 * config/i386/i386.md (*pushtf): Emit "#" instead of
541 calling gcc_unreachable in insn output.
544 (*pushsf_rex64): Ditto for alternatives other than 1.
545 (*pushsf): Ditto for alternatives other than 1.
547 2020-02-06 Martin Liska <mliska@suse.cz>
549 PR gcov-profile/91971
550 PR gcov-profile/93466
551 * coverage.c (coverage_init): Revert mangling of
552 path into filename. It can lead to huge filename length.
553 Creation of subfolders seem more natural.
555 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
558 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
559 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
560 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
562 2020-02-06 Jakub Jelinek <jakub@redhat.com>
565 * config/i386/predicates.md (avx_identity_operand): New predicate.
566 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
567 define_insn_and_split.
570 * omp-low.c (use_pointer_for_field): For nested constructs, also
571 look for map clauses on target construct.
572 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
573 taskreg_nesting_level.
576 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
577 shared clause, call omp_notice_variable on outer context if any.
579 2020-02-05 Jason Merrill <jason@redhat.com>
582 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
583 non-zero address even if weak and not yet defined.
585 2020-02-05 Martin Sebor <msebor@redhat.com>
587 PR tree-optimization/92765
588 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
589 * tree-ssa-strlen.c (compute_string_length): Remove.
590 (determine_min_objsize): Remove.
591 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
592 Avoid using type size as the upper bound on string length.
593 (handle_builtin_string_cmp): Add an argument. Adjust.
594 (strlen_check_and_optimize_call): Pass additional argument to
595 handle_builtin_string_cmp.
597 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
599 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
600 (*pushdi2_rex64 peephole2): Unconditionally split after
602 (*ashl<mode>3_doubleword): Ditto.
603 (*<shift_insn><mode>3_doubleword): Ditto.
605 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
608 * config/rs6000/rs6000.c (get_vector_offset): Fix
610 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
612 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
614 2020-02-05 David Malcolm <dmalcolm@redhat.com>
617 (Special Functions for Debugging the Analyzer): Update description
618 of __analyzer_dump_exploded_nodes.
620 2020-02-05 Jakub Jelinek <jakub@redhat.com>
623 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
624 include sets and not clobbers in the vzeroupper pattern.
625 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
626 the parallel has 17 (64-bit) or 9 (32-bit) elts.
627 (*avx_vzeroupper_1): New define_insn_and_split.
630 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
631 don't run when !optimize.
632 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
635 2020-02-05 Richard Biener <rguenther@suse.de>
638 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
639 checks before matching calls.
641 2020-02-05 Jakub Jelinek <jakub@redhat.com>
643 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
644 function comment typo.
647 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
648 simd_clone_create failed when i == 0, adjust clone->nargs by
651 2020-02-05 Martin Liska <mliska@suse.cz>
654 * doc/invoke.texi: Document that one should
655 not combine ASLR and -fpch.
657 2020-02-04 Richard Biener <rguenther@suse.de>
659 PR tree-optimization/93538
660 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
662 2020-02-04 Richard Biener <rguenther@suse.de>
664 PR tree-optimization/91123
665 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
666 (vn_walk_cb_data::last_vuse): New member.
667 (vn_walk_cb_data::saved_operands): Likewsie.
668 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
669 (vn_walk_cb_data::push_partial_def): Use finish.
670 (vn_reference_lookup_2): Update last_vuse and use finish if
671 we've saved operands.
672 (vn_reference_lookup_3): Use finish and update calls to
673 push_partial_defs everywhere. When translating through
674 memcpy or aggregate copies save off operands and alias-set.
675 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
676 operation for redundant store removal.
678 2020-02-04 Richard Biener <rguenther@suse.de>
680 PR tree-optimization/92819
681 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
682 generating more stmts than before.
684 2020-02-04 Martin Liska <mliska@suse.cz>
686 * config/arm/arm.c (arm_gen_far_branch): Move the function
687 outside of selftests.
689 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
691 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
692 function to adjust PC-relative vector addresses.
693 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
694 handle vectors with PC-relative addresses.
696 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
698 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
700 (hard_reg_and_mode_to_addr_mask): Delete.
701 (rs6000_adjust_vec_address): If the original vector address
702 was REG+REG or REG+OFFSET and the element is not zero, do the add
703 of the elements in the original address before adding the offset
704 for the vector element. Use address_to_insn_form to validate the
705 address using the register being loaded, rather than guessing
706 whether the address is a DS-FORM or DQ-FORM address.
708 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
710 * config/rs6000/rs6000.c (get_vector_offset): New helper function
711 to calculate the offset in memory from the start of a vector of a
712 particular element. Add code to keep the element number in
713 bounds if the element number is variable.
714 (rs6000_adjust_vec_address): Move calculation of offset of the
715 vector element to get_vector_offset.
716 (rs6000_split_vec_extract_var): Do not do the initial AND of
717 element here, move the code to get_vector_offset.
719 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
721 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
724 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
726 * config/rs6000/constraints.md: Improve documentation.
728 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
731 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
732 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
734 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
736 * config.gcc: Remove "carrizo" support.
737 * config/gcn/gcn-opts.h (processor_type): Likewise.
738 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
739 * config/gcn/gcn.opt (gpu_type): Likewise.
740 * config/gcn/t-omp-device: Likewise.
742 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
745 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
746 * config/arm/arm.c (arm_gen_far_branch): New function
748 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
750 2020-02-03 Julian Brown <julian@codesourcery.com>
751 Tobias Burnus <tobias@codesourcery.com>
753 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
755 2020-02-03 Jakub Jelinek <jakub@redhat.com>
758 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
759 valid RTL to sum up the lowest and second lowest bytes of the popcnt
762 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
764 PR rtl-optimization/91333
765 * ira-color.c (struct allocno_color_data): Add member
767 (init_allocno_threads): Set the member up.
768 (bucket_allocno_compare_func): Add compare hard reg
771 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
773 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
775 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
776 * config.in: Regenerated.
777 * configure: Regenerated.
778 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
779 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
780 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
782 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
784 * configure: Regenerate.
786 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
788 PR rtl-optimization/91333
789 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
790 reg preferences comparison up.
792 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
794 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
795 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
796 aarch64-sve-builtins-base.h.
797 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
798 aarch64-sve-builtins-base.cc.
799 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
800 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
802 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
803 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
804 (svcvtnt): New functions.
805 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
806 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
807 (svcvtnt): New functions.
808 (svcvt): Add a form that converts f32 to bf16.
809 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
810 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
812 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
813 Treat B as bfloat16_t.
814 (ternary_bfloat_lane_base): New class.
815 (ternary_bfloat_def): Likewise.
816 (ternary_bfloat): New shape.
817 (ternary_bfloat_lane_def): New class.
818 (ternary_bfloat_lane): New shape.
819 (ternary_bfloat_lanex2_def): New class.
820 (ternary_bfloat_lanex2): New shape.
821 (ternary_bfloat_opt_n_def): New class.
822 (ternary_bfloat_opt_n): New shape.
823 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
824 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
825 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
826 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
827 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
828 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
829 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
830 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
831 the pattern off the narrow mode instead of the wider one.
832 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
833 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
834 (sve_fp_op): Handle them.
835 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
836 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
838 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
840 * config/aarch64/arm_sve.h: Include arm_bf16.h.
841 * config/aarch64/aarch64-modes.def (BF): Move definition before
842 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
843 (SVE_MODES): Handle BF modes.
844 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
846 (aarch64_full_sve_mode): Likewise.
847 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
849 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
850 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
851 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
852 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
854 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
856 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
857 (TYPES_all_data): Add bf16.
858 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
859 (register_tuple_type): Increase buffer size.
860 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
861 (bf16): New type suffix.
862 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
863 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
864 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
865 Change type from all_data to all_arith.
866 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
869 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
870 Matthew Malcomson <matthew.malcomson@arm.com>
871 Richard Sandiford <richard.sandiford@arm.com>
873 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
874 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
875 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
876 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
877 __ARM_FEATURE_MATMUL_FP64.
878 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
879 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
880 be disabled at the same time.
881 (f32mm): New extension.
882 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
883 (AARCH64_FL_F64MM): Bump to the next bit up.
884 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
885 (TARGET_SVE_F64MM): New macros.
886 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
887 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
888 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
889 (UNSPEC_ZIP2Q): New unspeccs.
890 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
891 (optab, sur, perm_insn): Handle the new unspecs.
892 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
893 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
894 TARGET_SVE_F64MM instead of separate tests.
895 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
896 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
897 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
898 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
899 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
900 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
901 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
902 (TYPES_s_signed): New macro.
903 (TYPES_s_integer): Use it.
904 (TYPES_d_float): New macro.
905 (TYPES_d_data): Use it.
906 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
907 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
908 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
909 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
911 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
913 (ternary_resize2_lane_base): Likewise.
914 (ternary_resize2_base): New class.
915 (ternary_qq_lane_base): Likewise.
916 (ternary_intq_uintq_lane_def): Likewise.
917 (ternary_intq_uintq_lane): New shape.
918 (ternary_intq_uintq_opt_n_def): New class
919 (ternary_intq_uintq_opt_n): New shape.
920 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
921 (ternary_uintq_intq_def): New class.
922 (ternary_uintq_intq): New shape.
923 (ternary_uintq_intq_lane_def): New class.
924 (ternary_uintq_intq_lane): New shape.
925 (ternary_uintq_intq_opt_n_def): New class.
926 (ternary_uintq_intq_opt_n): New shape.
927 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
928 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
929 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
930 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
932 (svdotprod_lane_impl): ...this new class.
933 (svmmla_impl, svusdot_impl): New classes.
934 (svdot_lane): Update to use svdotprod_lane_impl.
935 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
936 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
938 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
939 function, with no types defined.
940 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
941 AARCH64_FL_I8MM functions.
942 (svmmla): New AARCH64_FL_F32MM function.
943 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
944 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
945 AARCH64_FL_F64MM function.
946 (REQUIRED_EXTENSIONS):
948 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
950 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
953 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
955 * config/i386/i386.md (*movoi_internal_avx): Do not check for
956 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
957 (*movti_internal): Do not check for
958 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
959 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
960 just after check for TARGET_AVX.
961 (*movdf_internal): Ditto.
962 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
963 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
964 * config/i386/sse.md (mov<mode>_internal): Only check
965 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
966 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
967 (<sse>_andnot<mode>3<mask_name>): Move check for
968 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
969 (<code><mode>3<mask_name>): Ditto.
970 (*andnot<mode>3): Ditto.
972 (*<code><mode>3): Ditto.
974 (*andnot<VI:mode>3): Remove
975 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
976 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
977 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
978 (sse4_1_blendv<ssemodesuffix>): Ditto.
979 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
980 Explain that tune applies to 128bit instructions only.
982 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
984 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
985 to definition of hsa_kernel_description. Parse assembly to find SGPR
986 and VGPR count of kernel and store in hsa_kernel_description.
988 2020-01-31 Tamar Christina <tamar.christina@arm.com>
990 PR rtl-optimization/91838
991 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
992 to truncate if allowed or reject combination.
994 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
996 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
997 (find_inv_vars_cb): Likewise.
999 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1001 * calls.c (special_function_p): Split out the check for DECL_NAME
1002 being non-NULL and fndecl being extern at file scope into a
1003 new maybe_special_function_p and call it. Drop check for fndecl
1004 being non-NULL that was after a usage of DECL_NAME (fndecl).
1005 * tree.h (maybe_special_function_p): New inline function.
1007 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1009 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1010 (mask_gather_load<mode>): ... here, and zero-initialize the
1012 (maskload<mode>di): Zero-initialize the destination.
1015 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1018 * doc/analyzer.texi (Limitations): Note that constraints on
1019 floating-point values are currently ignored.
1021 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1024 * symtab.c (symtab_node::noninterposable_alias): If localalias
1025 already exists, but is not usable, append numbers after it until
1026 a unique name is found. Formatting fix.
1029 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1032 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1034 * config/gcn/gcn.c (print_operand): Handle LTGT.
1035 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1037 2020-01-30 Richard Biener <rguenther@suse.de>
1039 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1040 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1042 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1044 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1045 without a DECL in .data.rel.ro.local.
1047 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1050 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1054 * config/i386/sse.md
1055 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1056 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1057 any_extend code iterator instead of always zero_extend.
1058 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1059 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1060 Use any_extend code iterator instead of always zero_extend.
1061 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1062 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1063 Use any_extend code iterator instead of always zero_extend.
1064 (*sse2_pmovmskb_ext): New define_insn.
1065 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1068 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1069 (*popcountsi2_zext_falsedep): New define_insn.
1071 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1073 * config.in: Regenerated.
1074 * configure: Regenerated.
1076 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1079 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1080 LLVM's assembler changed the default in version 9.
1082 2020-01-24 Jeff Law <law@redhat.com>
1084 PR tree-optimization/89689
1085 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1087 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1091 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1093 PR rtl-optimization/87763
1094 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1095 simplification to handle subregs as well as bare regs.
1096 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1098 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1101 * ira.c (ira): Revert use of simplified LRA algorithm.
1103 2020-01-29 Martin Jambor <mjambor@suse.cz>
1105 PR tree-optimization/92706
1106 * tree-sra.c (struct access): Fields first_link, last_link,
1107 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1108 next_rhs_queued and grp_rhs_queued respectively, new fields
1109 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1110 (struct assign_link): Field next renamed to next_rhs, new field
1111 next_lhs. Updated comment.
1112 (work_queue_head): Renamed to rhs_work_queue_head.
1113 (lhs_work_queue_head): New variable.
1114 (add_link_to_lhs): New function.
1115 (relink_to_new_repr): Also relink LHS lists.
1116 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1117 (add_access_to_lhs_work_queue): New function.
1118 (pop_access_from_work_queue): Renamed to
1119 pop_access_from_rhs_work_queue.
1120 (pop_access_from_lhs_work_queue): New function.
1121 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1123 (child_would_conflict_in_lacc): Renamed to
1124 child_would_conflict_in_acc. Adjusted parameter names.
1125 (create_artificial_child_access): New parameter set_grp_read, use it.
1126 (subtree_mark_written_and_enqueue): Renamed to
1127 subtree_mark_written_and_rhs_enqueue.
1128 (propagate_subaccesses_across_link): Renamed to
1129 propagate_subaccesses_from_rhs.
1130 (propagate_subaccesses_from_lhs): New function.
1131 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1134 2020-01-29 Martin Jambor <mjambor@suse.cz>
1136 PR tree-optimization/92706
1137 * tree-sra.c (struct access): Adjust comment of
1138 grp_total_scalarization.
1139 (find_access_in_subtree): Look for single children spanning an entire
1141 (scalarizable_type_p): Allow register accesses, adjust callers.
1142 (completely_scalarize): Remove function.
1143 (scalarize_elem): Likewise.
1144 (create_total_scalarization_access): Likewise.
1145 (sort_and_splice_var_accesses): Do not track total scalarization
1147 (analyze_access_subtree): New parameter totally, adjust to new meaning
1148 of grp_total_scalarization.
1149 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1150 (can_totally_scalarize_forest_p): New function.
1151 (create_total_scalarization_access): Likewise.
1152 (create_total_access_and_reshape): Likewise.
1153 (total_should_skip_creating_access): Likewise.
1154 (totally_scalarize_subtree): Likewise.
1155 (analyze_all_variable_accesses): Perform total scalarization after
1156 subaccess propagation using the new functions above.
1157 (initialize_constant_pool_replacements): Output initializers by
1158 traversing the access tree.
1160 2020-01-29 Martin Jambor <mjambor@suse.cz>
1162 * tree-sra.c (verify_sra_access_forest): New function.
1163 (verify_all_sra_access_forests): Likewise.
1164 (create_artificial_child_access): Set parent.
1165 (analyze_all_variable_accesses): Call the verifier.
1167 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1169 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1170 if called on indirect edge.
1171 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1172 speculative call if needed.
1174 2020-01-29 Richard Biener <rguenther@suse.de>
1176 PR tree-optimization/93428
1177 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1178 permutation when the load node is created.
1179 (vect_analyze_slp_instance): Re-use it here.
1181 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1183 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1185 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1187 PR rtl-optimization/93272
1188 * ira-lives.c (process_out_of_region_eh_regs): New function.
1189 (process_bb_node_lives): Call it.
1191 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1193 * coverage.c (read_counts_file): Make error message lowercase.
1195 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1197 * profile-count.c (profile_quality_display_names): Fix ordering.
1199 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1202 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1203 hash only when edge is first within the sequence.
1204 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1205 (symbol_table::create_edge): Do not set target_prob.
1206 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1208 (cgraph_edge::make_speculative): Drop target_prob parameter.
1209 (cgraph_edge::speculative_call_info): Remove.
1210 (cgraph_edge::first_speculative_call_target): New member function.
1211 (update_call_stmt_hash_for_removing_direct_edge): New function.
1212 (cgraph_edge::resolve_speculation): Rewrite to new API.
1213 (cgraph_edge::speculative_call_for_target): New member function.
1214 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1215 multiple speculation targets.
1216 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1218 (verify_speculative_call): Verify that targets form an interval.
1219 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1220 (cgraph_edge::first_speculative_call_target): New member function.
1221 (cgraph_edge::next_speculative_call_target): New member function.
1222 (cgraph_edge::speculative_call_target_ref): New member function.
1223 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1224 (cgraph_edge): Remove target_prob.
1225 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1226 Fix handling of speculative calls.
1227 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1228 * ipa-fnsummary.c (analyze_function_body): Likewise.
1229 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1230 * ipa-profile.c (dump_histogram): Fix formating.
1231 (ipa_profile_generate_summary): Watch for overflows.
1232 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1233 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1234 (update_indirect_edges_after_inlining): Update to new API.
1235 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1237 * profile-count.h: (profile_probability::adjusted): New.
1238 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1239 updating of profile.
1240 * value-prof.c (gimple_ic_transform): Rename to ...
1241 (dump_ic_profile): ... this one; update dumping.
1242 (stream_in_histogram_value): Fix formating.
1243 (gimple_value_profile_transformations): Update.
1245 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1248 * config/i386/i386.md (*movoi_internal_avx): Remove
1249 TARGET_SSE_TYPELESS_STORES check.
1250 (*movti_internal): Prefer TARGET_AVX over
1251 TARGET_SSE_TYPELESS_STORES.
1252 (*movtf_internal): Likewise.
1253 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1254 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1255 from TARGET_SSE_TYPELESS_STORES.
1257 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1259 * diagnostic-core.h (warning_at): Rename overload to...
1260 (warning_meta): ...this.
1261 (emit_diagnostic_valist): Delete decl of overload taking
1262 diagnostic_metadata.
1263 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1264 (warning_at): Rename overload taking diagnostic_metadata to...
1265 (warning_meta): ...this.
1267 2020-01-28 Richard Biener <rguenther@suse.de>
1269 PR tree-optimization/93439
1270 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1271 * tree-cfg.c (move_sese_region_to_fn): ... here.
1272 (verify_types_in_gimple_reference): Verify used cliques are
1275 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1278 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1279 argument of a pointer to struct gcc_options and pass it to
1280 parse_mtune_ctrl_str.
1281 (ix86_function_specific_restore): Pass opts to
1282 set_ix86_tune_features.
1283 (ix86_option_override_internal): Likewise.
1284 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1285 gcc_options and use it for x_ix86_tune_ctrl_string.
1287 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1289 PR rtl-optimization/87763
1290 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1291 simplification to handle subregs as well as bare regs.
1292 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1294 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1296 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1297 for reduction chains that (now) include a call.
1299 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1301 PR tree-optimization/92822
1302 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1303 out the don't-care elements of a vector whose significant elements
1304 are duplicates, make the don't-care elements duplicates too.
1306 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1308 PR tree-optimization/93434
1309 * tree-predcom.c (split_data_refs_to_components): Record which
1310 components have had aliasing loads removed. Prevent store-store
1311 commoning for all such components.
1313 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1316 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1317 -1 or is_vshift is true, use new_vector with number of elts npatterns
1318 rather than new_unary_operation.
1320 PR tree-optimization/93454
1321 * gimple-fold.c (fold_array_ctor_reference): Perform
1322 elt_size.to_uhwi () just once, instead of calling it in every
1323 iteration. Punt if that value is above size of the temporary
1324 buffer. Decrease third native_encode_expr argument when
1325 bufoff + elt_sz is above size of buf.
1327 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1329 * config/mips/mips.c (mips_declare_object_name)
1330 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1332 2020-01-27 Martin Liska <mliska@suse.cz>
1334 PR gcov-profile/93403
1335 * tree-profile.c (gimple_init_gcov_profiler): Generate
1336 both __gcov_indirect_call_profiler_v4 and
1337 __gcov_indirect_call_profiler_v4_atomic.
1339 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1342 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1344 (@aarch64_split_simd_mov<mode>): Use it.
1345 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1346 Leave the vec_extract patterns to handle 2-element vectors.
1347 (aarch64_simd_mov_from_<mode>high): Likewise.
1348 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1349 (vec_extractv2dfv1df): Likewise.
1351 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1353 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1354 jump conditions for *compare_condjump<GPI:mode>.
1356 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1359 * digraph.cc (test_edge::test_edge): Specify template for base
1362 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1364 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1366 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1368 * config/arc/arc-protos.h (gen_mlo): Remove.
1369 (gen_mhi): Likewise.
1370 * config/arc/arc.c (AUX_MULHI): Define.
1371 (arc_must_save_reister): Special handling for r58/59.
1372 (arc_compute_frame_size): Consider mlo/mhi registers.
1373 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1375 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1376 mlo/mhi name selection.
1377 (arc_restore_callee_saves): Don't early restore blink when ISR.
1378 (arc_expand_prologue): Add mlo/mhi saving.
1379 (arc_expand_epilogue): Add mlo/mhi restoring.
1382 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1383 numbering when MUL64 option is used.
1384 (DWARF2_FRAME_REG_OUT): Define.
1385 * config/arc/arc.md (arc600_stall): New pattern.
1386 (VUNSPEC_ARC_ARC600_STALL): Define.
1387 (mulsi64): Use correct mlo/mhi registers.
1388 (mulsi_600): Clean it up.
1389 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1391 (mhi_operand): Likewise.
1393 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1394 Petro Karashchenko <petro.karashchenko@ring.com>
1396 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1397 attributes if needed.
1398 (prepare_move_operands): Generate special unspec instruction for
1400 (arc_isuncached_mem_p): Propagate uncached attribute to each
1402 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1403 (VUNSPEC_ARC_STDI): Likewise.
1404 (ALLI): New mode iterator.
1405 (mALLI): New mode attribute.
1406 (lddi): New instruction pattern.
1408 (stdidi_split): Split instruction for architectures which are not
1409 supporting ll64 option.
1410 (lddidi_split): Likewise.
1412 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1414 PR rtl-optimization/92989
1415 * lra-lives.c (process_bb_lives): Update the live-in set before
1416 processing additional clobbers.
1418 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1420 PR rtl-optimization/93170
1421 * cselib.c (cselib_invalidate_regno_val): New function, split out
1423 (cselib_invalidate_regno): ...here.
1424 (cselib_invalidated_by_call_p): New function.
1425 (cselib_process_insn): Iterate over all the hard-register entries in
1426 REG_VALUES and invalidate any that cross call-clobbered registers.
1428 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1430 * dojump.c (split_comparison): Use HONOR_NANS rather than
1431 HONOR_SNANS when splitting LTGT.
1433 2020-01-27 Martin Liska <mliska@suse.cz>
1436 * opts.c (print_filtered_help): Exclude language-specific
1437 options from --help=common unless enabled in all FEs.
1439 2020-01-27 Martin Liska <mliska@suse.cz>
1441 * opts.c (print_help): Exclude params from
1442 all except --help=param.
1444 2020-01-27 Martin Liska <mliska@suse.cz>
1447 * config/i386/i386-features.c (make_resolver_func):
1448 Align the code with ppc64 target implementation.
1449 Do not generate a unique name for resolver function.
1451 2020-01-27 Richard Biener <rguenther@suse.de>
1453 PR tree-optimization/93397
1454 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1455 converted reduction chain SLP graph adjustment.
1457 2020-01-26 Marek Polacek <polacek@redhat.com>
1460 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1463 2020-01-26 Jason Merrill <jason@redhat.com>
1466 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1469 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1471 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1472 (rx_setmem): Likewise.
1474 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1477 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1478 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1479 drop <di> from constraint of last operand.
1482 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1483 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1484 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1486 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1489 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1492 2020-01-24 Jeff Law <law@redhat.com>
1494 PR tree-optimization/92788
1495 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1498 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1501 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1502 *avx_vperm_broadcast_<mode>,
1503 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1504 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1505 Move before avx2_perm<mode>/avx512f_perm<mode>.
1508 * simplify-rtx.c (simplify_const_unary_operation,
1509 simplify_const_binary_operation): Punt for mode precision above
1510 MAX_BITSIZE_MODE_ANY_INT.
1512 2020-01-24 Andrew Pinski <apinski@marvell.com>
1514 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1517 2020-01-24 Jeff Law <law@redhat.com>
1520 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1521 for REGs. Call output_operand_lossage to get more reasonable
1524 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1526 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1527 gcn_fp_compare_operator.
1528 (vec_cmpu<mode>di): Use gcn_compare_operator.
1529 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1530 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1531 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1532 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1533 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1534 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1535 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1536 gcn_fp_compare_operator.
1537 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1538 gcn_fp_compare_operator.
1539 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1540 gcn_fp_compare_operator.
1541 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1542 gcn_fp_compare_operator.
1544 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1546 * doc/install.texi (Cross-Compiler-Specific Options): Document
1547 `--with-toolexeclibdir' option.
1549 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1551 * target.def (flags_regnum): Also mention effect on delay slot filling.
1552 * doc/tm.texi: Regenerate.
1554 2020-01-23 Jeff Law <law@redhat.com>
1556 PR translation/90162
1557 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1559 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1562 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1565 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1567 PR rtl-optimization/93402
1568 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1571 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1573 * config.in: Regenerated.
1574 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1575 for TARGET_LIBC_GNUSTACK.
1576 * configure: Regenerated.
1577 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1578 found to be 2.31 or greater.
1580 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1582 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1584 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1585 (mips_asm_file_end): New function. Delegate to
1586 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1587 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1589 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1592 * config/i386/i386-modes.def (POImode): New mode.
1593 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1594 * config/i386/i386.md (DPWI): New mode attribute.
1595 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1597 (QPWI): ... this. Use POI instead of OI for TImode.
1598 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1599 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1602 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1605 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1607 (speculation_tracker_rev): New pattern.
1608 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1609 Use speculation_tracker_rev to track the inverse condition.
1611 2020-01-23 Richard Biener <rguenther@suse.de>
1613 PR tree-optimization/93381
1614 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1615 alias-set of the def as argument and record the first one.
1616 (vn_walk_cb_data::first_set): New member.
1617 (vn_reference_lookup_3): Pass the alias-set of the current def
1618 to push_partial_def. Fix alias-set used in the aggregate copy
1620 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1621 * real.c (clear_significand_below): Fix out-of-bound access.
1623 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1626 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1627 New define_insn patterns.
1629 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1631 * doc/sourcebuild.texi (check-function-bodies): Add an
1632 optional target/xfail selector.
1634 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1636 PR rtl-optimization/93124
1637 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1638 bare USE and CLOBBER insns.
1640 2020-01-22 Andrew Pinski <apinski@marvell.com>
1642 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1644 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1647 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1648 diagnostic_manager into "ana" namespace.
1649 * selftest-run-tests.c (selftest::run_tests): Update for move of
1650 selftest::run_analyzer_selftests to
1651 ana::selftest::run_analyzer_selftests.
1653 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1655 * cfgexpand.c (union_stack_vars): Update the size.
1657 2020-01-22 Richard Biener <rguenther@suse.de>
1659 PR tree-optimization/93381
1660 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1661 throughout, handle all conversions the same.
1663 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1666 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1667 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1668 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1669 Call force_reg on high_in2 unconditionally.
1671 2020-01-22 Martin Liska <mliska@suse.cz>
1673 PR tree-optimization/92924
1674 * profile.c (compute_value_histograms): Divide
1677 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1680 * output.h (assemble_name_resolve): Declare.
1681 * varasm.c (assemble_name_resolve): New function.
1682 (assemble_name): Use it.
1683 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1685 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1687 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1688 update_web_docs_git instead of update_web_docs_svn.
1690 2020-01-21 Andrew Pinski <apinski@marvell.com>
1693 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1694 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1695 (*tlsgd_small_<mode>): Likewise.
1696 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1697 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1698 register. Convert that register back to dest using convert_mode.
1700 2020-01-21 Jim Wilson <jimw@sifive.com>
1702 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1705 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1706 Uros Bizjak <ubizjak@gmail.com>
1709 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1711 (legitimize_tls_address): Do GNU2 TLS address computation in
1712 ptr_mode and zero-extend result to Pmode.
1713 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1714 :P with :PTR and Pmode with ptr_mode.
1715 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1716 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1717 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1719 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1722 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1723 the last two operands are CONST_INT_P before using them as such.
1725 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1727 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1728 to get the integer element types.
1730 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1732 * config/aarch64/aarch64-sve-builtins.h
1733 (function_expander::convert_to_pmode): Declare.
1734 * config/aarch64/aarch64-sve-builtins.cc
1735 (function_expander::convert_to_pmode): New function.
1736 (function_expander::get_contiguous_base): Use it.
1737 (function_expander::prepare_gather_address_operands): Likewise.
1738 * config/aarch64/aarch64-sve-builtins-sve2.cc
1739 (svwhilerw_svwhilewr_impl::expand): Likewise.
1741 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1744 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1745 cfun->machine->label_is_assembled.
1746 (aarch64_print_patchable_function_entry): New.
1747 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1748 * config/aarch64/aarch64.h (struct machine_function): New field,
1751 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1754 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1757 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1760 * cgraph.c (cgraph_edge::resolve_speculation,
1761 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1762 call_stmt_site_hash.
1764 2020-01-21 Martin Liska <mliska@suse.cz>
1766 * config/rs6000/rs6000.c (common_mode_defined): Remove
1769 2020-01-21 Richard Biener <rguenther@suse.de>
1771 PR tree-optimization/92328
1772 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1773 type when value-numbering same-sized store by inserting a
1775 (eliminate_dom_walker::eliminate_stmt): When eliminating
1776 a redundant store handle bit-reinterpretation of the same value.
1778 2020-01-21 Andrew Pinski <apinski@marvel.com>
1781 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1783 (prepare_block_for_update): This. Use a worklist instead of
1786 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1788 * gcc/config/arm/arm.c (clear_operation_p):
1789 Initialise last_regno, skip first iteration
1790 based on the first_set value and use ints instead
1791 of the unnecessary HOST_WIDE_INTs.
1793 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1796 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1797 compare_mode other than SFmode or DFmode.
1799 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1802 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1803 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1804 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1806 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1808 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1810 2020-01-20 Andrew Pinski <apinski@marvell.com>
1813 * targhooks.c (default_print_patchable_function_entry): Use
1814 output_asm_insn to emit the nop instruction.
1816 2020-01-20 Fangrui Song <maskray@google.com>
1819 * targhooks.c (default_print_patchable_function_entry): Align to
1822 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
1825 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1826 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
1827 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1828 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
1829 (*tls_dynamic_gnu2_lea_64): Renamed to ...
1830 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
1831 Remove the {q} suffix from lea.
1832 (*tls_dynamic_gnu2_call_64): Renamed to ...
1833 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
1834 (*tls_dynamic_gnu2_combine_64): Renamed to ...
1835 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
1836 Pass Pmode to gen_tls_dynamic_gnu2_64.
1838 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1840 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1842 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
1844 * config/aarch64/aarch64-sve-builtins-base.cc
1845 (svld1ro_impl::memory_vector_mode): Remove parameter name.
1847 2020-01-20 Richard Biener <rguenther@suse.de>
1850 * dwarf2out.c (prune_unused_types): Unconditionally mark
1851 called function DIEs.
1853 2020-01-20 Martin Liska <mliska@suse.cz>
1855 PR tree-optimization/93199
1856 * tree-eh.c (struct leh_state): Add
1857 new field outer_non_cleanup.
1858 (cleanup_is_dead_in): Pass leh_state instead
1859 of eh_region. Add a checking that state->outer_non_cleanup
1860 points to outer non-clean up region.
1861 (lower_try_finally): Record outer_non_cleanup
1863 (lower_catch): Likewise.
1864 (lower_eh_filter): Likewise.
1865 (lower_eh_must_not_throw): Likewise.
1866 (lower_cleanup): Likewise.
1868 2020-01-20 Richard Biener <rguenther@suse.de>
1870 PR tree-optimization/93094
1871 * tree-vectorizer.h (vect_loop_versioning): Adjust.
1872 (vect_transform_loop): Likewise.
1873 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1874 loop_vectorized_call to vect_transform_loop.
1875 * tree-vect-loop.c (vect_transform_loop): Pass down
1876 loop_vectorized_call to vect_loop_versioning.
1877 * tree-vect-loop-manip.c (vect_loop_versioning): Use
1878 the earlier discovered loop_vectorized_call.
1880 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
1882 * doc/contribute.texi: Update for SVN -> Git transition.
1883 * doc/install.texi: Likewise.
1885 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1888 * cgraph.c (cgraph_edge::make_speculative): Increase number of
1889 speculative targets.
1890 (verify_speculative_call): New function
1891 (cgraph_node::verify_node): Use it.
1892 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
1895 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1898 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
1899 (cgraph_edge::make_direct): Remove all indirect targets.
1900 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
1901 (cgraph_node::verify_node): Verify that only one call_stmt or
1902 lto_stmt_uid is set.
1903 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
1905 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
1906 (lto_output_ref): Simplify streaming of stmt.
1907 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
1909 2020-01-18 Tamar Christina <tamar.christina@arm.com>
1911 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
1912 Mark parameter unused.
1914 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
1916 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
1918 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
1920 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
1922 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
1924 * Makefile.in: Add coroutine-passes.o.
1925 * builtin-types.def (BT_CONST_SIZE): New.
1926 (BT_FN_BOOL_PTR): New.
1927 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
1928 * builtins.def (DEF_COROUTINE_BUILTIN): New.
1929 * coroutine-builtins.def: New file.
1930 * coroutine-passes.cc: New file.
1931 * function.h (struct GTY function): Add a bit to indicate that the
1932 function is a coroutine component.
1933 * internal-fn.c (expand_CO_FRAME): New.
1934 (expand_CO_YIELD): New.
1935 (expand_CO_SUSPN): New.
1936 (expand_CO_ACTOR): New.
1937 * internal-fn.def (CO_ACTOR): New.
1941 * passes.def: Add pass_coroutine_lower_builtins,
1942 pass_coroutine_early_expand_ifns.
1943 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
1944 (make_pass_coroutine_early_expand_ifns): New.
1945 * doc/invoke.texi: Document the fcoroutines command line
1948 2020-01-18 Jakub Jelinek <jakub@redhat.com>
1950 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
1953 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
1954 after checking the argument is a REG. Don't use REGNO (reg)
1955 again to set last_regno, reuse regno variable instead.
1957 2020-01-17 David Malcolm <dmalcolm@redhat.com>
1959 * doc/analyzer.texi (Limitations): Add note about NaN.
1961 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1962 Sudakshina Das <sudi.das@arm.com>
1964 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
1965 and valid immediate.
1966 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
1967 (lshrdi3): Generate thumb2_lsrl for valid immediates.
1968 * config/arm/constraints.md (Pg): New.
1969 * config/arm/predicates.md (long_shift_imm): New.
1970 (arm_reg_or_long_shift_imm): Likewise.
1971 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
1972 (thumb2_lsll): Likewise.
1975 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1976 Sudakshina Das <sudi.das@arm.com>
1978 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
1979 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
1980 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
1981 register pairs for doubleword quantities for ARMv8.1M-Mainline.
1982 * config/arm/thumb2.md (thumb2_asrl): New.
1983 (thumb2_lsll): Likewise.
1985 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1987 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
1990 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
1992 * gdbinit.in (help-gcc-hooks): New command.
1993 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
1994 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
1997 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1999 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2000 correct target macro.
2002 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2004 * config/aarch64/aarch64-protos.h
2005 (aarch64_sve_ld1ro_operand_p): New.
2006 * config/aarch64/aarch64-sve-builtins-base.cc
2007 (class load_replicate): New.
2008 (class svld1ro_impl): New.
2009 (class svld1rq_impl): Change to inherit from load_replicate.
2010 (svld1ro): New sve intrinsic function base.
2011 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2012 New DEF_SVE_FUNCTION.
2013 * config/aarch64/aarch64-sve-builtins-base.h
2014 (svld1ro): New decl.
2015 * config/aarch64/aarch64-sve-builtins.cc
2016 (function_expander::add_mem_operand): Modify assert to allow
2018 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2020 * config/aarch64/aarch64.c
2021 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2022 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2023 (aarch64_sve_ld1ro_operand_p): New.
2024 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2025 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2026 * config/aarch64/predicates.md
2027 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2029 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2031 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2032 Introduce this ACLE specified predefined macro.
2033 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2034 (fp): Disabling this disables f64mm.
2035 (simd): Disabling this disables f64mm.
2036 (fp16): Disabling this disables f64mm.
2037 (sve): Disabling this disables f64mm.
2038 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2039 (AARCH64_ISA_F64MM): New.
2040 (TARGET_F64MM): New.
2041 * doc/invoke.texi (f64mm): Document new option.
2043 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2045 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2046 (neoversen1_tunings): Likewise.
2048 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2051 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2052 Add assert to ensure prolog has been emitted.
2053 (aarch64_split_atomic_op): Likewise.
2054 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2055 Use epilogue_completed rather than reload_completed.
2056 (aarch64_atomic_exchange<mode>): Likewise.
2057 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2058 (atomic_nand<mode>): Likewise.
2059 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2060 (atomic_fetch_nand<mode>): Likewise.
2061 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2062 (atomic_nand_fetch<mode>): Likewise.
2064 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2067 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2069 (REVERSE_CONDITION): Delete.
2070 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2071 (CCFP_CCFPE): Likewise.
2072 (e): New mode attribute.
2073 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2074 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2075 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2076 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2077 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2078 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2079 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2080 name of generator from gen_ccmpdi to gen_ccmpccdi.
2081 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2082 the previous comparison but aren't able to, use the new ccmp_rev
2085 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2087 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2088 than testing directly for INTEGER_CST.
2089 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2091 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2093 PR tree-optimization/93292
2094 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2095 get_vectype_for_scalar_type returns NULL.
2097 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2099 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2100 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2102 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2104 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2106 * params.opt: (max-predicted-iterations): Set bounds.
2107 * predict.c (real_almost_one, real_br_prob_base,
2108 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2109 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2110 probabilities; do not truncate to reg_br_prob_bases.
2111 (estimate_loops_at_level): Pass max_cyclic_prob.
2112 (estimate_loops): Compute max_cyclic_prob.
2113 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2115 * profile-count.c (profile_probability::to_sreal): New.
2116 * profile-count.h (class sreal): Move up in file.
2117 (profile_probability::to_sreal): Declare.
2119 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2122 (arm_invalid_conversion): New function for target hook.
2123 (arm_invalid_unary_op): New function for target hook.
2124 (arm_invalid_binary_op): New function for target hook.
2126 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2128 * config.gcc: Add arm_bf16.h.
2129 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2130 (arm_simd_builtin_std_type): Add BFmode.
2131 (arm_init_simd_builtin_types): Define element types for vector types.
2132 (arm_init_bf16_types): New function.
2133 (arm_init_builtins): Add arm_init_bf16_types function call.
2134 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2135 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2136 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2137 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2138 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2139 (arm_mangle_type): Add __bf16.
2140 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2141 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2142 arm_bf16_ptr_type_node.
2143 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2144 define_split between ARM registers.
2145 * config/arm/arm_bf16.h: New file.
2146 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2147 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2149 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2150 * config/arm/vfp.md: Add BFmode to movhf patterns.
2152 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2153 Andre Vieira <andre.simoesdiasvieira@arm.com>
2155 * config/arm/arm-cpus.in (mve, mve_float): New features.
2156 (dsp, mve, mve.fp): New options.
2157 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2158 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2159 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2161 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2162 Thomas Preud'homme <thomas.preudhomme@arm.com>
2164 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2166 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2167 error for using -mcmse when targeting Armv8.1-M Mainline.
2169 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2170 Thomas Preud'homme <thomas.preudhomme@arm.com>
2172 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2173 address in r4 when targeting Armv8.1-M Mainline.
2174 (nonsecure_call_value_internal): Likewise.
2175 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2176 a register match_operand again. Emit BLXNS when targeting
2178 (nonsecure_call_value_reg_thumb2): Likewise.
2180 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2181 Thomas Preud'homme <thomas.preudhomme@arm.com>
2183 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2184 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2185 variable as true when floating-point ABI is not hard. Replace
2186 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2187 Generate VLSTM and VLLDM instruction respectively before and
2188 after a function call to cmse_nonsecure_call function.
2189 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2190 (VUNSPEC_VLLDM): Likewise.
2191 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2192 (lazy_load_multiple_insn): Likewise.
2194 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2195 Thomas Preud'homme <thomas.preudhomme@arm.com>
2197 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2198 (arm_emit_vfp_multi_reg_pop): Likewise.
2199 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2200 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2201 restore callee-saved VFP registers.
2203 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2204 Thomas Preud'homme <thomas.preudhomme@arm.com>
2206 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2207 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2208 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2209 callee-saved GPRs as well as clear ip register before doing a nonsecure
2210 call then restore callee-saved GPRs after it when targeting
2212 (arm_reorg): Adapt to function rename.
2214 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2215 Thomas Preud'homme <thomas.preudhomme@arm.com>
2217 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2218 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2219 clear_vfp_multiple pattern based on a new vfp parameter.
2220 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2221 targeting Armv8.1-M Mainline.
2222 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2223 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2224 whether VFP registers are available before looking call_used_regs for a
2226 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2227 of prototype of clear_operation_p.
2228 (clear_vfp_multiple_operation): New predicate.
2229 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2230 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
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): Declare.
2236 * config/arm/arm.c (clear_operation_p): New function.
2237 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2238 targeting Armv8.1-M Mainline or successor.
2239 (output_return_instruction): Only output APSR register clearing if
2240 Armv8.1-M Mainline instructions not available.
2241 (thumb_exit): Likewise.
2242 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2243 * config/arm/thumb2.md (clear_apsr): New define_insn.
2244 (clear_multiple): Likewise.
2245 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2247 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2248 Thomas Preud'homme <thomas.preudhomme@arm.com>
2250 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2251 (use_return_insn): Also return false for Armv8.1-M Mainline.
2252 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2253 Mainline instructions are available.
2254 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2255 when targeting Armv8.1-M Mainline Security Extensions.
2256 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2257 Mainline entry function.
2258 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2259 targeting Armv8.1-M Mainline or successor.
2260 (arm_expand_epilogue): Fix indentation of caller-saved register
2261 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2263 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2264 (FP_SYSREGS): Likewise.
2265 (enum vfp_sysregs_encoding): Define enum.
2266 (fp_sysreg_names): Declare.
2267 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2268 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2269 (pop_fpsysreg_insn): Likewise.
2271 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2272 Thomas Preud'homme <thomas.preudhomme@arm.com>
2274 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2275 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2276 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2277 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2278 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2279 (ARMv8_1m_main): New feature group.
2280 (armv8.1-m.main): New architecture.
2281 * config/arm/arm-tables.opt: Regenerate.
2282 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2283 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2284 (arm_options_perform_arch_sanity_checks): Error out when targeting
2285 Armv8.1-M Mainline Security Extensions.
2286 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2288 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2290 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2291 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2292 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2293 aarch64_bfdot_laneq): New.
2294 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2295 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2296 vbfdotq_laneq_f32): New.
2297 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2298 VBFMLA_W, VBF): New.
2299 (isquadop): Add V4BF, V8BF.
2301 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2303 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2304 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2305 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2306 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2307 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2308 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2309 usdot_laneq, sudot_lane,sudot_laneq): New.
2310 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2311 (aarch64_<sur>dot_lane): New.
2312 * config/aarch64/arm_neon.h (vusdot_s32): New.
2314 (vusdot_lane_s32): New.
2315 (vsudot_lane_s32): New.
2316 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2317 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2319 2020-01-16 Martin Liska <mliska@suse.cz>
2321 * value-prof.c (dump_histogram_value): Fix
2322 obvious spacing issue.
2324 2020-01-16 Andrew Pinski <apinski@marvell.com>
2326 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2327 !storage_order_barrier_p.
2329 2020-01-16 Andrew Pinski <apinski@marvell.com>
2331 * sched-int.h (_dep): Add unused bit-field field for the padding.
2332 * sched-deps.c (init_dep_1): Init unused field.
2334 2020-01-16 Andrew Pinski <apinski@marvell.com>
2336 * optabs.h (create_expand_operand): Initialize target field also.
2338 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2340 PR tree-optimization/92429
2341 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2342 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2344 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2347 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2349 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2350 aarch64_sve_int_mode to each mode.
2352 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2354 * doc/analyzer.texi (Overview): Add note about
2355 -fdump-ipa-analyzer.
2357 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2359 PR tree-optimization/93231
2360 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2361 input_type is unsigned. Use tree_to_shwi for shift constant.
2362 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2363 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2364 inputs more efficiently.
2366 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2368 * config/i386/i386.md (*movsf_internal): Do not require
2369 SSE2 ISA for alternatives 14 and 15.
2371 2020-01-15 Richard Biener <rguenther@suse.de>
2374 * tree-eh.c (sink_clobbers): If we already visited the destination
2375 block do not defer insertion.
2376 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2377 the purpose of defered insertion.
2379 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2381 * BASE-VER: Bump to 10.0.1.
2383 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2385 PR tree-optimization/93247
2386 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2387 type of the stmt that we're going to vectorize.
2389 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2391 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2392 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2395 2020-01-15 Martin Liska <mliska@suse.cz>
2397 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2398 2 calls of streamer_read_hwi in a function call.
2400 2020-01-15 Richard Biener <rguenther@suse.de>
2402 * alias.c (record_alias_subset): Avoid redundant work when
2403 subset is already recorded.
2405 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2407 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2408 the analyzer options provide CWE identifiers.
2410 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2412 * tree-diagnostic-path.cc (path_summary::event_range::print):
2413 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2414 using get_pure_location.
2416 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2418 PR tree-optimization/93262
2419 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2420 perform head trimming only if the last argument is constant,
2421 either all ones, or larger or equal to head trim, in the latter
2422 case decrease the last argument by head_trim.
2424 PR tree-optimization/93249
2425 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2426 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2427 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2428 perform head trim unless we can prove there are no '\0' chars
2429 from the source among the first head_trim chars.
2431 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2433 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2435 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2438 * config/i386/sse.md
2439 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2440 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2441 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2442 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2443 just a single alternative instead of two, make operands 1 and 2
2446 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2449 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2452 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2454 * Makefile.in (lang_opt_files): Add analyzer.opt.
2455 (ANALYZER_OBJS): New.
2456 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2457 tristate.o and ANALYZER_OBJS.
2458 (TEXI_GCCINT_FILES): Add analyzer.texi.
2459 * common.opt (-fanalyzer): New driver option.
2460 * config.in: Regenerate.
2461 * configure: Regenerate.
2462 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2463 (gccdepdir): Also create depdir for "analyzer" subdir.
2464 * digraph.cc: New file.
2465 * digraph.h: New file.
2466 * doc/analyzer.texi: New file.
2467 * doc/gccint.texi ("Static Analyzer") New menu item.
2468 (analyzer.texi): Include it.
2469 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2470 ("Warning Options"): Add static analysis warnings to the list.
2471 (-Wno-analyzer-double-fclose): New option.
2472 (-Wno-analyzer-double-free): New option.
2473 (-Wno-analyzer-exposure-through-output-file): New option.
2474 (-Wno-analyzer-file-leak): New option.
2475 (-Wno-analyzer-free-of-non-heap): New option.
2476 (-Wno-analyzer-malloc-leak): New option.
2477 (-Wno-analyzer-possible-null-argument): New option.
2478 (-Wno-analyzer-possible-null-dereference): New option.
2479 (-Wno-analyzer-null-argument): New option.
2480 (-Wno-analyzer-null-dereference): New option.
2481 (-Wno-analyzer-stale-setjmp-buffer): New option.
2482 (-Wno-analyzer-tainted-array-index): New option.
2483 (-Wno-analyzer-use-after-free): New option.
2484 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2485 (-Wno-analyzer-use-of-uninitialized-value): New option.
2486 (-Wanalyzer-too-complex): New option.
2487 (-fanalyzer-call-summaries): New warning.
2488 (-fanalyzer-checker=): New warning.
2489 (-fanalyzer-fine-grained): New warning.
2490 (-fno-analyzer-state-merge): New warning.
2491 (-fno-analyzer-state-purge): New warning.
2492 (-fanalyzer-transitivity): New warning.
2493 (-fanalyzer-verbose-edges): New warning.
2494 (-fanalyzer-verbose-state-changes): New warning.
2495 (-fanalyzer-verbosity=): New warning.
2496 (-fdump-analyzer): New warning.
2497 (-fdump-analyzer-callgraph): New warning.
2498 (-fdump-analyzer-exploded-graph): New warning.
2499 (-fdump-analyzer-exploded-nodes): New warning.
2500 (-fdump-analyzer-exploded-nodes-2): New warning.
2501 (-fdump-analyzer-exploded-nodes-3): New warning.
2502 (-fdump-analyzer-supergraph): New warning.
2503 * doc/sourcebuild.texi (dg-require-dot): New.
2504 (dg-check-dot): New.
2505 * gdbinit.in (break-on-saved-diagnostic): New command.
2506 * graphviz.cc: New file.
2507 * graphviz.h: New file.
2508 * ordered-hash-map-tests.cc: New file.
2509 * ordered-hash-map.h: New file.
2510 * passes.def (pass_analyzer): Add before
2511 pass_ipa_whole_program_visibility.
2512 * selftest-run-tests.c (selftest::run_tests): Call
2513 selftest::ordered_hash_map_tests_cc_tests.
2514 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2516 * shortest-paths.h: New file.
2517 * timevar.def (TV_ANALYZER): New timevar.
2518 (TV_ANALYZER_SUPERGRAPH): Likewise.
2519 (TV_ANALYZER_STATE_PURGE): Likewise.
2520 (TV_ANALYZER_PLAN): Likewise.
2521 (TV_ANALYZER_SCC): Likewise.
2522 (TV_ANALYZER_WORKLIST): Likewise.
2523 (TV_ANALYZER_DUMP): Likewise.
2524 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2525 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2526 * tree-pass.h (make_pass_analyzer): New decl.
2527 * tristate.cc: New file.
2528 * tristate.h: New file.
2530 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2533 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2534 alternatives 9 and 10.
2536 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2538 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2539 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2540 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2541 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2542 (selftest::hash_map_tests_c_tests): Call it.
2543 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2544 New static constant, using the value of = H::empty_zero_p.
2545 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2546 from default_hash_traits <Value>.
2547 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2549 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2550 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2551 calls to mark_empty with !Descriptor::empty_zero_p.
2552 (hash_table::empty_slow): Conditionalize the memset call with a
2553 check that Descriptor::empty_zero_p; otherwise, loop through the
2554 entries calling mark_empty on them.
2555 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2556 (pointer_hash::empty_zero_p): Likewise.
2557 (pair_hash::empty_zero_p): Likewise.
2558 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2560 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2561 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2562 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2563 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2564 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2565 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2566 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2567 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2569 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2572 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2574 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2575 fix typo on return value.
2577 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2580 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2582 (cgraph_edge::make_speculative): Add param for setting speculative_id
2584 (cgraph_edge::speculative_call_info): Update comments and find reference
2585 by speculative_id for multiple indirect targets.
2586 (cgraph_edge::resolve_speculation): Decrease the speculations
2587 for indirect edge, drop it's speculative if not direct target
2588 left. Update comments.
2589 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2590 (cgraph_node::dump): Print num_speculative_call_targets.
2591 (cgraph_node::verify_node): Don't report error if speculative
2592 edge not include statement.
2593 (cgraph_edge::num_speculative_call_targets_p): New function.
2594 * cgraph.h (int common_target_id): Remove.
2595 (int common_target_probability): Remove.
2596 (num_speculative_call_targets): New variable.
2597 (make_speculative): Add param for setting speculative_id.
2598 (cgraph_edge::num_speculative_call_targets_p): New declare.
2599 (target_prob): New variable.
2600 (speculative_id): New variable.
2601 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2602 call summaries for multiple speculative call targets.
2603 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2604 * ipa-profile.c (struct speculative_call_target): New struct.
2605 (class speculative_call_summary): New class.
2606 (class speculative_call_summaries): New class.
2607 (call_sums): New variable.
2608 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2609 (ipa_profile_write_edge_summary): New function.
2610 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2611 (ipa_profile_dump_all_summaries): New function.
2612 (ipa_profile_read_edge_summary): New function.
2613 (ipa_profile_read_summary_section): New function.
2614 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2615 (ipa_profile): Generate num_speculative_call_targets from
2617 * ipa-ref.h (speculative_id): New variable.
2618 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2619 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2620 common_target_probability. Stream out speculative_id and
2621 num_speculative_call_targets.
2622 (input_edge): Likewise.
2623 * predict.c (dump_prediction): Remove edges count assert to be
2625 * symtab.c (symtab_node::create_reference): Init speculative_id.
2626 (symtab_node::clone_references): Clone speculative_id.
2627 (symtab_node::clone_referring): Clone speculative_id.
2628 (symtab_node::clone_reference): Clone speculative_id.
2629 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2630 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2631 if indirect call contains multiple speculative targets.
2632 * value-prof.h (check_ic_target): Remove.
2633 * value-prof.c (gimple_value_profile_transformations):
2634 Use void function gimple_ic_transform.
2635 * value-prof.c (gimple_ic_transform): Handle topn case.
2636 Fix comment typos. Change it to a void function.
2638 2020-01-13 Andrew Pinski <apinski@marvell.com>
2640 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2641 (octeontx2t98): New define.
2642 (octeontx2t96): New define.
2643 (octeontx2t93): New define.
2644 (octeontx2f95): New define.
2645 (octeontx2f95n): New define.
2646 (octeontx2f95mm): New define.
2647 * config/aarch64/aarch64-tune.md: Regenerate.
2648 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2650 2020-01-13 Jason Merrill <jason@redhat.com>
2652 PR c++/33799 - destroy return value if local cleanup throws.
2653 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2655 2020-01-13 Martin Liska <mliska@suse.cz>
2657 * ipa-cp.c (get_max_overall_size): Use newly
2658 renamed param param_ipa_cp_unit_growth.
2659 * params.opt: Remove legacy param name.
2661 2020-01-13 Martin Sebor <msebor@redhat.com>
2663 PR tree-optimization/93213
2664 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2665 stores to be eliminated.
2667 2020-01-13 Martin Liska <mliska@suse.cz>
2669 * opts.c (print_help): Do not print CL_PARAM
2670 and CL_WARNING for CL_OPTIMIZATION.
2672 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2675 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2676 depending on optimization settings.
2678 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2680 PR tree-optimization/90838
2681 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2682 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2683 argument rather than to initialize temporary for targets that
2684 don't use the mode argument at all. Initialize ctzval to avoid
2687 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2689 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2690 * tree-core.h: Document it.
2691 * gimplify.c (gimplify_omp_workshare): Set it.
2692 * omp-low.c (lower_omp_target): Use it.
2693 * tree-pretty-print.c (dump_omp_clause): Print it.
2695 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2696 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2698 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2700 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2701 * common.opt (fdiagnostics-path-format=): New option.
2702 (diagnostic_path_format): New enum.
2703 (fdiagnostics-show-path-depths): New option.
2704 * coretypes.h (diagnostic_event_id_t): New forward decl.
2705 * diagnostic-color.c (color_dict): Add "path".
2706 * diagnostic-event-id.h: New file.
2707 * diagnostic-format-json.cc (json_from_expanded_location): Make
2709 (json_end_diagnostic): Call context->make_json_for_path if it
2710 exists and the diagnostic has a path.
2711 (diagnostic_output_format_init): Clear context->print_path.
2712 * diagnostic-path.h: New file.
2713 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2714 when printing a run of events in a diagnostic_path so that they
2715 all get the same color.
2716 (layout::m_diagnostic_path_p): New field.
2717 (layout::layout): Initialize it.
2718 (layout::print_any_labels): Don't colorize the label text for an
2719 event in a diagnostic_path.
2720 (gcc_rich_location::add_location_if_nearby): Add
2721 "restrict_to_current_line_spans" and "label" params. Pass the
2722 former to layout.maybe_add_location_range; pass the latter
2723 when calling add_range.
2724 * diagnostic.c: Include "diagnostic-path.h".
2725 (diagnostic_initialize): Initialize context->path_format and
2726 context->show_path_depths.
2727 (diagnostic_show_any_path): New function.
2728 (diagnostic_path::interprocedural_p): New function.
2729 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2730 (simple_diagnostic_path::num_events): New function.
2731 (simple_diagnostic_path::get_event): New function.
2732 (simple_diagnostic_path::add_event): New function.
2733 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2734 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2735 (debug): New overload taking a diagnostic_path *.
2736 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2737 * diagnostic.h (enum diagnostic_path_format): New enum.
2738 (json::value): New forward decl.
2739 (diagnostic_context::path_format): New field.
2740 (diagnostic_context::show_path_depths): New field.
2741 (diagnostic_context::print_path): New callback field.
2742 (diagnostic_context::make_json_for_path): New callback field.
2743 (diagnostic_show_any_path): New decl.
2744 (json_from_expanded_location): New decl.
2745 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2746 (-fdiagnostics-show-path-depths): New option.
2747 (-fdiagnostics-color): Add "path" to description of default
2748 GCC_COLORS; describe it.
2749 (-fdiagnostics-format=json): Document how diagnostic paths are
2750 represented in the JSON output format.
2751 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2752 Add optional params "restrict_to_current_line_spans" and "label".
2753 * opts.c (common_handle_option): Handle
2754 OPT_fdiagnostics_path_format_ and
2755 OPT_fdiagnostics_show_path_depths.
2756 * pretty-print.c: Include "diagnostic-event-id.h".
2757 (pp_format): Implement "%@" format code for printing
2758 diagnostic_event_id_t *.
2759 (selftest::test_pp_format): Add tests for "%@".
2760 * selftest-run-tests.c (selftest::run_tests): Call
2761 selftest::tree_diagnostic_path_cc_tests.
2762 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2763 * toplev.c (general_init): Initialize global_dc->path_format and
2764 global_dc->show_path_depths.
2765 * tree-diagnostic-path.cc: New file.
2766 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2767 non-static. Drop "diagnostic" param in favor of storing the
2768 original value of "where" and re-using it.
2769 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2770 maybe_unwind_expanded_macro_loc.
2771 (tree_diagnostics_defaults): Initialize context->print_path and
2772 context->make_json_for_path.
2773 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2775 (default_tree_make_json_for_path): New decl.
2776 (maybe_unwind_expanded_macro_loc): New decl.
2778 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2780 PR tree-optimization/93210
2781 * fold-const.h (native_encode_initializer,
2782 can_native_interpret_type_p): Declare.
2783 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2785 (native_encode_initializer): New function, moved from dwarf2out.c.
2786 Adjust to native_encode_expr compatible arguments, including dry-run
2787 and partial extraction modes. Don't handle STRING_CST.
2788 (can_native_interpret_type_p): No longer static.
2789 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2790 offset / BITS_PER_UNIT fits into int and don't call it if
2791 can_native_interpret_type_p fails. If suboff is NULL and for
2792 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2793 native_encode_initializer.
2794 (fold_const_aggregate_ref_1): Formatting fix.
2795 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2796 (tree_add_const_value_attribute): Adjust caller.
2798 PR tree-optimization/90838
2799 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2800 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2801 CTZ_DEFINED_VALUE_AT_ZERO.
2803 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2806 * lra-constraints.c (match_reload): Permit input operands have the
2807 same mode as output while other input operands have a different
2810 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
2812 PR tree-optimization/90838
2813 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2814 (check_ctz_string): Likewise.
2815 (optimize_count_trailing_zeroes): Likewise.
2816 (simplify_count_trailing_zeroes): Likewise.
2817 (pass_forwprop::execute): Try ctz simplification.
2818 * match.pd: Add matching for ctz idioms.
2820 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2822 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2824 (aarch64_invalid_unary_op): New function for target hook.
2825 (aarch64_invalid_binary_op): New function for target hook.
2827 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2829 * config.gcc: Add arm_bf16.h.
2830 * config/aarch64/aarch64-builtins.c
2831 (aarch64_simd_builtin_std_type): Add BFmode.
2832 (aarch64_init_simd_builtin_types): Define element types for vector
2834 (aarch64_init_bf16_types): New function.
2835 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2836 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2838 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2839 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2841 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2842 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2843 * config/aarch64/aarch64.c
2844 (aarch64_classify_vector_mode): Add support for BF types.
2845 (aarch64_gimplify_va_arg_expr): Add support for BF types.
2846 (aarch64_vq_mode): Add support for BF types.
2847 (aarch64_simd_container_mode): Add support for BF types.
2848 (aarch64_mangle_type): Add support for BF scalar type.
2849 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2850 * config/aarch64/arm_bf16.h: New file.
2851 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2852 * config/aarch64/iterators.md: Add BF types to mode attributes.
2853 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2855 2020-01-10 Jason Merrill <jason@redhat.com>
2857 PR c++/93173 - incorrect tree sharing.
2858 * gimplify.c (copy_if_shared): No longer static.
2859 * gimplify.h: Declare it.
2861 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2863 * doc/invoke.texi (-msve-vector-bits=): Document that
2864 -msve-vector-bits=128 now generates VL-specific code for
2865 little-endian targets.
2866 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2867 build_vector_type_for_mode to construct the data vector types.
2868 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2869 VL-specific code for -msve-vector-bits=128 on little-endian targets.
2870 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2871 for 128-bit vectors.
2873 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2875 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2878 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2880 * config/aarch64/aarch64-builtins.c
2881 (aarch64_builtin_vectorized_function): Check for specific vector modes,
2882 rather than checking the number of elements and the element mode.
2884 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2886 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2887 get_related_vectype_for_scalar_type rather than build_vector_type
2888 to create the index type for a conditional reduction.
2890 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2892 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
2893 for any type of gather or scatter, including strided accesses.
2895 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2897 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
2900 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2902 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
2904 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
2905 parameter and its use to reset DR_OFFSET's.
2906 (vect_transform_loop): Remove orig_drs_init argument.
2907 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
2908 member of dr_vec_info rather than the offset of the associated
2909 data_reference's innermost_loop_behavior.
2910 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
2911 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
2912 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
2913 get_dr_vinfo_offset.
2914 (vectorizable_store): Likewise.
2915 (vectorizable_load): Likewise.
2917 2020-01-10 Richard Biener <rguenther@suse.de>
2919 * gimple-ssa-store-merging
2920 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
2922 2020-01-10 Martin Liska <mliska@suse.cz>
2925 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
2926 encapsulation that was there before r280040.
2928 2020-01-10 Richard Biener <rguenther@suse.de>
2931 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
2932 sequences to avoid walking them again for secondary opportunities.
2933 (pass_lower_eh_dispatch::execute): Instead actually insert
2936 2020-01-10 Richard Biener <rguenther@suse.de>
2939 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
2940 (cleanup_all_empty_eh): Walk landing pads in reverse order to
2941 avoid quadraticness.
2943 2020-01-10 Martin Jambor <mjambor@suse.cz>
2945 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
2946 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
2947 to get param_ipa_sra_max_replacements.
2948 (param_splitting_across_edge): Pass the caller to
2949 pull_accesses_from_callee.
2951 2020-01-10 Martin Jambor <mjambor@suse.cz>
2953 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
2954 * ipa-cp.c (max_new_size): Removed.
2955 (orig_overall_size): New variable.
2956 (get_max_overall_size): New function.
2957 (estimate_local_effects): Use it. Adjust dump.
2958 (decide_about_value): Likewise.
2959 (ipcp_propagate_stage): Do not calculate max_new_size, just store
2960 orig_overall_size. Adjust dump.
2961 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
2963 2020-01-10 Martin Jambor <mjambor@suse.cz>
2965 * params.opt (param_ipa_max_agg_items): Mark as Optimization
2966 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
2967 instead of param_ipa_max_agg_items.
2968 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
2969 optimization info for the callee.
2971 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
2973 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
2974 markers if debug_inline_points is false.
2976 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2978 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
2980 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
2981 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
2982 aarch64-sve-builtins-sve2.h.
2983 (aarch64-sve-builtins-sve2.o): New rule.
2984 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
2985 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
2986 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
2987 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
2988 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
2989 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
2991 * config/aarch64/aarch64-sve.md: Update comments with SVE2
2992 instructions that are handled here.
2993 (@cond_asrd<mode>): Generalize to...
2994 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
2995 (*cond_asrd<mode>_2): Generalize to...
2996 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
2997 (*cond_asrd<mode>_z): Generalize to...
2998 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
2999 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3000 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3001 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3002 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3004 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3005 (@aarch64_scatter_stnt<mode>): Likewise.
3006 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3007 (@aarch64_mul_lane_<mode>): Likewise.
3008 (@aarch64_sve_suqadd<mode>_const): Likewise.
3009 (*<sur>h<addsub><mode>): Generalize to...
3010 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3012 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3013 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3014 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3015 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3016 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3017 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3018 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3019 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3020 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3021 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3022 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3023 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3024 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3025 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3026 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3027 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3028 (@aarch64_sve2_xar<mode>): Likewise.
3029 (@aarch64_sve2_bcax<mode>): Likewise.
3030 (*aarch64_sve2_eor3<mode>): Rename to...
3031 (@aarch64_sve2_eor3<mode>): ...this.
3032 (@aarch64_sve2_bsl<mode>): New expander.
3033 (@aarch64_sve2_nbsl<mode>): Likewise.
3034 (@aarch64_sve2_bsl1n<mode>): Likewise.
3035 (@aarch64_sve2_bsl2n<mode>): Likewise.
3036 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3037 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3038 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3039 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3040 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3041 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3042 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3043 (<su>mull<bt><Vwide>): Generalize to...
3044 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3046 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3047 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3048 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3049 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3050 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3051 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3052 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3053 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3054 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3055 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3056 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3057 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3058 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3059 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3060 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3061 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3062 (<SHRNB:r>shrnb<mode>): Generalize to...
3063 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3065 (<SHRNT:r>shrnt<mode>): Generalize to...
3066 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3068 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3069 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3070 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3071 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3072 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3073 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3074 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3075 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3076 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3077 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3078 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3079 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3080 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3081 (@aarch64_sve2_cvtnt<mode>): Likewise.
3082 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3083 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3084 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3085 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3086 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3087 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3088 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3089 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3090 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3091 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3092 (@aarch64_sve2_pmul<mode>): Likewise.
3093 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3094 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3095 (@aarch64_sve2_tbl2<mode>): Likewise.
3096 (@aarch64_sve2_tbx<mode>): Likewise.
3097 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3098 (@aarch64_sve2_histcnt<mode>): Likewise.
3099 (@aarch64_sve2_histseg<mode>): Likewise.
3100 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3101 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3102 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3103 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3104 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3105 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3106 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3107 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3108 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3109 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3110 (SVE2_PMULL_PAIR_I): New mode iterators.
3111 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3112 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3113 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3114 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3115 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3116 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3117 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3118 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3119 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3120 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3121 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3122 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3123 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3124 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3125 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3126 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3127 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3128 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3129 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3130 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3131 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3132 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3133 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3134 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3135 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3136 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3137 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3138 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3139 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3140 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3141 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3143 (VNARROW, Ventype): New mode attributes.
3144 (Vewtype): Handle VNx2DI. Fix typo in comment.
3145 (VDOUBLE): New mode attribute.
3146 (sve_lane_con): Handle VNx8HI.
3147 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3148 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3149 (sve_int_op, sve_int_op_rev): Handle the above codes.
3150 (sve_pred_int_rhs2_operand): Likewise.
3151 (MULLBT, SHRNB, SHRNT): Delete.
3152 (SVE_INT_SHIFT_IMM): New int iterator.
3153 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3154 and UNSPEC_WHILEHS for TARGET_SVE2.
3155 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3156 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3157 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3158 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3159 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3160 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3161 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3162 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3163 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3164 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3165 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3166 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3167 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3168 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3169 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3170 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3171 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3172 (optab): Handle the new unspecs.
3173 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3175 (lr): Handle the new unspecs.
3177 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3178 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3179 (sve_int_qsub_op): New int attributes.
3180 (sve_fp_op, rot): Handle the new unspecs.
3181 * config/aarch64/aarch64-sve-builtins.h
3182 (function_resolver::require_matching_pointer_type): Declare.
3183 (function_resolver::resolve_unary): Add an optional boolean argument.
3184 (function_resolver::finish_opt_n_resolution): Add an optional
3185 type_suffix_index argument.
3186 (gimple_folder::redirect_call): Declare.
3187 (gimple_expander::prepare_gather_address_operands): Add an optional
3189 * config/aarch64/aarch64-sve-builtins.cc: Include
3190 aarch64-sve-builtins-sve2.h.
3191 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3192 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3193 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3194 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3195 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3196 (TYPES_s_unsigned): Likewise.
3197 (TYPES_s_integer): Use TYPES_s_unsigned.
3198 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3199 (TYPES_sd_integer): Use them.
3200 (TYPES_d_unsigned): New macro.
3201 (TYPES_d_integer): Use it.
3202 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3203 (TYPES_cvt_narrow): Likewise.
3204 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3205 (preds_mx): New variable.
3206 (function_builder::add_overloaded_function): Allow the new feature
3207 set to be more restrictive than the original one.
3208 (function_resolver::infer_pointer_type): Remove qualifiers from
3209 the pointer type before printing it.
3210 (function_resolver::require_matching_pointer_type): New function.
3211 (function_resolver::resolve_sv_displacement): Handle functions
3212 that don't support 32-bit vector indices or svint32_t vector offsets.
3213 (function_resolver::finish_opt_n_resolution): Take the inferred type
3214 as a separate argument.
3215 (function_resolver::resolve_unary): Optionally treat all forms in
3216 the same way as normal merging functions.
3217 (gimple_folder::redirect_call): New function.
3218 (function_expander::prepare_gather_address_operands): Add an argument
3219 that says whether scaled forms are available. If they aren't,
3220 handle scaling of vector indices and don't add the extension and
3222 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3223 fall back to using cond_* instead.
3224 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3225 Split out the member variables into...
3226 (rtx_code_function_base): ...this new base class.
3227 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3228 (unspec_based_function): Split out the member variables into...
3229 (unspec_based_function_base): ...this new base class.
3230 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3231 (unspec_based_function_exact_insn): New class.
3232 (unspec_based_add_function, unspec_based_add_lane_function)
3233 (unspec_based_lane_function, unspec_based_pred_function)
3234 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3235 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3236 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3238 (unspec_based_fused_function): New class.
3239 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3240 (unspec_based_fused_lane_function): New class.
3241 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3243 (CODE_FOR_MODE1): New macro.
3244 (fixed_insn_function): New class.
3245 (while_comparison): Likewise.
3246 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3247 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3248 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3249 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3250 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3251 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3252 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3253 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3254 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3255 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3256 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3257 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3258 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3259 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3260 Also add an initial argument for unary_convert_narrowt, regardless
3261 of the predication type.
3262 (build_32_64): Allow loads and stores to specify MODE_none.
3263 (build_sv_index64, build_sv_uint_offset): New functions.
3264 (long_type_suffix): New function.
3265 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3266 (binary_imm_long_base, load_gather_sv_base): Likewise.
3267 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3268 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3269 (unary_narrowb_base, unary_narrowt_base): Likewise.
3270 (binary_long_lane_def, binary_long_lane): New shape.
3271 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3272 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3273 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3274 (binary_to_uint_def, binary_to_uint): Likewise.
3275 (binary_wide_def, binary_wide): Likewise.
3276 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3277 (compare_def, compare): Likewise.
3278 (compare_ptr_def, compare_ptr): Likewise.
3279 (load_ext_gather_index_restricted_def,
3280 load_ext_gather_index_restricted): Likewise.
3281 (load_ext_gather_offset_restricted_def,
3282 load_ext_gather_offset_restricted): Likewise.
3283 (load_gather_sv_def): Inherit from load_gather_sv_base.
3284 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3285 (shift_left_imm_def, shift_left_imm): Likewise.
3286 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3287 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3288 (store_scatter_index_restricted_def,
3289 store_scatter_index_restricted): Likewise.
3290 (store_scatter_offset_restricted_def,
3291 store_scatter_offset_restricted): Likewise.
3292 (tbl_tuple_def, tbl_tuple): Likewise.
3293 (ternary_long_lane_def, ternary_long_lane): Likewise.
3294 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3295 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3296 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3297 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3298 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3299 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3300 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3301 (ternary_uint_def, ternary_uint): Likewise.
3302 (unary_convert): Fix typo in comment.
3303 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3304 (unary_long_def, unary_long): Likewise.
3305 (unary_narrowb_def, unary_narrowb): Likewise.
3306 (unary_narrowt_def, unary_narrowt): Likewise.
3307 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3308 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3309 (unary_to_int_def, unary_to_int): Likewise.
3310 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3311 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3312 (svasrd_impl): Delete.
3313 (svcadd_impl::expand): Handle integer operations too.
3314 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3315 new functions to derive the unspec numbers.
3316 (svmla_svmls_lane_impl): Replace with...
3317 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3318 integer operations too.
3319 (svwhile_impl): Rename to...
3320 (svwhilelx_impl): ...this and inherit from while_comparison.
3321 (svasrd): Use unspec_based_function.
3322 (svmla_lane): Use svmla_lane_impl.
3323 (svmls_lane): Use svmls_lane_impl.
3324 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3325 (svwhilele, svwhilelt): Use svwhilelx_impl.
3326 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3327 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3328 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3329 * config/aarch64/aarch64-sve-builtins.def: Include
3330 aarch64-sve-builtins-sve2.def.
3332 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3334 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3335 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3336 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3337 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3338 immediates as well as vector ones.
3339 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3340 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3341 (aarch64_sve_qsub_immediate): Update calls accordingly.
3343 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3345 * config/aarch64/aarch64-sve2.md: Add banner comments.
3346 (<su>mulh<r>s<mode>3): Move further up file.
3347 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3348 (*aarch64_sve2_sra<mode>): Move further down file.
3349 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3351 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3353 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3355 (while_optab_cmp): Handle them.
3356 * config/aarch64/aarch64-sve.md
3357 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3358 and add a "@" marker.
3359 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3360 instead of gen_aarch64_sve2_while_ptest.
3361 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3363 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3365 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3366 (UNSPEC_WHILELE): ...this.
3367 (UNSPEC_WHILE_LO): Rename to...
3368 (UNSPEC_WHILELO): ...this.
3369 (UNSPEC_WHILE_LS): Rename to...
3370 (UNSPEC_WHILELS): ...this.
3371 (UNSPEC_WHILE_LT): Rename to...
3372 (UNSPEC_WHILELT): ...this.
3373 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3374 (cmp_op, while_optab_cmp): Likewise.
3375 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3376 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3377 (svwhilelt): Likewise.
3379 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3381 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3382 (unary_to_uint): Define.
3383 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3384 (unary_count): Rename to...
3385 (unary_to_uint_def, unary_to_uint): ...this.
3386 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3388 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3390 * config/aarch64/aarch64-sve-builtins-functions.h
3391 (code_for_mode_function): New class.
3392 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3393 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3394 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3395 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3396 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3398 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3400 * config/aarch64/iterators.md (addsub): New code attribute.
3401 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3403 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3404 in the asm string and attributes. Fix indentation.
3405 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3407 (@aarch64_sve_<optab><mode>): ...this.
3408 * config/aarch64/aarch64-sve-builtins.h
3409 (function_expander::expand_signed_unpred_op): Delete.
3410 * config/aarch64/aarch64-sve-builtins.cc
3411 (function_expander::expand_signed_unpred_op): Likewise.
3412 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3413 try using code_for_aarch64_sve instead.
3414 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3415 (svqsub_impl): Likewise.
3416 (svqadd, svqsub): Use rtx_code_function instead.
3418 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3420 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3421 (HADDSUB, sur, addsub): Remove them.
3423 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3425 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3426 functions the same way, rather than singling out those that
3427 aren't mapped directly to optabs.
3429 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3431 * target.def (compatible_vector_types_p): New target hook.
3432 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3433 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3434 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3435 * doc/tm.texi: Regenerate.
3436 * gimple-expr.c: Include target.h.
3437 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3438 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3440 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3441 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3442 Use the original predicate if it already has a suitable type.
3444 2020-01-09 Martin Jambor <mjambor@suse.cz>
3446 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3447 resolve_speculation and redirect_call_stmt_to_callee static. Change
3448 return type of set_call_stmt to cgraph_edge *.
3449 * auto-profile.c (afdo_indirect_call): Adjust call to
3450 redirect_call_stmt_to_callee.
3451 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3452 make the this pointer explicit, adjust self-recursive calls and the
3453 call top make_direct. Return the resulting edge.
3454 (cgraph_edge::remove): Make this pointer explicit.
3455 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3456 (cgraph_edge::make_direct): Likewise, adjust call to
3457 resolve_speculation.
3458 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3459 call to set_call_stmt.
3460 (cgraph_update_edges_for_call_stmt_node): Update call to
3461 set_call_stmt and remove.
3462 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3463 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3464 (cgraph_node::create_edge_including_clones): Moved "first" definition
3465 of edge to the block where it was used. Adjusted calls to
3467 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3468 cgraph_edge::remove.
3469 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3470 make_direct and redirect_call_stmt_to_callee.
3471 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3472 resolve_speculation and make_direct.
3473 * ipa-inline-transform.c (inline_transform): Adjust call to
3474 redirect_call_stmt_to_callee.
3475 (check_speculations_1):: Adjust call to resolve_speculation.
3476 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3477 resolve-speculation.
3478 (inline_small_functions): Adjust call to resolve_speculation.
3479 (ipa_inline): Likewise.
3480 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3482 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3483 safe with regards to edge removal, adjust calls to
3484 redirect_call_stmt_to_callee.
3485 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3486 and redirect_call_stmt_to_callee.
3487 * multiple_target.c (create_dispatcher_calls): Adjust call to
3488 redirect_call_stmt_to_callee
3489 (redirect_to_specific_clone): Likewise.
3490 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3491 Adjust calls to cgraph_edge::remove.
3492 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3493 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3494 (expand_call_inline): Adjust call to cgraph_edge::remove.
3496 2020-01-09 Martin Liska <mliska@suse.cz>
3498 * params.opt: Set Optimization for
3499 param_max_speculative_devirt_maydefs.
3501 2020-01-09 Martin Sebor <msebor@redhat.com>
3505 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3507 2020-01-09 Martin Liska <mliska@suse.cz>
3509 * auto-profile.c (auto_profile): Use opt_for_fn
3511 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3512 (propagate_vals_across_arith_jfunc): Likewise.
3513 (hint_time_bonus): Likewise.
3514 (incorporate_penalties): Likewise.
3515 (good_cloning_opportunity_p): Likewise.
3516 (perform_estimation_of_a_value): Likewise.
3517 (estimate_local_effects): Likewise.
3518 (ipcp_propagate_stage): Likewise.
3519 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3520 (set_switch_stmt_execution_predicate): Likewise.
3521 (analyze_function_body): Likewise.
3522 * ipa-inline-analysis.c (offline_size): Likewise.
3523 * ipa-inline.c (early_inliner): Likewise.
3524 * ipa-prop.c (ipa_analyze_node): Likewise.
3525 (ipcp_transform_function): Likewise.
3526 * ipa-sra.c (process_scan_results): Likewise.
3527 (ipa_sra_summarize_function): Likewise.
3528 * params.opt: Rename ipcp-unit-growth to
3529 ipa-cp-unit-growth. Add Optimization for various
3530 IPA-related parameters.
3532 2020-01-09 Richard Biener <rguenther@suse.de>
3535 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3537 2020-01-09 Richard Biener <rguenther@suse.de>
3539 PR tree-optimization/93040
3540 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3542 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3544 * common/config/avr/avr-common.c (avr_option_optimization_table)
3545 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3547 2020-01-09 Martin Liska <mliska@suse.cz>
3549 * cgraphclones.c (symbol_table::materialize_all_clones):
3550 Use cgraph_node::dump_name.
3552 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3555 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3556 output_operand_lossage instead of gcc_unreachable.
3557 * doc/md.texi (riscv f constraint): Fix typo.
3560 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3561 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3562 CONST_SCALAR_INT_P instead of CONST_INT_P.
3563 (*subv<mode>4_1): Rename to ...
3564 (subv<mode>4_1): ... this.
3565 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3566 define_insn_and_split patterns.
3567 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3570 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3572 * vec.c (class selftest::count_dtor): New class.
3573 (selftest::test_auto_delete_vec): New test.
3574 (selftest::vec_c_tests): Call it.
3575 * vec.h (class auto_delete_vec): New class template.
3576 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3578 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3580 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3582 2020-01-08 Jim Wilson <jimw@sifive.com>
3584 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3585 use of TLS_MODEL_LOCAL_EXEC when not pic.
3587 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3589 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3592 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3595 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3596 *stack_protect_set_3 peephole2): Also check that the second
3597 insns source is general_operand.
3600 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3601 predicate for output operand instead of register_operand.
3602 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3603 memory destination and non-memory operands[2].
3605 2020-01-08 Martin Liska <mliska@suse.cz>
3607 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3608 ::dump_asm_name instead of (::name or ::asm_name).
3609 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3610 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3611 (analyze_functions): Likewise.
3612 (expand_all_functions): Likewise.
3613 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3614 (propagate_bits_across_jump_function): Likewise.
3615 (dump_profile_updates): Likewise.
3616 (ipcp_store_bits_results): Likewise.
3617 (ipcp_store_vr_results): Likewise.
3618 * ipa-devirt.c (dump_targets): Likewise.
3619 * ipa-fnsummary.c (analyze_function_body): Likewise.
3620 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3621 (process_hsa_functions): Likewise.
3622 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3623 (set_alias_uids): Likewise.
3624 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3625 * ipa-inline.c (recursive_inlining): Likewise.
3626 (inline_to_all_callers_1): Likewise.
3627 (ipa_inline): Likewise.
3628 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3629 (ipa_propagate_frequency): Likewise.
3630 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3631 (remove_described_reference): Likewise.
3632 * ipa-pure-const.c (worse_state): Likewise.
3633 (check_retval_uses): Likewise.
3634 (analyze_function): Likewise.
3635 (propagate_pure_const): Likewise.
3636 (propagate_nothrow): Likewise.
3637 (dump_malloc_lattice): Likewise.
3638 (propagate_malloc): Likewise.
3639 (pass_local_pure_const::execute): Likewise.
3640 * ipa-visibility.c (optimize_weakref): Likewise.
3641 (function_and_variable_visibility): Likewise.
3642 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3643 (ipa_discover_variable_flags): Likewise.
3644 * lto-streamer-out.c (output_function): Likewise.
3645 (output_constructor): Likewise.
3646 * tree-inline.c (copy_bb): Likewise.
3647 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3648 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3650 2020-01-08 Richard Biener <rguenther@suse.de>
3653 * tree-eh.c (sink_clobbers): Update virtual operands for
3654 the first and last stmt only. Add a dry-run capability.
3655 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3656 after CFG manipulations and in RPO order to catch all
3657 secondary opportunities reliably.
3659 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3662 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3664 2019-01-08 Richard Biener <rguenther@suse.de>
3667 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3668 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3669 virtual operand, also updating SSA use.
3670 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3671 Update stmt after resetting virtual operand.
3672 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3673 * gimple-iterator.c (gsi_remove): When not removing the stmt
3674 permanently do not delink immediate uses or mark the stmt modified.
3676 2020-01-08 Martin Liska <mliska@suse.cz>
3678 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3679 (ipa_call_context::estimate_size_and_time): Likewise.
3680 (inline_analyze_function): Likewise.
3682 2020-01-08 Martin Liska <mliska@suse.cz>
3684 * cgraph.c (cgraph_node::dump): Use systematically
3687 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3689 Add -nodevicespecs option for avr.
3692 * config/avr/avr.opt (-nodevicespecs): New driver option.
3693 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3694 "-specs=device-specs/..." if that option is not set.
3695 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3697 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3699 Implement 64-bit double functions for avr.
3702 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3703 --with-double-comparison.
3704 * doc/install.texi: Document them.
3705 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3706 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3707 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3708 * doc/invoke.texi (AVR Built-in Macros): Document them.
3709 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3710 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3711 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3713 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3716 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3717 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3718 when only building rm-profile multilibs.
3720 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3723 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3724 lattice for a value to check.
3725 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3726 finite propagation in self-recursive scc.
3728 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3730 * ipa-inline.c (caller_growth_limits): Restore the AND.
3732 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3734 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3735 (VEC_ALLREG_ALT): New iterator.
3736 (VEC_ALLREG_INT_MODE): New iterator.
3737 (VCMP_MODE): New iterator.
3738 (VCMP_MODE_INT): New iterator.
3739 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3740 (vec_cmp<u>v64qidi): New define_expand.
3741 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3742 (vec_cmpu<mode>di_exec): New define_expand.
3743 (vec_cmp<u>v64qidi_exec): New define_expand.
3744 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3745 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3746 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3747 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3748 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3749 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3750 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3751 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3752 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3753 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3755 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3756 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3758 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3760 * config/gcn/constraints.md (DA): Update description and match.
3762 (Db): New constraint.
3763 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3765 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3766 Implement 'Db' mixed immediate type.
3767 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3768 (addcv64si3_dup<exec_vcc>): Delete.
3769 (subcv64si3<exec_vcc>): Rework constraints.
3770 (addv64di3): Rework constraints.
3771 (addv64di3_exec): Rework constraints.
3772 (subv64di3): Rework constraints.
3773 (addv64di3_dup): Delete.
3774 (addv64di3_dup_exec): Delete.
3775 (addv64di3_zext): Rework constraints.
3776 (addv64di3_zext_exec): Rework constraints.
3777 (addv64di3_zext_dup): Rework constraints.
3778 (addv64di3_zext_dup_exec): Rework constraints.
3779 (addv64di3_zext_dup2): Rework constraints.
3780 (addv64di3_zext_dup2_exec): Rework constraints.
3781 (addv64di3_sext_dup2): Rework constraints.
3782 (addv64di3_sext_dup2_exec): Rework constraints.
3784 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3786 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3787 existing target checks.
3789 2020-01-07 Richard Biener <rguenther@suse.de>
3791 * doc/install.texi: Bump minimal supported MPC version.
3793 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3795 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3796 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3797 * langhooks.c: Include stor-layout.h.
3798 (lhd_simulate_enum_decl): New function.
3799 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3800 handle_arm_sve_h for the LTO frontend.
3801 (register_vector_type): Cope with null returns from pushdecl.
3803 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3805 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3806 (aarch64_sve::nvectors_if_data_type): Replace with...
3807 (aarch64_sve::builtin_type_p): ...this.
3808 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3809 (find_vector_type): Delete.
3810 (add_sve_type_attribute): New function.
3811 (lookup_sve_type_attribute): Likewise.
3812 (register_builtin_types): Add an "SVE type" attribute to each type.
3813 (register_tuple_type): Likewise.
3814 (svbool_type_p, nvectors_if_data_type): Delete.
3815 (mangle_builtin_type): Use lookup_sve_type_attribute.
3816 (builtin_type_p): Likewise. Add an overload that returns the
3817 number of constituent vector and predicate registers.
3818 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3819 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3820 instead of aarch64_sve_argument_p.
3821 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3822 (aarch64_pass_by_reference): Likewise.
3823 (aarch64_function_value_1): Likewise.
3824 (aarch64_return_in_memory): Likewise.
3825 (aarch64_layout_arg): Likewise.
3827 2020-01-07 Jakub Jelinek <jakub@redhat.com>
3829 PR tree-optimization/93156
3830 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3831 least significant bit is always clear.
3833 PR tree-optimization/93118
3834 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
3835 simplifier with two intermediate conversions.
3837 2020-01-07 Martin Liska <mliska@suse.cz>
3839 * params.opt: Add Optimization for various parameters.
3841 2020-01-07 Martin Liska <mliska@suse.cz>
3844 * doc/extend.texi: Explain cloning for target_clone
3847 2020-01-07 Martin Liska <mliska@suse.cz>
3849 PR tree-optimization/92860
3850 * common.opt: Make in Optimization option
3851 as it is affected by -O0, which is an Optimization
3853 * tree-inline.c (tree_inlinable_function_p):
3854 Use opt_for_fn for warn_inline.
3855 (expand_call_inline): Likewise.
3857 2020-01-07 Martin Liska <mliska@suse.cz>
3859 PR tree-optimization/92860
3860 * common.opt: Make flag_ree as optimization
3863 2020-01-07 Martin Liska <mliska@suse.cz>
3865 PR optimization/92860
3866 * params.opt: Mark param_min_crossjump_insns with Optimization
3869 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
3871 * ipa-inline-analysis.c (estimate_growth): Fix typo.
3872 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3874 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
3876 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3877 helper function to return the valid addressing formats for a given
3878 hard register and mode.
3879 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3881 * config/rs6000/constraints.md (Q constraint): Update
3883 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3886 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3887 Use 'Q' for doing vector extract from memory.
3888 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
3890 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
3891 doing vector extract from memory.
3892 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
3893 extract from memory.
3895 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
3896 for the offset being 34-bits when -mcpu=future is used.
3898 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
3900 * config/pa/pa.md: Revert change to use ordered_comparison_operator
3901 instead of cmpib_comparison_operator in cmpib patterns.
3902 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
3903 of cmpib_comparison_operator. Revise comment.
3905 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3907 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
3908 in an IFN_DIV_POW2 node to be equal.
3910 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3912 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
3913 (vect_check_scalar_mask): ...this.
3914 (vectorizable_store, vectorizable_load): Update call accordingly.
3915 (vectorizable_call): Use vect_check_scalar_mask to check the mask
3916 argument in calls to conditional internal functions.
3918 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3920 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
3921 '0' matching inputs.
3922 (subv64di3_exec): Likewise.
3924 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
3926 * config/mips/mips.c (vr4130_align_insns): Fix typo.
3927 * doc/md.texi (movstr): Likewise.
3929 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3931 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
3934 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3936 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
3938 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
3939 to a temporary file and use move-if-change to update the real
3940 file where necessary.
3942 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3944 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
3945 rather than Upa for CPY /M.
3947 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3949 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
3952 2020-01-06 Martin Liska <mliska@suse.cz>
3954 PR tree-optimization/92860
3955 * params.opt: Mark param_max_combine_insns with Optimization
3958 2020-01-05 Jakub Jelinek <jakub@redhat.com>
3961 * config/i386/i386.md (SWIDWI): New mode iterator.
3962 (DWI, dwi): Add TImode variants.
3963 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
3964 <general_hilo_operand> instead of <general_operand>. Use
3965 CONST_SCALAR_INT_P instead of CONST_INT_P.
3966 (*addv<mode>4_1): Rename to ...
3967 (addv<mode>4_1): ... this.
3968 (QWI): New mode attribute.
3969 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3970 define_insn_and_split patterns.
3971 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3973 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
3974 <general_hilo_operand> instead of <general_operand>.
3975 (*addcarry<mode>_1): New define_insn.
3976 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
3978 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
3980 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
3981 Use "call" instead of "set".
3983 2020-01-03 Martin Jambor <mjambor@suse.cz>
3986 * ipa-cp.c (print_all_lattices): Skip functions without info.
3988 2020-01-03 Jakub Jelinek <jakub@redhat.com>
3991 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
3992 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
3993 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
3994 for 'e' simd clones.
3997 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
3999 (mprefer-vector-width=): Add Save.
4000 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4001 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4002 (ix86_debug_options, ix86_function_specific_print): Adjust
4003 ix86_target_string callers.
4004 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4005 (ix86_valid_target_attribute_tree): Likewise.
4006 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4007 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4008 ix86_target_string caller.
4011 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4012 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4013 instead of gen_int_shift_amount + convert_modes.
4015 PR rtl-optimization/93088
4016 * loop-iv.c (find_single_def_src): Punt after looking through
4017 128 reg copies for regs with single definitions. Move definitions
4020 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4022 * config/arm/arm-c.c (arm_cpu_builtins): Define
4023 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4024 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4025 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4026 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4027 * config/arm/arm-tables.opt: Regenerated.
4028 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4029 arm_arch_i8mm and arm_arch_bf16 when enabled.
4030 * config/arm/arm.h (TARGET_I8MM): New macro.
4031 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4032 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4033 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4034 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4035 (v8_6_a_simd_variants): New.
4036 (v8_*_a_simd_variants): Add i8mm and bf16.
4037 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4039 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4042 * predict.c (compute_function_frequency): Don't call
4043 warn_function_cold on functions that already have cold attribute.
4045 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4048 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4049 COMDAT group function labels in .data.rel.ro.local section.
4050 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4053 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4054 comparison_operator in B and S integer comparisons. Likewise, use
4055 ordered_comparison_operator instead of cmpib_comparison_operator in
4057 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4059 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4061 Update copyright years.
4063 * gcc.c (process_command): Update copyright notice dates.
4064 * gcov-dump.c (print_version): Ditto.
4065 * gcov.c (print_version): Ditto.
4066 * gcov-tool.c (print_version): Ditto.
4067 * gengtype.c (create_file): Ditto.
4068 * doc/cpp.texi: Bump @copying's copyright year.
4069 * doc/cppinternals.texi: Ditto.
4070 * doc/gcc.texi: Ditto.
4071 * doc/gccint.texi: Ditto.
4072 * doc/gcov.texi: Ditto.
4073 * doc/install.texi: Ditto.
4074 * doc/invoke.texi: Ditto.
4076 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4078 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4081 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4083 PR tree-optimization/93098
4084 * match.pd (popcount): For shift amounts, use integer_onep
4085 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4086 tests. Make sure that precision is power of two larger than or equal
4087 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4088 instead of ULL suffixed constants. Formatting fixes.
4090 Copyright (C) 2020 Free Software Foundation, Inc.
4092 Copying and distribution of this file, with or without modification,
4093 are permitted in any medium without royalty provided the copyright
4094 notice and this notice are preserved.