| blob | f5d4e6dd582697565e0622079fb18f68b1b873b6 |
1 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
3 PR target/93932
4 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
5 Split the insn into two parts. This insn only does variable
6 extract from a register.
7 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
8 variable extract from memory.
9 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
10 only does variable extract from a register.
11 (vsx_extract_v4sf_var_load): New insn, do variable extract from
12 memory.
13 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
14 into two parts. This insn only does variable extract from a
15 register.
16 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
17 do variable extract from memory.
19 2020-02-27 Martin Jambor <mjambor@suse.cz>
20 Feng Xue <fxue@os.amperecomputing.com>
22 PR ipa/93707
23 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
24 new function calls_same_node_or_its_all_contexts_clone_p.
25 (cgraph_edge_brings_value_p): Use it.
26 (cgraph_edge_brings_value_p): Likewise.
27 (self_recursive_pass_through_p): Return false if caller is a clone.
28 (self_recursive_agg_pass_through_p): Likewise.
30 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
32 PR middle-end/92152
33 * alias.c (ends_tbaa_access_path_p): Break out from ...
34 (component_uses_parent_alias_set_from): ... here.
35 * alias.h (ends_tbaa_access_path_p): Declare.
36 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
37 handle trailing arrays past end of tbaa access path.
38 (aliasing_component_refs_p): ... here; likewise.
39 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
40 path; disambiguate also past end of it.
41 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
42 path.
44 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
46 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
47 beginning of the file.
48 (vcreate_bf16, vcombine_bf16): New.
49 (vdup_n_bf16, vdupq_n_bf16): New.
50 (vdup_lane_bf16, vdup_laneq_bf16): New.
51 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
52 (vduph_lane_bf16, vduph_laneq_bf16): New.
53 (vset_lane_bf16, vsetq_lane_bf16): New.
54 (vget_lane_bf16, vgetq_lane_bf16): New.
55 (vget_high_bf16, vget_low_bf16): New.
56 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
57 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
58 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
59 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
60 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
61 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
62 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
63 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
64 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
65 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
66 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
67 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
68 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
69 (vreinterpretq_bf16_p128): New.
70 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
71 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
72 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
73 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
74 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
75 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
76 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
77 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
78 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
79 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
80 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
81 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
82 (vreinterpretq_p128_bf16): New.
83 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
84 (V_elem): Likewise.
85 (V_elem_l): Likewise.
86 (VD_LANE): Likewise.
87 (VQX) Add V8BF.
88 (V_DOUBLE): Likewise.
89 (VDQX): Add V4BF and V8BF.
90 (V_two_elem, V_three_elem, V_four_elem): Likewise.
91 (V_reg): Likewise.
92 (V_HALF): Likewise.
93 (V_double_vector_mode): Likewise.
94 (V_cmp_result): Likewise.
95 (V_uf_sclr): Likewise.
96 (V_sz_elem): Likewise.
97 (Is_d_reg): Likewise.
98 (V_mode_nunits): Likewise.
99 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
101 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
103 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
104 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
105 (<expander><mode>3<exec>): Likewise.
106 (<expander><mode>3): New.
107 (v<expander><mode>3): New.
108 (<expander><mode>3): New.
109 (<expander><mode>3<exec>): Rename to ...
110 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
111 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
113 2020-02-27 Alexandre Oliva <oliva@adacore.com>
115 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
116 them alone on vx7.
118 2020-02-27 Richard Biener <rguenther@suse.de>
120 PR tree-optimization/93508
121 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
122 non-_CHK variants. Valueize their length arguments.
124 2020-02-27 Richard Biener <rguenther@suse.de>
126 PR tree-optimization/93953
127 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
128 to the hash-map entry.
130 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
132 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
134 2020-02-27 Mark Williams <mwilliams@fb.com>
136 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
137 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
138 -ffile-prefix-map and -fmacro-prefix-map.
139 * lto-streamer-out.c: Include file-prefix-map.h.
140 (lto_output_location): Remap the file part of locations.
142 2020-02-27 Jakub Jelinek <jakub@redhat.com>
144 PR c/93949
145 * gimplify.c (gimplify_init_constructor): Don't promote readonly
146 DECL_REGISTER variables to TREE_STATIC.
148 PR tree-optimization/93582
149 PR tree-optimization/93945
150 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
151 non-zero INTEGER_CST second argument and ref->offset or ref->size
152 not a multiple of BITS_PER_UNIT.
154 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
156 * doc/install.texi (Binaries): Update description of BullFreeware.
158 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
160 PR c++/90467
162 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
163 C++ Language Options, Warning Options, and Static Analyzer
164 Options lists. Document negative form of options enabled by
165 default. Move some things around to more accurately sort
166 warnings by category.
167 (C++ Dialect Options, Warning Options, Static Analyzer
168 Options): Document negative form of options when enabled by
169 default. Move some things around to more accurately sort
170 warnings by category. Add some missing index entries.
171 Light copy-editing.
173 2020-02-26 Carl Love <cel@us.ibm.com>
175 PR target/91276
176 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
177 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
178 for the vector unsigned short arguments. It is also listed as the
179 name of the built-in for arguments vector unsigned short,
180 vector unsigned int and vector unsigned long long built-ins. The
181 name of the builtins for these arguments should be:
182 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
183 __builtin_crypto_vpmsumd respectively.
185 2020-02-26 Richard Biener <rguenther@suse.de>
187 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
188 and load permutation.
190 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
192 PR middle-end/93843
193 * optabs-tree.c (supportable_convert_operation): Reject types with
194 scalar modes.
196 2020-02-26 David Malcolm <dmalcolm@redhat.com>
198 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
200 2020-02-26 Jakub Jelinek <jakub@redhat.com>
202 PR tree-optimization/93820
203 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
204 argument to ALL_INTEGER_CST_P boolean.
205 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
206 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
207 adjacent INTEGER_CST store into merged_store->only_constants like
208 overlapping one.
210 2020-02-25 Jakub Jelinek <jakub@redhat.com>
212 PR other/93912
213 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
214 -> probability.
215 * cfghooks.c (verify_flow_info): Likewise.
216 * predict.c (combine_predictions_for_bb): Likewise.
217 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
218 sucessor -> successor.
219 (find_traces_1_round): Fix comment typo, destinarion -> destination.
220 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
221 successors.
222 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
223 message typo, sucessors -> successors.
225 2020-02-25 Martin Sebor <msebor@redhat.com>
227 * doc/extend.texi (attribute access): Correct an example.
229 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
231 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
232 Add simd_bf.
233 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
234 (VAR15, VAR16): New.
235 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
236 (VD): Enable for V4BF.
237 (VDC): Likewise.
238 (VQ): Enable for V8BF.
239 (VQ2): Likewise.
240 (VQ_NO2E): Likewise.
241 (VDBL, Vdbl): Add V4BF.
242 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
243 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
244 (bfloat16x8x2_t): Likewise.
245 (bfloat16x4x3_t): Likewise.
246 (bfloat16x8x3_t): Likewise.
247 (bfloat16x4x4_t): Likewise.
248 (bfloat16x8x4_t): Likewise.
249 (vcombine_bf16): New.
250 (vld1_bf16, vld1_bf16_x2): New.
251 (vld1_bf16_x3, vld1_bf16_x4): New.
252 (vld1q_bf16, vld1q_bf16_x2): New.
253 (vld1q_bf16_x3, vld1q_bf16_x4): New.
254 (vld1_lane_bf16): New.
255 (vld1q_lane_bf16): New.
256 (vld1_dup_bf16): New.
257 (vld1q_dup_bf16): New.
258 (vld2_bf16): New.
259 (vld2q_bf16): New.
260 (vld2_dup_bf16): New.
261 (vld2q_dup_bf16): New.
262 (vld3_bf16): New.
263 (vld3q_bf16): New.
264 (vld3_dup_bf16): New.
265 (vld3q_dup_bf16): New.
266 (vld4_bf16): New.
267 (vld4q_bf16): New.
268 (vld4_dup_bf16): New.
269 (vld4q_dup_bf16): New.
270 (vst1_bf16, vst1_bf16_x2): New.
271 (vst1_bf16_x3, vst1_bf16_x4): New.
272 (vst1q_bf16, vst1q_bf16_x2): New.
273 (vst1q_bf16_x3, vst1q_bf16_x4): New.
274 (vst1_lane_bf16): New.
275 (vst1q_lane_bf16): New.
276 (vst2_bf16): New.
277 (vst2q_bf16): New.
278 (vst3_bf16): New.
279 (vst3q_bf16): New.
280 (vst4_bf16): New.
281 (vst4q_bf16): New.
283 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
285 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
286 (VALL_F16): Likewise.
287 (VALLDI_F16): Likewise.
288 (Vtype): Likewise.
289 (Vetype): Likewise.
290 (vswap_width_name): Likewise.
291 (VSWAP_WIDTH): Likewise.
292 (Vel): Likewise.
293 (VEL): Likewise.
294 (q): Likewise.
295 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
296 (vget_lane_bf16, vgetq_lane_bf16): New.
297 (vcreate_bf16): New.
298 (vdup_n_bf16, vdupq_n_bf16): New.
299 (vdup_lane_bf16, vdup_laneq_bf16): New.
300 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
301 (vduph_lane_bf16, vduph_laneq_bf16): New.
302 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
303 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
304 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
305 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
306 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
307 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
308 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
309 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
310 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
311 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
312 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
313 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
314 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
315 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
316 (vreinterpretq_bf16_p128): New.
317 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
318 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
319 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
320 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
321 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
322 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
323 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
324 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
325 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
326 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
327 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
328 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
329 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
330 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
331 (vreinterpretq_p128_bf16): New.
333 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
335 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
336 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
337 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
338 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
339 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
340 * config/arm/iterators.md (VSF2BF): New attribute.
341 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
342 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
343 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
345 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
347 * config/arm/arm.md (required_for_purecode): New attribute.
348 (enabled): Handle required_for_purecode.
349 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
350 work with -mpure-code.
352 2020-02-25 Jakub Jelinek <jakub@redhat.com>
354 PR rtl-optimization/93908
355 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
356 with mask.
358 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
360 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
362 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
364 * doc/install.texi (--enable-checking): Adjust wording.
366 2020-02-25 Richard Biener <rguenther@suse.de>
368 PR tree-optimization/93868
369 * tree-vect-slp.c (slp_copy_subtree): New function.
370 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
371 re-arranging stmts in it.
373 2020-02-25 Jakub Jelinek <jakub@redhat.com>
375 PR middle-end/93874
376 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
377 dummy function and remove it at the end.
379 PR translation/93864
380 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
381 paramter -> parameter.
382 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
383 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
385 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
387 * doc/install.texi (--enable-checking): Properly document current
388 behavior.
389 (--enable-stage1-checking): Minor clarification about bootstrap.
391 2020-02-24 David Malcolm <dmalcolm@redhat.com>
393 PR analyzer/93032
394 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
395 -fanalyzer-checker=taint is also required.
396 (-fanalyzer-checker=): Note that providing this option enables the
397 given checker, and doing so may be required for checkers that are
398 disabled by default.
400 2020-02-24 David Malcolm <dmalcolm@redhat.com>
402 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
403 significant control flow events; add a "3" which shows all
404 control flow events; the old "3" becomes "4".
406 2020-02-24 Jakub Jelinek <jakub@redhat.com>
408 PR tree-optimization/93582
409 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
410 pd.offset and pd.size to be counted in bits rather than bytes, add
411 support for maxsizei that is not a multiple of BITS_PER_UNIT and
412 handle bitfield stores and loads.
413 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
414 uncomparable quantities - bytes vs. bits. Allow push_partial_def
415 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
416 pd.offset/pd.size to be counted in bits rather than bytes.
417 Formatting fix. Rename shadowed len variable to buflen.
419 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
420 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
422 PR driver/47785
423 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
424 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
425 * opts-common.c (parse_options_from_collect_gcc_options): New function.
426 (prepend_xassembler_to_collect_as_options): Likewise.
427 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
428 (prepend_xassembler_to_collect_as_options): Likewise.
429 * lto-opts.c (lto_write_options): Stream assembler options
430 in COLLECT_AS_OPTIONS.
431 * lto-wrapper.c (xassembler_options_error): New static variable.
432 (get_options_from_collect_gcc_options): Move parsing options code to
433 parse_options_from_collect_gcc_options and call it.
434 (merge_and_complain): Validate -Xassembler options.
435 (append_compiler_options): Handle OPT_Xassembler.
436 (run_gcc): Append command line -Xassembler options to
437 collect_gcc_options.
438 * doc/invoke.texi: Add documentation about using Xassembler
439 options with LTO.
441 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
443 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
444 for LTGT.
445 (riscv_rtx_costs): Update cost model for LTGT.
447 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
449 PR rtl-optimization/93564
450 * ira-color.c (struct update_cost_queue_elem): New member start.
451 (queue_update_cost, get_next_update_cost): Add new arg start.
452 (allocnos_conflict_p): New function.
453 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
454 Add checking conflicts with allocnos_conflict_p.
455 (update_costs_from_prefs, restore_costs_from_copies): Adjust
456 update_costs_from_allocno calls.
457 (update_conflict_hard_regno_costs): Add checking conflicts with
458 allocnos_conflict_p. Adjust calls of queue_update_cost and
459 get_next_update_cost.
460 (assign_hard_reg): Adjust calls of queue_update_cost. Add
461 debugging print.
462 (bucket_allocno_compare_func): Restore previous version.
464 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
466 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
467 double-word size when handling aggregate return values.
468 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
469 that homogeneous SFmode and DFmode aggregates are passed and returned
470 in general registers.
472 2020-02-21 Jakub Jelinek <jakub@redhat.com>
474 PR translation/93759
475 * opts.c (print_filtered_help): Translate help before appending
476 messages to it rather than after that.
478 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
480 PR rtl-optimization/PR92989
481 * lra-lives.c (process_bb_lives): Restore the original order
482 of the bb liveness update. Call make_hard_regno_dead for each
483 register clobbered at the start of an EH receiver.
485 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
487 PR ipa/93763
488 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
489 self-recursively generated.
491 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
493 PR target/93860
494 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
495 error string.
497 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
499 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
500 Document new target supports option.
502 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
504 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
505 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
506 * config/arm/iterators.md (MATMUL): New iterator.
507 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
508 (mmla_sfx): New attribute.
509 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
510 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
511 (UNSPEC_MATMUL_US): New.
513 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
515 * config/arm/arm.md: Prevent scalar shifts from being used when big
516 endian is enabled.
518 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
519 Richard Biener <rguenther@suse.de>
521 PR tree-optimization/93586
522 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
523 after mismatched array refs; do not sure type size information to
524 recover from unmatched referneces with !flag_strict_aliasing_p.
526 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
528 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
529 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
530 (scatter_store<mode>): Rename to ...
531 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
532 (scatter<mode>_exec): Delete. Move contents ...
533 (mask_scatter_store<mode>): ... here, and rename that to ...
534 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
535 Remove mode conversion.
536 (mask_gather_load<mode>): Rename to ...
537 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
538 Remove mode conversion.
539 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
541 2020-02-21 Martin Jambor <mjambor@suse.cz>
543 PR tree-optimization/93845
544 * tree-sra.c (verify_sra_access_forest): Only test access size of
545 scalar types.
547 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
549 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
550 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
551 (addv64di3_exec): Likewise.
552 (subv64di3): Likewise.
553 (subv64di3_exec): Likewise.
554 (addv64di3_zext): Likewise.
555 (addv64di3_zext_exec): Likewise.
556 (addv64di3_zext_dup): Likewise.
557 (addv64di3_zext_dup_exec): Likewise.
558 (addv64di3_zext_dup2): Likewise.
559 (addv64di3_zext_dup2_exec): Likewise.
560 (addv64di3_sext_dup2): Likewise.
561 (addv64di3_sext_dup2_exec): Likewise.
562 (<expander>v64di3): Likewise.
563 (<expander>v64di3_exec): Likewise.
564 (*<reduc_op>_dpp_shr_v64di): Likewise.
565 (*plus_carry_dpp_shr_v64di): Likewise.
566 * config/gcn/gcn.md (adddi3): Likewise.
567 (addptrdi3): Likewise.
568 (<expander>di3): Likewise.
570 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
572 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
574 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
576 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
577 support. Use aarch64_emit_mult instead of emitting multiplication
578 instructions directly.
579 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
580 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
582 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
584 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
585 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
586 instead of emitting multiplication instructions directly.
587 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
588 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
589 (@aarch64_frecps<mode>): New expanders.
591 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
593 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
594 on and produce uint64_ts rather than ints.
595 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
596 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
598 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
600 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
601 an unused xmsk register when handling approximate rsqrt.
603 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
605 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
606 flag_finite_math_only condition.
608 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
610 PR target/93828
611 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
612 to destination operand for shufps alternative.
613 (*vec_extractv2si_1): Ditto.
615 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
617 PR target/93658
618 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
619 vector modes.
621 2020-02-20 Martin Liska <mliska@suse.cz>
623 PR translation/93831
624 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
626 2020-02-20 Martin Liska <mliska@suse.cz>
628 PR translation/93830
629 * common/config/avr/avr-common.c: Remote trailing "|".
631 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
633 * collect2.c (maybe_run_lto_and_relink): Fix typo in
634 comment.
636 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
638 PR tree-optimization/93767
639 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
640 access-size bias from the offset calculations for negative strides.
642 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
644 * collect2.c (c_file, o_file): Make const again.
645 (ldout,lderrout, dump_ld_file): Remove.
646 (tool_cleanup): Avoid calling not signal-safe functions.
647 (maybe_run_lto_and_relink): Avoid possible signal handler
648 access to unintialzed memory (lto_o_files).
649 (main): Avoid leaking temp files in $TMPDIR.
650 Initialize c_file/o_file with concat, which avoids exposing
651 uninitialized memory to signal handler, which calls unlink(!).
652 Avoid calling maybe_unlink when the main function returns,
653 since the atexit handler is already doing this.
654 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
656 2020-02-19 Martin Jambor <mjambor@suse.cz>
658 PR tree-optimization/93776
659 * tree-sra.c (create_access): Do not create zero size accesses.
660 (get_access_for_expr): Do not search for zero sized accesses.
662 2020-02-19 Martin Jambor <mjambor@suse.cz>
664 PR tree-optimization/93667
665 * tree-sra.c (scalarizable_type_p): Return false if record fields
666 do not follow wach other.
668 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
670 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
671 rather than fmv.x.s/fmv.s.x.
673 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
675 * config/aarch64/aarch64-simd-builtins.def
676 (intrinsic_vec_smult_lo_): New.
677 (intrinsic_vec_umult_lo_): Likewise.
678 (vec_widen_smult_hi_): Likewise.
679 (vec_widen_umult_hi_): Likewise.
680 * config/aarch64/aarch64-simd.md
681 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
682 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
683 (vmull_high_s16): Likewise.
684 (vmull_high_s32): Likewise.
685 (vmull_high_u8): Likewise.
686 (vmull_high_u16): Likewise.
687 (vmull_high_u32): Likewise.
688 (vmull_s8): Likewise.
689 (vmull_s16): Likewise.
690 (vmull_s32): Likewise.
691 (vmull_u8): Likewise.
692 (vmull_u16): Likewise.
693 (vmull_u32): Likewise.
695 2020-02-18 Martin Liska <mliska@suse.cz>
697 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
698 bootstrap by missing removal of invalid sanity check.
700 2020-02-18 Martin Liska <mliska@suse.cz>
702 PR ipa/92518
703 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
704 Always compare LHS of gimple_assign.
706 2020-02-18 Martin Liska <mliska@suse.cz>
708 PR ipa/93583
709 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
710 and return type of functions.
711 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
712 Drop MALLOC attribute for void functions.
713 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
714 malloc_state for a new VOID clone.
716 2020-02-18 Martin Liska <mliska@suse.cz>
718 PR ipa/92924
719 * common.opt: Add -fprofile-reproducibility.
720 * doc/invoke.texi: Document it.
721 * value-prof.c (dump_histogram_value):
722 Document and support behavior for counters[0]
723 being a negative value.
724 (get_nth_most_common_value): Handle negative
725 counters[0] in respect to flag_profile_reproducible.
727 2020-02-18 Jakub Jelinek <jakub@redhat.com>
729 PR ipa/93797
730 * cgraph.c (verify_speculative_call): Use speculative_id instead of
731 speculative_uid in messages. Remove trailing whitespace from error
732 message. Use num_speculative_call_targets instead of
733 num_speculative_targets in a message.
734 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
735 edge messages and stmt instead of cal_stmt in reference message.
737 PR tree-optimization/93780
738 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
739 before calling build_vector_type.
740 (execute_update_addresses_taken): Likewise.
742 PR driver/93796
743 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
744 typo, functoin -> function.
745 * tree.c (free_lang_data_in_decl): Fix comment typo,
746 functoin -> function.
747 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
749 2020-02-17 David Malcolm <dmalcolm@redhat.com>
751 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
752 won't be printed.
753 (print_option_information): Don't call get_option_url if URLs
754 won't be printed.
756 2020-02-17 Alexandre Oliva <oliva@adacore.com>
758 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
759 handling of register_common-less targets.
761 2020-02-17 Martin Liska <mliska@suse.cz>
763 PR ipa/93760
764 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
766 2020-02-17 Martin Liska <mliska@suse.cz>
768 PR translation/93755
769 * config/rs6000/rs6000.c (rs6000_option_override_internal):
770 Fix double quotes.
772 2020-02-17 Martin Liska <mliska@suse.cz>
774 PR other/93756
775 * config/rx/elf.opt: Fix typo.
777 2020-02-17 Richard Biener <rguenther@suse.de>
779 PR c/86134
780 * opts-global.c (print_ignored_options): Use inform and
781 amend message.
783 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
785 PR target/93047
786 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
788 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
790 PR target/93743
791 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
792 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
794 2020-02-15 Jason Merrill <jason@redhat.com>
796 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
798 2020-02-15 Jakub Jelinek <jakub@redhat.com>
800 PR tree-optimization/93744
801 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
802 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
803 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
804 sure @2 in the first and @1 in the other patterns has no side-effects.
806 2020-02-15 David Malcolm <dmalcolm@redhat.com>
807 Bernd Edlinger <bernd.edlinger@hotmail.de>
809 PR 87488
810 PR other/93168
811 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
812 * configure.ac (--with-diagnostics-urls): New configuration
813 option, based on --with-diagnostics-color.
814 (DIAGNOSTICS_URLS_DEFAULT): New define.
815 * config.h: Regenerate.
816 * configure: Regenerate.
817 * diagnostic.c (diagnostic_urls_init): Handle -1 for
818 DIAGNOSTICS_URLS_DEFAULT from configure-time
819 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
820 and TERM_URLS environment variable.
821 * diagnostic-url.h (diagnostic_url_format): New enum type.
822 (diagnostic_urls_enabled_p): rename to...
823 (determine_url_format): ... this, and change return type.
824 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
825 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
826 the linux console, and mingw.
827 (diagnostic_urls_enabled_p): rename to...
828 (determine_url_format): ... this, and adjust.
829 * pretty-print.h (pretty_printer::show_urls): rename to...
830 (pretty_printer::url_format): ... this, and change to enum.
831 * pretty-print.c (pretty_printer::pretty_printer,
832 pp_begin_url, pp_end_url, test_urls): Adjust.
833 * doc/install.texi (--with-diagnostics-urls): Document the new
834 configuration option.
835 (--with-diagnostics-color): Document the existing interaction
836 with GCC_COLORS better.
837 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
838 vindex reference. Update description of defaults based on the above.
839 (-fdiagnostics-color): Update description of how -fdiagnostics-color
840 interacts with GCC_COLORS.
842 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
844 PR target/93704
845 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
846 conjunction with TARGET_GNU_TLS in early return.
848 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
850 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
851 the mode is not wider than UNITS_PER_WORD.
853 2020-02-14 Martin Jambor <mjambor@suse.cz>
855 PR tree-optimization/93516
856 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
857 access of the same type as the parent.
858 (propagate_subaccesses_from_lhs): Likewise.
860 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
862 PR target/93724
863 * config/i386/avx512vbmi2intrin.h
864 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
865 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
866 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
867 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
868 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
869 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
870 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
871 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
872 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
873 of lacking a closing parenthesis.
874 * config/i386/avx512vbmi2vlintrin.h
875 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
876 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
877 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
878 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
879 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
880 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
881 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
882 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
883 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
884 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
885 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
886 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
887 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
888 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
889 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
890 _mm_shldi_epi32, _mm_mask_shldi_epi32,
891 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
892 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
894 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
896 PR target/93656
897 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
898 the target function entry.
900 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
902 * common/config/arc/arc-common.c (arc_option_optimization_table):
903 Disable if-conversion step when optimized for size.
905 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
907 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
908 R12-R15 are always in ARCOMPACT16_REGS register class.
909 * config/arc/arc.opt (mq-class): Deprecate.
910 * config/arc/constraint.md ("q"): Remove dependency on mq-class
911 option.
912 * doc/invoke.texi (mq-class): Update text.
913 * common/config/arc/arc-common.c (arc_option_optimization_table):
914 Update list.
916 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
918 * config/arc/arc.c (arc_insn_cost): New function.
919 (TARGET_INSN_COST): Define.
920 * config/arc/arc.md (cost): New attribute.
921 (add_n): Use arc_nonmemory_operand.
922 (ashlsi3_insn): Likewise, also update constraints.
923 (ashrsi3_insn): Likewise.
924 (rotrsi3): Likewise.
925 (add_shift): Likewise.
926 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
928 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
930 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
931 registers.
932 (umulsidi_600): Likewise.
934 2020-02-13 Jakub Jelinek <jakub@redhat.com>
936 PR target/93696
937 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
938 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
939 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
940 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
941 pass __A to the builtin followed by __W instead of __A followed by
942 __B.
943 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
944 _mm512_mask_popcnt_epi64): Likewise.
945 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
946 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
947 _mm256_mask_popcnt_epi64): Likewise.
949 PR tree-optimization/93582
950 * fold-const.h (shift_bytes_in_array_left,
951 shift_bytes_in_array_right): Declare.
952 * fold-const.c (shift_bytes_in_array_left,
953 shift_bytes_in_array_right): New function, moved from
954 gimple-ssa-store-merging.c, no longer static.
955 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
956 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
957 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
958 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
959 shift_bytes_in_array.
960 (verify_shift_bytes_in_array): Rename to ...
961 (verify_shift_bytes_in_array_left): ... this. Use
962 shift_bytes_in_array_left instead of shift_bytes_in_array.
963 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
964 instead of verify_shift_bytes_in_array.
965 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
966 / native_interpret_expr where the store covers all needed bits,
967 punt on PDP-endian, otherwise allow all involved offsets and sizes
968 not to be byte-aligned.
970 PR target/93673
971 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
972 use const_0_to_255_operand predicate instead of immediate_operand.
973 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
974 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
975 vgf2p8affineinvqb_<mode><mask_name>,
976 vgf2p8affineqb_<mode><mask_name>): Drop mode from
977 const_0_to_255_operand predicated operands.
979 2020-02-12 Jeff Law <law@redhat.com>
981 * config/h8300/h8300.md (comparison shortening peepholes): Use
982 a mode iterator to merge the HImode and SImode peepholes.
984 2020-02-12 Jakub Jelinek <jakub@redhat.com>
986 PR middle-end/93663
987 * real.c (is_even): Make static. Function comment fix.
988 (is_halfway_below): Make static, don't assert R is not inf/nan,
989 instead return false for those. Small formatting fixes.
991 2020-02-12 Martin Sebor <msebor@redhat.com>
993 PR middle-end/93646
994 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
995 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
996 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
997 (strlen_check_and_optimize_call): Adjust callee name.
999 2020-02-12 Jeff Law <law@redhat.com>
1001 * config/h8300/h8300.md (comparison shortening peepholes): Drop
1002 (and (xor)) variant. Combine other two into single peephole.
1004 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1006 PR rtl-optimization/93565
1007 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1009 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1011 * config/aarch64/aarch64-simd.md
1012 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1013 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1014 generating separate ADDV and zero_extend patterns.
1015 * config/aarch64/iterators.md (VDQV_E): New iterator.
1017 2020-02-12 Jeff Law <law@redhat.com>
1019 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1020 expanders, splits, etc.
1021 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1022 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1023 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1024 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1025 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1026 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1027 function prototype.
1028 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1030 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1032 PR target/93670
1033 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1034 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
1035 TARGET_AVX512DQ from condition.
1036 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1037 instead of <mask_mode512bit_condition> in condition. If
1038 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1039 vextract*32x8.
1040 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1041 from condition.
1043 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
1045 PR target/91052
1046 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1048 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
1050 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1051 where strlen is more legible.
1052 (rs6000_builtin_vectorized_libmass): Ditto.
1053 (rs6000_print_options_internal): Ditto.
1055 2020-02-11 Martin Sebor <msebor@redhat.com>
1057 PR tree-optimization/93683
1058 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1060 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
1062 * config/rs6000/predicates.md (cint34_operand): Rename the
1063 -mprefixed-addr option to be -mprefixed.
1064 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1065 the -mprefixed-addr option to be -mprefixed.
1066 (OTHER_FUTURE_MASKS): Likewise.
1067 (POWERPC_MASKS): Likewise.
1068 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1069 the -mprefixed-addr option to be -mprefixed. Change error
1070 messages to refer to -mprefixed.
1071 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1072 -mprefixed.
1073 (rs6000_legitimate_offset_address_p): Likewise.
1074 (rs6000_mode_dependent_address): Likewise.
1075 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1076 "-mprefixed" for target attributes and pragmas.
1077 (address_to_insn_form): Rename the -mprefixed-addr option to be
1078 -mprefixed.
1079 (rs6000_adjust_insn_length): Likewise.
1080 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1081 -mprefixed-addr option to be -mprefixed.
1082 (ASM_OUTPUT_OPCODE): Likewise.
1083 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1084 -mprefixed-addr option to be -mprefixed.
1085 * config/rs6000/rs6000.opt (-mprefixed): Rename the
1086 -mprefixed-addr option to be prefixed. Change the option from
1087 being undocumented to being documented.
1088 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1089 -mprefixed option. Update the -mpcrel documentation to mention
1090 -mprefixed.
1092 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
1094 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1095 including FIRST_PSEUDO_REGISTER - 1.
1096 * ira-color.c (print_hard_reg_set): Ditto.
1098 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1100 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
1101 (USTERNOP_QUALIFIERS): New define.
1102 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1103 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1104 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1105 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1106 * config/arm/arm_neon.h (vusdot_s32): New.
1107 (vusdot_lane_s32): New.
1108 (vusdotq_lane_s32): New.
1109 (vsudot_lane_s32): New.
1110 (vsudotq_lane_s32): New.
1111 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1112 * config/arm/iterators.md (DOTPROD_I8MM): New.
1113 (sup, opsuffix): Add <us/su>.
1114 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1115 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1117 2020-02-11 Richard Biener <rguenther@suse.de>
1119 PR tree-optimization/93661
1120 PR tree-optimization/93662
1121 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1122 tree_to_poly_int64.
1123 * tree-sra.c (get_access_for_expr): Likewise.
1125 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1127 PR target/93637
1128 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1129 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1130 Change condition from TARGET_AVX2 to TARGET_AVX.
1132 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1134 PR other/93641
1135 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1136 argument of strncmp.
1138 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1140 Try to generate zero-based comparisons.
1141 * config/cris/cris.c (cris_reduce_compare): New function.
1142 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1143 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1144 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1146 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1148 PR target/91913
1149 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1150 in Thumb state and also as a destination in Arm state. Add T16
1151 variants.
1153 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1155 * md.texi (Define Subst): Match closing paren in example.
1157 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1159 PR target/58218
1160 PR other/93641
1161 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1162 arguments of strncmp.
1164 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1166 PR ipa/93203
1167 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1168 but different source value.
1169 (adjust_callers_for_value_intersection): New function.
1170 (gather_edges_for_value): Adjust order of callers to let a
1171 non-self-recursive caller be the first element.
1172 (self_recursive_pass_through_p): Add a new parameter "simple", and
1173 check generalized self-recursive pass-through jump function.
1174 (self_recursive_agg_pass_through_p): Likewise.
1175 (find_more_scalar_values_for_callers_subset): Compute value from
1176 pass-through jump function for self-recursive.
1177 (intersect_with_plats): Cleanup previous implementation code for value
1178 itersection with self-recursive call edge.
1179 (intersect_with_agg_replacements): Likewise.
1180 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1181 function for self-recursive call edge. Cleanup previous implementation
1182 code for value intersection with self-recursive call edge.
1183 (decide_whether_version_node): Remove dead callers and adjust order
1184 to let a non-self-recursive caller be the first element.
1186 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1188 * recog.c: Move pass_split_before_sched2 code in front of
1189 pass_split_before_regstack.
1190 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1191 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1192 (rest_of_handle_split_before_sched2): Remove.
1193 (pass_split_before_sched2::execute): Unconditionally call
1194 split_all_insns.
1195 (enable_split_before_sched2): New function.
1196 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1197 (pass_split_before_regstack::gate): Ditto.
1198 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1199 Update name check for renamed split4 pass.
1200 * config/sh/sh.c (register_sh_passes): Update pass insertion
1201 point for renamed split4 pass.
1203 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1205 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1206 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1207 copying them around between host and target.
1209 2020-02-08 Andrew Pinski <apinski@marvell.com>
1211 PR target/91927
1212 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1213 STRICT_ALIGNMENT also.
1215 2020-02-08 Jim Wilson <jimw@sifive.com>
1217 PR target/93532
1218 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1220 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1221 Jakub Jelinek <jakub@redhat.com>
1223 PR target/65782
1224 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1225 xmm16-xmm31 call-used even in 64-bit ms-abi.
1227 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1229 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1230 (simd_ummla, simd_usmmla): Likewise.
1231 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1232 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1233 (vusmmlaq_s32): New.
1235 2020-02-07 Richard Biener <rguenther@suse.de>
1237 PR middle-end/93519
1238 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1239 skipping unreachable regions.
1240 (optimize_inline_calls): Skip folding stmts when we didn't
1241 inline.
1243 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1245 PR target/85667
1246 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1247 Don't return aggregates with only SFmode and DFmode in SSE
1248 register.
1249 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1251 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1253 PR target/93122
1254 * config/rs6000/rs6000-logue.c
1255 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1256 if it fails, move rs into end_addr and retry. Add
1257 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1258 the insn pattern doesn't describe well what exactly happens to
1259 dwarf2cfi.c.
1261 PR target/93594
1262 * config/i386/predicates.md (avx_identity_operand): Remove.
1263 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1264 (avx_<castmode><avxsizesuffix>_<castmode>,
1265 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1266 a VEC_CONCAT of the operand and UNSPEC_CAST.
1267 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1268 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1269 UNSPEC_CAST.
1271 PR target/93611
1272 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1273 recog_data.insn if distance_non_agu_define changed it.
1275 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1277 PR target/93569
1278 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1279 we only had X-FORM (reg+reg) addressing for vectors. Also before
1280 ISA 3.0, we only had X-FORM addressing for scalars in the
1281 traditional Altivec registers.
1283 2020-02-06 <zhongyunde@huawei.com>
1284 Vladimir Makarov <vmakarov@redhat.com>
1286 PR rtl-optimization/93561
1287 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1288 hard register range.
1290 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1292 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1293 attribute.
1295 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1297 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1298 where the low and the high 32 bits are equal to each other specially,
1299 with an rldimi instruction.
1301 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1303 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1305 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1307 * config/arm/arm-tables.opt: Regenerate.
1309 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1311 PR target/87763
1312 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1313 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1314 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1316 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1318 PR rtl-optimization/87763
1319 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1321 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1323 * config/aarch64/aarch64-simd-builtins.def
1324 (bfmlaq): New built-in function.
1325 (bfmlalb): New built-in function.
1326 (bfmlalt): New built-in function.
1327 (bfmlalb_lane): New built-in function.
1328 (bfmlalt_lane): New built-in function.
1329 * config/aarch64/aarch64-simd.md
1330 (aarch64_bfmmlaqv4sf): New pattern.
1331 (aarch64_bfmlal<bt>v4sf): New pattern.
1332 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1333 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1334 (vbfmlalbq_f32): New intrinsic.
1335 (vbfmlaltq_f32): New intrinsic.
1336 (vbfmlalbq_lane_f32): New intrinsic.
1337 (vbfmlaltq_lane_f32): New intrinsic.
1338 (vbfmlalbq_laneq_f32): New intrinsic.
1339 (vbfmlaltq_laneq_f32): New intrinsic.
1340 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1341 (bt): New int attribute.
1343 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1345 * config/i386/i386.md (*pushtf): Emit "#" instead of
1346 calling gcc_unreachable in insn output.
1347 (*pushxf): Ditto.
1348 (*pushdf): Ditto.
1349 (*pushsf_rex64): Ditto for alternatives other than 1.
1350 (*pushsf): Ditto for alternatives other than 1.
1352 2020-02-06 Martin Liska <mliska@suse.cz>
1354 PR gcov-profile/91971
1355 PR gcov-profile/93466
1356 * coverage.c (coverage_init): Revert mangling of
1357 path into filename. It can lead to huge filename length.
1358 Creation of subfolders seem more natural.
1360 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1362 PR target/93300
1363 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1364 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1365 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1367 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1369 PR target/93594
1370 * config/i386/predicates.md (avx_identity_operand): New predicate.
1371 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1372 define_insn_and_split.
1374 PR libgomp/93515
1375 * omp-low.c (use_pointer_for_field): For nested constructs, also
1376 look for map clauses on target construct.
1377 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1378 taskreg_nesting_level.
1380 PR libgomp/93515
1381 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1382 shared clause, call omp_notice_variable on outer context if any.
1384 2020-02-05 Jason Merrill <jason@redhat.com>
1386 PR c++/92003
1387 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1388 non-zero address even if weak and not yet defined.
1390 2020-02-05 Martin Sebor <msebor@redhat.com>
1392 PR tree-optimization/92765
1393 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1394 * tree-ssa-strlen.c (compute_string_length): Remove.
1395 (determine_min_objsize): Remove.
1396 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1397 Avoid using type size as the upper bound on string length.
1398 (handle_builtin_string_cmp): Add an argument. Adjust.
1399 (strlen_check_and_optimize_call): Pass additional argument to
1400 handle_builtin_string_cmp.
1402 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1404 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1405 (*pushdi2_rex64 peephole2): Unconditionally split after
1406 epilogue_completed.
1407 (*ashl<mode>3_doubleword): Ditto.
1408 (*<shift_insn><mode>3_doubleword): Ditto.
1410 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1412 PR target/93568
1413 * config/rs6000/rs6000.c (get_vector_offset): Fix
1415 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1417 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1419 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1421 * doc/analyzer.texi
1422 (Special Functions for Debugging the Analyzer): Update description
1423 of __analyzer_dump_exploded_nodes.
1425 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1427 PR target/92190
1428 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1429 include sets and not clobbers in the vzeroupper pattern.
1430 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1431 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1432 (*avx_vzeroupper_1): New define_insn_and_split.
1434 PR target/92190
1435 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1436 don't run when !optimize.
1437 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1438 when !optimize.
1440 2020-02-05 Richard Biener <rguenther@suse.de>
1442 PR middle-end/90648
1443 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1444 checks before matching calls.
1446 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1448 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1449 function comment typo.
1451 PR middle-end/93555
1452 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1453 simd_clone_create failed when i == 0, adjust clone->nargs by
1454 clone->inbranch.
1456 2020-02-05 Martin Liska <mliska@suse.cz>
1458 PR c++/92717
1459 * doc/invoke.texi: Document that one should
1460 not combine ASLR and -fpch.
1462 2020-02-04 Richard Biener <rguenther@suse.de>
1464 PR tree-optimization/93538
1465 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1467 2020-02-04 Richard Biener <rguenther@suse.de>
1469 PR tree-optimization/91123
1470 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1471 (vn_walk_cb_data::last_vuse): New member.
1472 (vn_walk_cb_data::saved_operands): Likewsie.
1473 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1474 (vn_walk_cb_data::push_partial_def): Use finish.
1475 (vn_reference_lookup_2): Update last_vuse and use finish if
1476 we've saved operands.
1477 (vn_reference_lookup_3): Use finish and update calls to
1478 push_partial_defs everywhere. When translating through
1479 memcpy or aggregate copies save off operands and alias-set.
1480 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1481 operation for redundant store removal.
1483 2020-02-04 Richard Biener <rguenther@suse.de>
1485 PR tree-optimization/92819
1486 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1487 generating more stmts than before.
1489 2020-02-04 Martin Liska <mliska@suse.cz>
1491 * config/arm/arm.c (arm_gen_far_branch): Move the function
1492 outside of selftests.
1494 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1496 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1497 function to adjust PC-relative vector addresses.
1498 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1499 handle vectors with PC-relative addresses.
1501 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1503 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1504 reference.
1505 (hard_reg_and_mode_to_addr_mask): Delete.
1506 (rs6000_adjust_vec_address): If the original vector address
1507 was REG+REG or REG+OFFSET and the element is not zero, do the add
1508 of the elements in the original address before adding the offset
1509 for the vector element. Use address_to_insn_form to validate the
1510 address using the register being loaded, rather than guessing
1511 whether the address is a DS-FORM or DQ-FORM address.
1513 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1515 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1516 to calculate the offset in memory from the start of a vector of a
1517 particular element. Add code to keep the element number in
1518 bounds if the element number is variable.
1519 (rs6000_adjust_vec_address): Move calculation of offset of the
1520 vector element to get_vector_offset.
1521 (rs6000_split_vec_extract_var): Do not do the initial AND of
1522 element here, move the code to get_vector_offset.
1524 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1526 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1527 gcc_asserts.
1529 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1531 * config/rs6000/constraints.md: Improve documentation.
1533 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1535 PR target/93548
1536 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1537 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1539 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1541 * config.gcc: Remove "carrizo" support.
1542 * config/gcn/gcn-opts.h (processor_type): Likewise.
1543 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1544 * config/gcn/gcn.opt (gpu_type): Likewise.
1545 * config/gcn/t-omp-device: Likewise.
1547 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1549 PR target/91816
1550 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1551 * config/arm/arm.c (arm_gen_far_branch): New function
1552 arm_gen_far_branch.
1553 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1555 2020-02-03 Julian Brown <julian@codesourcery.com>
1556 Tobias Burnus <tobias@codesourcery.com>
1558 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1560 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1562 PR target/93533
1563 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1564 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1565 result.
1567 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1569 PR rtl-optimization/91333
1570 * ira-color.c (struct allocno_color_data): Add member
1571 hard_reg_prefs.
1572 (init_allocno_threads): Set the member up.
1573 (bucket_allocno_compare_func): Add compare hard reg
1574 prefs.
1576 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1578 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1580 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1581 * config.in: Regenerated.
1582 * configure: Regenerated.
1583 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1584 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1585 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1587 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1589 * configure: Regenerate.
1591 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1593 PR rtl-optimization/91333
1594 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1595 reg preferences comparison up.
1597 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1599 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1600 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1601 aarch64-sve-builtins-base.h.
1602 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1603 aarch64-sve-builtins-base.cc.
1604 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1605 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1606 (svcvtnt): Declare.
1607 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1608 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1609 (svcvtnt): New functions.
1610 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1611 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1612 (svcvtnt): New functions.
1613 (svcvt): Add a form that converts f32 to bf16.
1614 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1615 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1616 Declare.
1617 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1618 Treat B as bfloat16_t.
1619 (ternary_bfloat_lane_base): New class.
1620 (ternary_bfloat_def): Likewise.
1621 (ternary_bfloat): New shape.
1622 (ternary_bfloat_lane_def): New class.
1623 (ternary_bfloat_lane): New shape.
1624 (ternary_bfloat_lanex2_def): New class.
1625 (ternary_bfloat_lanex2): New shape.
1626 (ternary_bfloat_opt_n_def): New class.
1627 (ternary_bfloat_opt_n): New shape.
1628 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1629 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1630 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1631 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1632 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1633 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1634 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1635 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1636 the pattern off the narrow mode instead of the wider one.
1637 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1638 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1639 (sve_fp_op): Handle them.
1640 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1641 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1643 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1645 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1646 * config/aarch64/aarch64-modes.def (BF): Move definition before
1647 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1648 (SVE_MODES): Handle BF modes.
1649 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1650 BF modes.
1651 (aarch64_full_sve_mode): Likewise.
1652 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1653 and VNx32BF.
1654 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1655 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1656 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1657 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1658 new SVE BF modes.
1659 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1660 type_class_index.
1661 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1662 (TYPES_all_data): Add bf16.
1663 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1664 (register_tuple_type): Increase buffer size.
1665 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1666 (bf16): New type suffix.
1667 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1668 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1669 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1670 Change type from all_data to all_arith.
1671 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1672 (svminp): Likewise.
1674 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1675 Matthew Malcomson <matthew.malcomson@arm.com>
1676 Richard Sandiford <richard.sandiford@arm.com>
1678 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1679 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1680 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1681 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1682 __ARM_FEATURE_MATMUL_FP64.
1683 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1684 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1685 be disabled at the same time.
1686 (f32mm): New extension.
1687 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1688 (AARCH64_FL_F64MM): Bump to the next bit up.
1689 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1690 (TARGET_SVE_F64MM): New macros.
1691 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1692 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1693 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1694 (UNSPEC_ZIP2Q): New unspeccs.
1695 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1696 (optab, sur, perm_insn): Handle the new unspecs.
1697 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1698 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1699 TARGET_SVE_F64MM instead of separate tests.
1700 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1701 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1702 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1703 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1704 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1705 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1706 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1707 (TYPES_s_signed): New macro.
1708 (TYPES_s_integer): Use it.
1709 (TYPES_d_float): New macro.
1710 (TYPES_d_data): Use it.
1711 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1712 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1713 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1714 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1715 (svmmla): New shape.
1716 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1717 template parameters.
1718 (ternary_resize2_lane_base): Likewise.
1719 (ternary_resize2_base): New class.
1720 (ternary_qq_lane_base): Likewise.
1721 (ternary_intq_uintq_lane_def): Likewise.
1722 (ternary_intq_uintq_lane): New shape.
1723 (ternary_intq_uintq_opt_n_def): New class
1724 (ternary_intq_uintq_opt_n): New shape.
1725 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1726 (ternary_uintq_intq_def): New class.
1727 (ternary_uintq_intq): New shape.
1728 (ternary_uintq_intq_lane_def): New class.
1729 (ternary_uintq_intq_lane): New shape.
1730 (ternary_uintq_intq_opt_n_def): New class.
1731 (ternary_uintq_intq_opt_n): New shape.
1732 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1733 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1734 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1735 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1736 Generalize to...
1737 (svdotprod_lane_impl): ...this new class.
1738 (svmmla_impl, svusdot_impl): New classes.
1739 (svdot_lane): Update to use svdotprod_lane_impl.
1740 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1741 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1742 functions.
1743 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1744 function, with no types defined.
1745 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1746 AARCH64_FL_I8MM functions.
1747 (svmmla): New AARCH64_FL_F32MM function.
1748 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1749 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1750 AARCH64_FL_F64MM function.
1751 (REQUIRED_EXTENSIONS):
1753 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1755 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1756 alternative only.
1758 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1760 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1761 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1762 (*movti_internal): Do not check for
1763 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1764 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1765 just after check for TARGET_AVX.
1766 (*movdf_internal): Ditto.
1767 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1768 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1769 * config/i386/sse.md (mov<mode>_internal): Only check
1770 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1771 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1772 (<sse>_andnot<mode>3<mask_name>): Move check for
1773 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1774 (<code><mode>3<mask_name>): Ditto.
1775 (*andnot<mode>3): Ditto.
1776 (*andnottf3): Ditto.
1777 (*<code><mode>3): Ditto.
1778 (*<code>tf3): Ditto.
1779 (*andnot<VI:mode>3): Remove
1780 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1781 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1782 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1783 (sse4_1_blendv<ssemodesuffix>): Ditto.
1784 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1785 Explain that tune applies to 128bit instructions only.
1787 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1789 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1790 to definition of hsa_kernel_description. Parse assembly to find SGPR
1791 and VGPR count of kernel and store in hsa_kernel_description.
1793 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1795 PR rtl-optimization/91838
1796 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1797 to truncate if allowed or reject combination.
1799 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1801 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1802 (find_inv_vars_cb): Likewise.
1804 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1806 * calls.c (special_function_p): Split out the check for DECL_NAME
1807 being non-NULL and fndecl being extern at file scope into a
1808 new maybe_special_function_p and call it. Drop check for fndecl
1809 being non-NULL that was after a usage of DECL_NAME (fndecl).
1810 * tree.h (maybe_special_function_p): New inline function.
1812 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1814 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1815 (mask_gather_load<mode>): ... here, and zero-initialize the
1816 destination.
1817 (maskload<mode>di): Zero-initialize the destination.
1818 * config/gcn/gcn.c:
1820 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1822 PR analyzer/93356
1823 * doc/analyzer.texi (Limitations): Note that constraints on
1824 floating-point values are currently ignored.
1826 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1828 PR lto/93384
1829 * symtab.c (symtab_node::noninterposable_alias): If localalias
1830 already exists, but is not usable, append numbers after it until
1831 a unique name is found. Formatting fix.
1833 PR middle-end/93505
1834 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1835 rotate counts.
1837 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1839 * config/gcn/gcn.c (print_operand): Handle LTGT.
1840 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1842 2020-01-30 Richard Biener <rguenther@suse.de>
1844 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1845 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1847 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1849 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1850 without a DECL in .data.rel.ro.local.
1852 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1854 PR target/93494
1855 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1856 returned.
1858 PR target/91824
1859 * config/i386/sse.md
1860 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1861 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1862 any_extend code iterator instead of always zero_extend.
1863 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1864 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1865 Use any_extend code iterator instead of always zero_extend.
1866 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1867 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1868 Use any_extend code iterator instead of always zero_extend.
1869 (*sse2_pmovmskb_ext): New define_insn.
1870 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1872 PR target/91824
1873 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1874 (*popcountsi2_zext_falsedep): New define_insn.
1876 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1878 * config.in: Regenerated.
1879 * configure: Regenerated.
1881 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1883 PR bootstrap/93409
1884 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1885 LLVM's assembler changed the default in version 9.
1887 2020-01-24 Jeff Law <law@redhat.com>
1889 PR tree-optimization/89689
1890 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1892 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1894 Revert:
1896 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1898 PR rtl-optimization/87763
1899 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1900 simplification to handle subregs as well as bare regs.
1901 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1903 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1905 PR target/93221
1906 * ira.c (ira): Revert use of simplified LRA algorithm.
1908 2020-01-29 Martin Jambor <mjambor@suse.cz>
1910 PR tree-optimization/92706
1911 * tree-sra.c (struct access): Fields first_link, last_link,
1912 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1913 next_rhs_queued and grp_rhs_queued respectively, new fields
1914 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1915 (struct assign_link): Field next renamed to next_rhs, new field
1916 next_lhs. Updated comment.
1917 (work_queue_head): Renamed to rhs_work_queue_head.
1918 (lhs_work_queue_head): New variable.
1919 (add_link_to_lhs): New function.
1920 (relink_to_new_repr): Also relink LHS lists.
1921 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1922 (add_access_to_lhs_work_queue): New function.
1923 (pop_access_from_work_queue): Renamed to
1924 pop_access_from_rhs_work_queue.
1925 (pop_access_from_lhs_work_queue): New function.
1926 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1927 work_queue.
1928 (child_would_conflict_in_lacc): Renamed to
1929 child_would_conflict_in_acc. Adjusted parameter names.
1930 (create_artificial_child_access): New parameter set_grp_read, use it.
1931 (subtree_mark_written_and_enqueue): Renamed to
1932 subtree_mark_written_and_rhs_enqueue.
1933 (propagate_subaccesses_across_link): Renamed to
1934 propagate_subaccesses_from_rhs.
1935 (propagate_subaccesses_from_lhs): New function.
1936 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1937 RHSs.
1939 2020-01-29 Martin Jambor <mjambor@suse.cz>
1941 PR tree-optimization/92706
1942 * tree-sra.c (struct access): Adjust comment of
1943 grp_total_scalarization.
1944 (find_access_in_subtree): Look for single children spanning an entire
1945 access.
1946 (scalarizable_type_p): Allow register accesses, adjust callers.
1947 (completely_scalarize): Remove function.
1948 (scalarize_elem): Likewise.
1949 (create_total_scalarization_access): Likewise.
1950 (sort_and_splice_var_accesses): Do not track total scalarization
1951 flags.
1952 (analyze_access_subtree): New parameter totally, adjust to new meaning
1953 of grp_total_scalarization.
1954 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1955 (can_totally_scalarize_forest_p): New function.
1956 (create_total_scalarization_access): Likewise.
1957 (create_total_access_and_reshape): Likewise.
1958 (total_should_skip_creating_access): Likewise.
1959 (totally_scalarize_subtree): Likewise.
1960 (analyze_all_variable_accesses): Perform total scalarization after
1961 subaccess propagation using the new functions above.
1962 (initialize_constant_pool_replacements): Output initializers by
1963 traversing the access tree.
1965 2020-01-29 Martin Jambor <mjambor@suse.cz>
1967 * tree-sra.c (verify_sra_access_forest): New function.
1968 (verify_all_sra_access_forests): Likewise.
1969 (create_artificial_child_access): Set parent.
1970 (analyze_all_variable_accesses): Call the verifier.
1972 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1974 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1975 if called on indirect edge.
1976 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1977 speculative call if needed.
1979 2020-01-29 Richard Biener <rguenther@suse.de>
1981 PR tree-optimization/93428
1982 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1983 permutation when the load node is created.
1984 (vect_analyze_slp_instance): Re-use it here.
1986 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1988 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1990 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1992 PR rtl-optimization/93272
1993 * ira-lives.c (process_out_of_region_eh_regs): New function.
1994 (process_bb_node_lives): Call it.
1996 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1998 * coverage.c (read_counts_file): Make error message lowercase.
2000 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2002 * profile-count.c (profile_quality_display_names): Fix ordering.
2004 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2006 PR lto/93318
2007 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2008 hash only when edge is first within the sequence.
2009 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2010 (symbol_table::create_edge): Do not set target_prob.
2011 (cgraph_edge::remove_caller): Watch for speculative calls when updating
2012 the call site hash.
2013 (cgraph_edge::make_speculative): Drop target_prob parameter.
2014 (cgraph_edge::speculative_call_info): Remove.
2015 (cgraph_edge::first_speculative_call_target): New member function.
2016 (update_call_stmt_hash_for_removing_direct_edge): New function.
2017 (cgraph_edge::resolve_speculation): Rewrite to new API.
2018 (cgraph_edge::speculative_call_for_target): New member function.
2019 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2020 multiple speculation targets.
2021 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2022 of profile.
2023 (verify_speculative_call): Verify that targets form an interval.
2024 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2025 (cgraph_edge::first_speculative_call_target): New member function.
2026 (cgraph_edge::next_speculative_call_target): New member function.
2027 (cgraph_edge::speculative_call_target_ref): New member function.
2028 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2029 (cgraph_edge): Remove target_prob.
2030 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2031 Fix handling of speculative calls.
2032 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2033 * ipa-fnsummary.c (analyze_function_body): Likewise.
2034 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2035 * ipa-profile.c (dump_histogram): Fix formating.
2036 (ipa_profile_generate_summary): Watch for overflows.
2037 (ipa_profile): Do not require probablity to be 1/2; update to new API.
2038 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2039 (update_indirect_edges_after_inlining): Update to new API.
2040 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2041 profiles.
2042 * profile-count.h: (profile_probability::adjusted): New.
2043 * tree-inline.c (copy_bb): Update to new speculative call API; fix
2044 updating of profile.
2045 * value-prof.c (gimple_ic_transform): Rename to ...
2046 (dump_ic_profile): ... this one; update dumping.
2047 (stream_in_histogram_value): Fix formating.
2048 (gimple_value_profile_transformations): Update.
2050 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2052 PR target/91461
2053 * config/i386/i386.md (*movoi_internal_avx): Remove
2054 TARGET_SSE_TYPELESS_STORES check.
2055 (*movti_internal): Prefer TARGET_AVX over
2056 TARGET_SSE_TYPELESS_STORES.
2057 (*movtf_internal): Likewise.
2058 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2059 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
2060 from TARGET_SSE_TYPELESS_STORES.
2062 2020-01-28 David Malcolm <dmalcolm@redhat.com>
2064 * diagnostic-core.h (warning_at): Rename overload to...
2065 (warning_meta): ...this.
2066 (emit_diagnostic_valist): Delete decl of overload taking
2067 diagnostic_metadata.
2068 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2069 (warning_at): Rename overload taking diagnostic_metadata to...
2070 (warning_meta): ...this.
2072 2020-01-28 Richard Biener <rguenther@suse.de>
2074 PR tree-optimization/93439
2075 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2076 * tree-cfg.c (move_sese_region_to_fn): ... here.
2077 (verify_types_in_gimple_reference): Verify used cliques are
2078 tracked.
2080 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2082 PR target/91399
2083 * config/i386/i386-options.c (set_ix86_tune_features): Add an
2084 argument of a pointer to struct gcc_options and pass it to
2085 parse_mtune_ctrl_str.
2086 (ix86_function_specific_restore): Pass opts to
2087 set_ix86_tune_features.
2088 (ix86_option_override_internal): Likewise.
2089 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2090 gcc_options and use it for x_ix86_tune_ctrl_string.
2092 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2094 PR rtl-optimization/87763
2095 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2096 simplification to handle subregs as well as bare regs.
2097 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2099 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2101 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2102 for reduction chains that (now) include a call.
2104 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2106 PR tree-optimization/92822
2107 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2108 out the don't-care elements of a vector whose significant elements
2109 are duplicates, make the don't-care elements duplicates too.
2111 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2113 PR tree-optimization/93434
2114 * tree-predcom.c (split_data_refs_to_components): Record which
2115 components have had aliasing loads removed. Prevent store-store
2116 commoning for all such components.
2118 2020-01-28 Jakub Jelinek <jakub@redhat.com>
2120 PR target/93418
2121 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2122 -1 or is_vshift is true, use new_vector with number of elts npatterns
2123 rather than new_unary_operation.
2125 PR tree-optimization/93454
2126 * gimple-fold.c (fold_array_ctor_reference): Perform
2127 elt_size.to_uhwi () just once, instead of calling it in every
2128 iteration. Punt if that value is above size of the temporary
2129 buffer. Decrease third native_encode_expr argument when
2130 bufoff + elt_sz is above size of buf.
2132 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2134 * config/mips/mips.c (mips_declare_object_name)
2135 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2137 2020-01-27 Martin Liska <mliska@suse.cz>
2139 PR gcov-profile/93403
2140 * tree-profile.c (gimple_init_gcov_profiler): Generate
2141 both __gcov_indirect_call_profiler_v4 and
2142 __gcov_indirect_call_profiler_v4_atomic.
2144 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2146 PR target/92822
2147 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2148 expander.
2149 (@aarch64_split_simd_mov<mode>): Use it.
2150 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2151 Leave the vec_extract patterns to handle 2-element vectors.
2152 (aarch64_simd_mov_from_<mode>high): Likewise.
2153 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2154 (vec_extractv2dfv1df): Likewise.
2156 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2158 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2159 jump conditions for *compare_condjump<GPI:mode>.
2161 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2163 PR analyzer/93276
2164 * digraph.cc (test_edge::test_edge): Specify template for base
2165 class initializer.
2167 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2169 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2171 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2173 * config/arc/arc-protos.h (gen_mlo): Remove.
2174 (gen_mhi): Likewise.
2175 * config/arc/arc.c (AUX_MULHI): Define.
2176 (arc_must_save_reister): Special handling for r58/59.
2177 (arc_compute_frame_size): Consider mlo/mhi registers.
2178 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2179 paramter is true.
2180 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2181 mlo/mhi name selection.
2182 (arc_restore_callee_saves): Don't early restore blink when ISR.
2183 (arc_expand_prologue): Add mlo/mhi saving.
2184 (arc_expand_epilogue): Add mlo/mhi restoring.
2185 (gen_mlo): Remove.
2186 (gen_mhi): Remove.
2187 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2188 numbering when MUL64 option is used.
2189 (DWARF2_FRAME_REG_OUT): Define.
2190 * config/arc/arc.md (arc600_stall): New pattern.
2191 (VUNSPEC_ARC_ARC600_STALL): Define.
2192 (mulsi64): Use correct mlo/mhi registers.
2193 (mulsi_600): Clean it up.
2194 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2195 TARGET_BIG_ENDIAN.
2196 (mhi_operand): Likewise.
2198 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2199 Petro Karashchenko <petro.karashchenko@ring.com>
2201 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2202 attributes if needed.
2203 (prepare_move_operands): Generate special unspec instruction for
2204 direct access.
2205 (arc_isuncached_mem_p): Propagate uncached attribute to each
2206 structure member.
2207 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2208 (VUNSPEC_ARC_STDI): Likewise.
2209 (ALLI): New mode iterator.
2210 (mALLI): New mode attribute.
2211 (lddi): New instruction pattern.
2212 (stdi): Likewise.
2213 (stdidi_split): Split instruction for architectures which are not
2214 supporting ll64 option.
2215 (lddidi_split): Likewise.
2217 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2219 PR rtl-optimization/92989
2220 * lra-lives.c (process_bb_lives): Update the live-in set before
2221 processing additional clobbers.
2223 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2225 PR rtl-optimization/93170
2226 * cselib.c (cselib_invalidate_regno_val): New function, split out
2227 from...
2228 (cselib_invalidate_regno): ...here.
2229 (cselib_invalidated_by_call_p): New function.
2230 (cselib_process_insn): Iterate over all the hard-register entries in
2231 REG_VALUES and invalidate any that cross call-clobbered registers.
2233 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2235 * dojump.c (split_comparison): Use HONOR_NANS rather than
2236 HONOR_SNANS when splitting LTGT.
2238 2020-01-27 Martin Liska <mliska@suse.cz>
2240 PR driver/91220
2241 * opts.c (print_filtered_help): Exclude language-specific
2242 options from --help=common unless enabled in all FEs.
2244 2020-01-27 Martin Liska <mliska@suse.cz>
2246 * opts.c (print_help): Exclude params from
2247 all except --help=param.
2249 2020-01-27 Martin Liska <mliska@suse.cz>
2251 PR target/93274
2252 * config/i386/i386-features.c (make_resolver_func):
2253 Align the code with ppc64 target implementation.
2254 Do not generate a unique name for resolver function.
2256 2020-01-27 Richard Biener <rguenther@suse.de>
2258 PR tree-optimization/93397
2259 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2260 converted reduction chain SLP graph adjustment.
2262 2020-01-26 Marek Polacek <polacek@redhat.com>
2264 PR sanitizer/93436
2265 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2266 null DECL_NAME.
2268 2020-01-26 Jason Merrill <jason@redhat.com>
2270 PR c++/92601
2271 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2272 of complete types.
2274 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2276 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2277 (rx_setmem): Likewise.
2279 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2281 PR target/93412
2282 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2283 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2284 drop <di> from constraint of last operand.
2286 PR target/93430
2287 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2288 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2289 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2291 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2293 PR ipa/93166
2294 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2295 check assertion.
2297 2020-01-24 Jeff Law <law@redhat.com>
2299 PR tree-optimization/92788
2300 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2301 not EDGE_ABNORMAL.
2303 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2305 PR target/93395
2306 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2307 *avx_vperm_broadcast_<mode>,
2308 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2309 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2310 Move before avx2_perm<mode>/avx512f_perm<mode>.
2312 PR target/93376
2313 * simplify-rtx.c (simplify_const_unary_operation,
2314 simplify_const_binary_operation): Punt for mode precision above
2315 MAX_BITSIZE_MODE_ANY_INT.
2317 2020-01-24 Andrew Pinski <apinski@marvell.com>
2319 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2320 alu.shift_reg to 0.
2322 2020-01-24 Jeff Law <law@redhat.com>
2324 PR target/13721
2325 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2326 for REGs. Call output_operand_lossage to get more reasonable
2327 diagnostics.
2329 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2331 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2332 gcn_fp_compare_operator.
2333 (vec_cmpu<mode>di): Use gcn_compare_operator.
2334 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2335 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2336 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2337 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2338 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2339 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2340 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2341 gcn_fp_compare_operator.
2342 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2343 gcn_fp_compare_operator.
2344 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2345 gcn_fp_compare_operator.
2346 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2347 gcn_fp_compare_operator.
2349 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2351 * doc/install.texi (Cross-Compiler-Specific Options): Document
2352 `--with-toolexeclibdir' option.
2354 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2356 * target.def (flags_regnum): Also mention effect on delay slot filling.
2357 * doc/tm.texi: Regenerate.
2359 2020-01-23 Jeff Law <law@redhat.com>
2361 PR translation/90162
2362 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2364 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2366 PR target/92269
2367 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2368 profiling label
2370 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2372 PR rtl-optimization/93402
2373 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2374 USE insns.
2376 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2378 * config.in: Regenerated.
2379 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2380 for TARGET_LIBC_GNUSTACK.
2381 * configure: Regenerated.
2382 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2383 found to be 2.31 or greater.
2385 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2387 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2388 TARGET_SOFT_FLOAT.
2389 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2390 (mips_asm_file_end): New function. Delegate to
2391 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2392 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2394 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2396 PR target/93376
2397 * config/i386/i386-modes.def (POImode): New mode.
2398 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2399 * config/i386/i386.md (DPWI): New mode attribute.
2400 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2401 (QWI): Rename to...
2402 (QPWI): ... this. Use POI instead of OI for TImode.
2403 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2404 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2405 instead of <QWI>.
2407 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2409 PR target/93341
2410 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2411 unspec.
2412 (speculation_tracker_rev): New pattern.
2413 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2414 Use speculation_tracker_rev to track the inverse condition.
2416 2020-01-23 Richard Biener <rguenther@suse.de>
2418 PR tree-optimization/93381
2419 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2420 alias-set of the def as argument and record the first one.
2421 (vn_walk_cb_data::first_set): New member.
2422 (vn_reference_lookup_3): Pass the alias-set of the current def
2423 to push_partial_def. Fix alias-set used in the aggregate copy
2424 case.
2425 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2426 * real.c (clear_significand_below): Fix out-of-bound access.
2428 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2430 PR target/93346
2431 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2432 New define_insn patterns.
2434 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2436 * doc/sourcebuild.texi (check-function-bodies): Add an
2437 optional target/xfail selector.
2439 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2441 PR rtl-optimization/93124
2442 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2443 bare USE and CLOBBER insns.
2445 2020-01-22 Andrew Pinski <apinski@marvell.com>
2447 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2449 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2451 PR analyzer/93307
2452 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2453 diagnostic_manager into "ana" namespace.
2454 * selftest-run-tests.c (selftest::run_tests): Update for move of
2455 selftest::run_analyzer_selftests to
2456 ana::selftest::run_analyzer_selftests.
2458 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2460 * cfgexpand.c (union_stack_vars): Update the size.
2462 2020-01-22 Richard Biener <rguenther@suse.de>
2464 PR tree-optimization/93381
2465 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2466 throughout, handle all conversions the same.
2468 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2470 PR target/93335
2471 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2472 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2473 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2474 Call force_reg on high_in2 unconditionally.
2476 2020-01-22 Martin Liska <mliska@suse.cz>
2478 PR tree-optimization/92924
2479 * profile.c (compute_value_histograms): Divide
2480 all counter values.
2482 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2484 PR target/91298
2485 * output.h (assemble_name_resolve): Declare.
2486 * varasm.c (assemble_name_resolve): New function.
2487 (assemble_name): Use it.
2488 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2490 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2492 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2493 update_web_docs_git instead of update_web_docs_svn.
2495 2020-01-21 Andrew Pinski <apinski@marvell.com>
2497 PR target/9311
2498 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2499 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2500 (*tlsgd_small_<mode>): Likewise.
2501 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2502 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2503 register. Convert that register back to dest using convert_mode.
2505 2020-01-21 Jim Wilson <jimw@sifive.com>
2507 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2508 instead of XINT.
2510 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2511 Uros Bizjak <ubizjak@gmail.com>
2513 PR target/93319
2514 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2515 with ptr_mode.
2516 (legitimize_tls_address): Do GNU2 TLS address computation in
2517 ptr_mode and zero-extend result to Pmode.
2518 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2519 :P with :PTR and Pmode with ptr_mode.
2520 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2521 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2522 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2524 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2526 PR target/93333
2527 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2528 the last two operands are CONST_INT_P before using them as such.
2530 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2532 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2533 to get the integer element types.
2535 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2537 * config/aarch64/aarch64-sve-builtins.h
2538 (function_expander::convert_to_pmode): Declare.
2539 * config/aarch64/aarch64-sve-builtins.cc
2540 (function_expander::convert_to_pmode): New function.
2541 (function_expander::get_contiguous_base): Use it.
2542 (function_expander::prepare_gather_address_operands): Likewise.
2543 * config/aarch64/aarch64-sve-builtins-sve2.cc
2544 (svwhilerw_svwhilewr_impl::expand): Likewise.
2546 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2548 PR target/92424
2549 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2550 cfun->machine->label_is_assembled.
2551 (aarch64_print_patchable_function_entry): New.
2552 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2553 * config/aarch64/aarch64.h (struct machine_function): New field,
2554 label_is_assembled.
2556 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2558 PR ipa/93315
2559 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2560 NULL on exit.
2562 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2564 PR lto/93318
2565 * cgraph.c (cgraph_edge::resolve_speculation,
2566 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2567 call_stmt_site_hash.
2569 2020-01-21 Martin Liska <mliska@suse.cz>
2571 * config/rs6000/rs6000.c (common_mode_defined): Remove
2572 unused variable.
2574 2020-01-21 Richard Biener <rguenther@suse.de>
2576 PR tree-optimization/92328
2577 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2578 type when value-numbering same-sized store by inserting a
2579 VIEW_CONVERT_EXPR.
2580 (eliminate_dom_walker::eliminate_stmt): When eliminating
2581 a redundant store handle bit-reinterpretation of the same value.
2583 2020-01-21 Andrew Pinski <apinski@marvel.com>
2585 PR tree-opt/93321
2586 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2587 from ...
2588 (prepare_block_for_update): This. Use a worklist instead of
2589 recursing.
2591 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2593 * gcc/config/arm/arm.c (clear_operation_p):
2594 Initialise last_regno, skip first iteration
2595 based on the first_set value and use ints instead
2596 of the unnecessary HOST_WIDE_INTs.
2598 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2600 PR target/93073
2601 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2602 compare_mode other than SFmode or DFmode.
2604 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2606 PR target/93304
2607 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2608 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2609 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2611 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2613 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2615 2020-01-20 Andrew Pinski <apinski@marvell.com>
2617 PR middle-end/93242
2618 * targhooks.c (default_print_patchable_function_entry): Use
2619 output_asm_insn to emit the nop instruction.
2621 2020-01-20 Fangrui Song <maskray@google.com>
2623 PR middle-end/93194
2624 * targhooks.c (default_print_patchable_function_entry): Align to
2625 POINTER_SIZE.
2627 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2629 PR target/93319
2630 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2631 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2632 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2633 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2634 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2635 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2636 Remove the {q} suffix from lea.
2637 (*tls_dynamic_gnu2_call_64): Renamed to ...
2638 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2639 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2640 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2641 Pass Pmode to gen_tls_dynamic_gnu2_64.
2643 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2645 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2647 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2649 * config/aarch64/aarch64-sve-builtins-base.cc
2650 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2652 2020-01-20 Richard Biener <rguenther@suse.de>
2654 PR debug/92763
2655 * dwarf2out.c (prune_unused_types): Unconditionally mark
2656 called function DIEs.
2658 2020-01-20 Martin Liska <mliska@suse.cz>
2660 PR tree-optimization/93199
2661 * tree-eh.c (struct leh_state): Add
2662 new field outer_non_cleanup.
2663 (cleanup_is_dead_in): Pass leh_state instead
2664 of eh_region. Add a checking that state->outer_non_cleanup
2665 points to outer non-clean up region.
2666 (lower_try_finally): Record outer_non_cleanup
2667 for this_state.
2668 (lower_catch): Likewise.
2669 (lower_eh_filter): Likewise.
2670 (lower_eh_must_not_throw): Likewise.
2671 (lower_cleanup): Likewise.
2673 2020-01-20 Richard Biener <rguenther@suse.de>
2675 PR tree-optimization/93094
2676 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2677 (vect_transform_loop): Likewise.
2678 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2679 loop_vectorized_call to vect_transform_loop.
2680 * tree-vect-loop.c (vect_transform_loop): Pass down
2681 loop_vectorized_call to vect_loop_versioning.
2682 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2683 the earlier discovered loop_vectorized_call.
2685 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2687 * doc/contribute.texi: Update for SVN -> Git transition.
2688 * doc/install.texi: Likewise.
2690 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2692 PR lto/93318
2693 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2694 speculative targets.
2695 (verify_speculative_call): New function
2696 (cgraph_node::verify_node): Use it.
2697 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2698 speculations.
2700 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2702 PR lto/93318
2703 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2704 (cgraph_edge::make_direct): Remove all indirect targets.
2705 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2706 (cgraph_node::verify_node): Verify that only one call_stmt or
2707 lto_stmt_uid is set.
2708 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2709 lto_stmt_uid.
2710 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2711 (lto_output_ref): Simplify streaming of stmt.
2712 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2714 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2716 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2717 Mark parameter unused.
2719 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2721 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2723 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2725 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2727 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2729 * Makefile.in: Add coroutine-passes.o.
2730 * builtin-types.def (BT_CONST_SIZE): New.
2731 (BT_FN_BOOL_PTR): New.
2732 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2733 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2734 * coroutine-builtins.def: New file.
2735 * coroutine-passes.cc: New file.
2736 * function.h (struct GTY function): Add a bit to indicate that the
2737 function is a coroutine component.
2738 * internal-fn.c (expand_CO_FRAME): New.
2739 (expand_CO_YIELD): New.
2740 (expand_CO_SUSPN): New.
2741 (expand_CO_ACTOR): New.
2742 * internal-fn.def (CO_ACTOR): New.
2743 (CO_YIELD): New.
2744 (CO_SUSPN): New.
2745 (CO_FRAME): New.
2746 * passes.def: Add pass_coroutine_lower_builtins,
2747 pass_coroutine_early_expand_ifns.
2748 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2749 (make_pass_coroutine_early_expand_ifns): New.
2750 * doc/invoke.texi: Document the fcoroutines command line
2751 switch.
2753 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2755 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2757 PR target/93312
2758 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2759 after checking the argument is a REG. Don't use REGNO (reg)
2760 again to set last_regno, reuse regno variable instead.
2762 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2764 * doc/analyzer.texi (Limitations): Add note about NaN.
2766 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2767 Sudakshina Das <sudi.das@arm.com>
2769 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2770 and valid immediate.
2771 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2772 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2773 * config/arm/constraints.md (Pg): New.
2774 * config/arm/predicates.md (long_shift_imm): New.
2775 (arm_reg_or_long_shift_imm): Likewise.
2776 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2777 (thumb2_lsll): Likewise.
2778 (thumb2_lsrl): New.
2780 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2781 Sudakshina Das <sudi.das@arm.com>
2783 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2784 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2785 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2786 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2787 * config/arm/thumb2.md (thumb2_asrl): New.
2788 (thumb2_lsll): Likewise.
2790 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2792 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2793 unused variable.
2795 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2797 * gdbinit.in (help-gcc-hooks): New command.
2798 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2799 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2800 documentation.
2802 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2804 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2805 correct target macro.
2807 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2809 * config/aarch64/aarch64-protos.h
2810 (aarch64_sve_ld1ro_operand_p): New.
2811 * config/aarch64/aarch64-sve-builtins-base.cc
2812 (class load_replicate): New.
2813 (class svld1ro_impl): New.
2814 (class svld1rq_impl): Change to inherit from load_replicate.
2815 (svld1ro): New sve intrinsic function base.
2816 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2817 New DEF_SVE_FUNCTION.
2818 * config/aarch64/aarch64-sve-builtins-base.h
2819 (svld1ro): New decl.
2820 * config/aarch64/aarch64-sve-builtins.cc
2821 (function_expander::add_mem_operand): Modify assert to allow
2822 OImode.
2823 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2824 pattern.
2825 * config/aarch64/aarch64.c
2826 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2827 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2828 (aarch64_sve_ld1ro_operand_p): New.
2829 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2830 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2831 * config/aarch64/predicates.md
2832 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2834 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2836 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2837 Introduce this ACLE specified predefined macro.
2838 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2839 (fp): Disabling this disables f64mm.
2840 (simd): Disabling this disables f64mm.
2841 (fp16): Disabling this disables f64mm.
2842 (sve): Disabling this disables f64mm.
2843 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2844 (AARCH64_ISA_F64MM): New.
2845 (TARGET_F64MM): New.
2846 * doc/invoke.texi (f64mm): Document new option.
2848 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2850 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2851 (neoversen1_tunings): Likewise.
2853 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2855 PR target/92692
2856 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2857 Add assert to ensure prolog has been emitted.
2858 (aarch64_split_atomic_op): Likewise.
2859 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2860 Use epilogue_completed rather than reload_completed.
2861 (aarch64_atomic_exchange<mode>): Likewise.
2862 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2863 (atomic_nand<mode>): Likewise.
2864 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2865 (atomic_fetch_nand<mode>): Likewise.
2866 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2867 (atomic_nand_fetch<mode>): Likewise.
2869 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2871 PR target/93133
2872 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2873 for FP modes.
2874 (REVERSE_CONDITION): Delete.
2875 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2876 (CCFP_CCFPE): Likewise.
2877 (e): New mode attribute.
2878 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2879 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2880 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2881 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2882 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2883 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2884 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2885 name of generator from gen_ccmpdi to gen_ccmpccdi.
2886 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2887 the previous comparison but aren't able to, use the new ccmp_rev
2888 patterns instead.
2890 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2892 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2893 than testing directly for INTEGER_CST.
2894 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2896 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2898 PR tree-optimization/93292
2899 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2900 get_vectype_for_scalar_type returns NULL.
2902 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2904 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2905 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2907 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2909 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2910 dump.
2911 * params.opt: (max-predicted-iterations): Set bounds.
2912 * predict.c (real_almost_one, real_br_prob_base,
2913 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2914 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2915 probabilities; do not truncate to reg_br_prob_bases.
2916 (estimate_loops_at_level): Pass max_cyclic_prob.
2917 (estimate_loops): Compute max_cyclic_prob.
2918 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2919 of back edge prob.
2920 * profile-count.c (profile_probability::to_sreal): New.
2921 * profile-count.h (class sreal): Move up in file.
2922 (profile_probability::to_sreal): Declare.
2924 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2926 * config/arm/arm.c
2927 (arm_invalid_conversion): New function for target hook.
2928 (arm_invalid_unary_op): New function for target hook.
2929 (arm_invalid_binary_op): New function for target hook.
2931 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2933 * config.gcc: Add arm_bf16.h.
2934 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2935 (arm_simd_builtin_std_type): Add BFmode.
2936 (arm_init_simd_builtin_types): Define element types for vector types.
2937 (arm_init_bf16_types): New function.
2938 (arm_init_builtins): Add arm_init_bf16_types function call.
2939 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2940 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2941 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2942 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2943 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2944 (arm_mangle_type): Add __bf16.
2945 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2946 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2947 arm_bf16_ptr_type_node.
2948 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2949 define_split between ARM registers.
2950 * config/arm/arm_bf16.h: New file.
2951 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2952 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2953 (VQXMOV): Add V8BF.
2954 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2955 * config/arm/vfp.md: Add BFmode to movhf patterns.
2957 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2958 Andre Vieira <andre.simoesdiasvieira@arm.com>
2960 * config/arm/arm-cpus.in (mve, mve_float): New features.
2961 (dsp, mve, mve.fp): New options.
2962 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2963 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2964 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2966 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2967 Thomas Preud'homme <thomas.preudhomme@arm.com>
2969 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2970 Armv8-M Mainline.
2971 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2972 error for using -mcmse when targeting Armv8.1-M Mainline.
2974 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2975 Thomas Preud'homme <thomas.preudhomme@arm.com>
2977 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2978 address in r4 when targeting Armv8.1-M Mainline.
2979 (nonsecure_call_value_internal): Likewise.
2980 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2981 a register match_operand again. Emit BLXNS when targeting
2982 Armv8.1-M Mainline.
2983 (nonsecure_call_value_reg_thumb2): Likewise.
2985 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2986 Thomas Preud'homme <thomas.preudhomme@arm.com>
2988 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2989 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2990 variable as true when floating-point ABI is not hard. Replace
2991 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2992 Generate VLSTM and VLLDM instruction respectively before and
2993 after a function call to cmse_nonsecure_call function.
2994 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2995 (VUNSPEC_VLLDM): Likewise.
2996 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2997 (lazy_load_multiple_insn): Likewise.
2999 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3000 Thomas Preud'homme <thomas.preudhomme@arm.com>
3002 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3003 (arm_emit_vfp_multi_reg_pop): Likewise.
3004 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3005 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
3006 restore callee-saved VFP registers.
3008 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3009 Thomas Preud'homme <thomas.preudhomme@arm.com>
3011 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3012 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3013 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
3014 callee-saved GPRs as well as clear ip register before doing a nonsecure
3015 call then restore callee-saved GPRs after it when targeting
3016 Armv8.1-M Mainline.
3017 (arm_reorg): Adapt to function rename.
3019 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3020 Thomas Preud'homme <thomas.preudhomme@arm.com>
3022 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3023 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3024 clear_vfp_multiple pattern based on a new vfp parameter.
3025 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3026 targeting Armv8.1-M Mainline.
3027 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3028 unconditionally when targeting Armv8.1-M Mainline architecture. Check
3029 whether VFP registers are available before looking call_used_regs for a
3030 VFP register.
3031 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3032 of prototype of clear_operation_p.
3033 (clear_vfp_multiple_operation): New predicate.
3034 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3035 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3037 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3038 Thomas Preud'homme <thomas.preudhomme@arm.com>
3040 * config/arm/arm-protos.h (clear_operation_p): Declare.
3041 * config/arm/arm.c (clear_operation_p): New function.
3042 (cmse_clear_registers): Generate clear_multiple instruction pattern if
3043 targeting Armv8.1-M Mainline or successor.
3044 (output_return_instruction): Only output APSR register clearing if
3045 Armv8.1-M Mainline instructions not available.
3046 (thumb_exit): Likewise.
3047 * config/arm/predicates.md (clear_multiple_operation): New predicate.
3048 * config/arm/thumb2.md (clear_apsr): New define_insn.
3049 (clear_multiple): Likewise.
3050 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3052 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3053 Thomas Preud'homme <thomas.preudhomme@arm.com>
3055 * config/arm/arm.c (fp_sysreg_names): Declare and define.
3056 (use_return_insn): Also return false for Armv8.1-M Mainline.
3057 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3058 Mainline instructions are available.
3059 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3060 when targeting Armv8.1-M Mainline Security Extensions.
3061 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3062 Mainline entry function.
3063 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3064 targeting Armv8.1-M Mainline or successor.
3065 (arm_expand_epilogue): Fix indentation of caller-saved register
3066 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
3067 entry function.
3068 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3069 (FP_SYSREGS): Likewise.
3070 (enum vfp_sysregs_encoding): Define enum.
3071 (fp_sysreg_names): Declare.
3072 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3073 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3074 (pop_fpsysreg_insn): Likewise.
3076 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3077 Thomas Preud'homme <thomas.preudhomme@arm.com>
3079 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3080 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3081 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3082 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3083 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3084 (ARMv8_1m_main): New feature group.
3085 (armv8.1-m.main): New architecture.
3086 * config/arm/arm-tables.opt: Regenerate.
3087 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3088 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3089 (arm_options_perform_arch_sanity_checks): Error out when targeting
3090 Armv8.1-M Mainline Security Extensions.
3091 * config/arm/arm.h (arm_arch8_1m_main): Declare.
3093 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3095 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3096 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3097 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3098 aarch64_bfdot_laneq): New.
3099 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3100 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3101 vbfdotq_laneq_f32): New.
3102 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3103 VBFMLA_W, VBF): New.
3104 (isquadop): Add V4BF, V8BF.
3106 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3108 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3109 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3110 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3111 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3112 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3113 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3114 usdot_laneq, sudot_lane,sudot_laneq): New.
3115 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3116 (aarch64_<sur>dot_lane): New.
3117 * config/aarch64/arm_neon.h (vusdot_s32): New.
3118 (vusdotq_s32): New.
3119 (vusdot_lane_s32): New.
3120 (vsudot_lane_s32): New.
3121 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3122 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3124 2020-01-16 Martin Liska <mliska@suse.cz>
3126 * value-prof.c (dump_histogram_value): Fix
3127 obvious spacing issue.
3129 2020-01-16 Andrew Pinski <apinski@marvell.com>
3131 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3132 !storage_order_barrier_p.
3134 2020-01-16 Andrew Pinski <apinski@marvell.com>
3136 * sched-int.h (_dep): Add unused bit-field field for the padding.
3137 * sched-deps.c (init_dep_1): Init unused field.
3139 2020-01-16 Andrew Pinski <apinski@marvell.com>
3141 * optabs.h (create_expand_operand): Initialize target field also.
3143 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3145 PR tree-optimization/92429
3146 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3147 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3148 control folding.
3149 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3150 tree.
3152 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3154 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3155 aarch64_sve_int_mode to each mode.
3157 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3159 * doc/analyzer.texi (Overview): Add note about
3160 -fdump-ipa-analyzer.
3162 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3164 PR tree-optimization/93231
3165 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3166 input_type is unsigned. Use tree_to_shwi for shift constant.
3167 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3168 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3169 inputs more efficiently.
3171 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3173 * config/i386/i386.md (*movsf_internal): Do not require
3174 SSE2 ISA for alternatives 14 and 15.
3176 2020-01-15 Richard Biener <rguenther@suse.de>
3178 PR middle-end/93273
3179 * tree-eh.c (sink_clobbers): If we already visited the destination
3180 block do not defer insertion.
3181 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3182 the purpose of defered insertion.
3184 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3186 * BASE-VER: Bump to 10.0.1.
3188 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3190 PR tree-optimization/93247
3191 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3192 type of the stmt that we're going to vectorize.
3194 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3196 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3197 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3198 type from the lhs.
3200 2020-01-15 Martin Liska <mliska@suse.cz>
3202 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3203 2 calls of streamer_read_hwi in a function call.
3205 2020-01-15 Richard Biener <rguenther@suse.de>
3207 * alias.c (record_alias_subset): Avoid redundant work when
3208 subset is already recorded.
3210 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3212 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3213 the analyzer options provide CWE identifiers.
3215 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3217 * tree-diagnostic-path.cc (path_summary::event_range::print):
3218 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3219 using get_pure_location.
3221 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3223 PR tree-optimization/93262
3224 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3225 perform head trimming only if the last argument is constant,
3226 either all ones, or larger or equal to head trim, in the latter
3227 case decrease the last argument by head_trim.
3229 PR tree-optimization/93249
3230 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3231 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3232 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3233 perform head trim unless we can prove there are no '\0' chars
3234 from the source among the first head_trim chars.
3236 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3238 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3240 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3242 PR target/93009
3243 * config/i386/sse.md
3244 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3245 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3246 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3247 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3248 just a single alternative instead of two, make operands 1 and 2
3249 commutative.
3251 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3253 PR lto/91576
3254 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3255 TYPE_MODE.
3257 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3259 * Makefile.in (lang_opt_files): Add analyzer.opt.
3260 (ANALYZER_OBJS): New.
3261 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3262 tristate.o and ANALYZER_OBJS.
3263 (TEXI_GCCINT_FILES): Add analyzer.texi.
3264 * common.opt (-fanalyzer): New driver option.
3265 * config.in: Regenerate.
3266 * configure: Regenerate.
3267 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3268 (gccdepdir): Also create depdir for "analyzer" subdir.
3269 * digraph.cc: New file.
3270 * digraph.h: New file.
3271 * doc/analyzer.texi: New file.
3272 * doc/gccint.texi ("Static Analyzer") New menu item.
3273 (analyzer.texi): Include it.
3274 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3275 ("Warning Options"): Add static analysis warnings to the list.
3276 (-Wno-analyzer-double-fclose): New option.
3277 (-Wno-analyzer-double-free): New option.
3278 (-Wno-analyzer-exposure-through-output-file): New option.
3279 (-Wno-analyzer-file-leak): New option.
3280 (-Wno-analyzer-free-of-non-heap): New option.
3281 (-Wno-analyzer-malloc-leak): New option.
3282 (-Wno-analyzer-possible-null-argument): New option.
3283 (-Wno-analyzer-possible-null-dereference): New option.
3284 (-Wno-analyzer-null-argument): New option.
3285 (-Wno-analyzer-null-dereference): New option.
3286 (-Wno-analyzer-stale-setjmp-buffer): New option.
3287 (-Wno-analyzer-tainted-array-index): New option.
3288 (-Wno-analyzer-use-after-free): New option.
3289 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3290 (-Wno-analyzer-use-of-uninitialized-value): New option.
3291 (-Wanalyzer-too-complex): New option.
3292 (-fanalyzer-call-summaries): New warning.
3293 (-fanalyzer-checker=): New warning.
3294 (-fanalyzer-fine-grained): New warning.
3295 (-fno-analyzer-state-merge): New warning.
3296 (-fno-analyzer-state-purge): New warning.
3297 (-fanalyzer-transitivity): New warning.
3298 (-fanalyzer-verbose-edges): New warning.
3299 (-fanalyzer-verbose-state-changes): New warning.
3300 (-fanalyzer-verbosity=): New warning.
3301 (-fdump-analyzer): New warning.
3302 (-fdump-analyzer-callgraph): New warning.
3303 (-fdump-analyzer-exploded-graph): New warning.
3304 (-fdump-analyzer-exploded-nodes): New warning.
3305 (-fdump-analyzer-exploded-nodes-2): New warning.
3306 (-fdump-analyzer-exploded-nodes-3): New warning.
3307 (-fdump-analyzer-supergraph): New warning.
3308 * doc/sourcebuild.texi (dg-require-dot): New.
3309 (dg-check-dot): New.
3310 * gdbinit.in (break-on-saved-diagnostic): New command.
3311 * graphviz.cc: New file.
3312 * graphviz.h: New file.
3313 * ordered-hash-map-tests.cc: New file.
3314 * ordered-hash-map.h: New file.
3315 * passes.def (pass_analyzer): Add before
3316 pass_ipa_whole_program_visibility.
3317 * selftest-run-tests.c (selftest::run_tests): Call
3318 selftest::ordered_hash_map_tests_cc_tests.
3319 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3320 decl.
3321 * shortest-paths.h: New file.
3322 * timevar.def (TV_ANALYZER): New timevar.
3323 (TV_ANALYZER_SUPERGRAPH): Likewise.
3324 (TV_ANALYZER_STATE_PURGE): Likewise.
3325 (TV_ANALYZER_PLAN): Likewise.
3326 (TV_ANALYZER_SCC): Likewise.
3327 (TV_ANALYZER_WORKLIST): Likewise.
3328 (TV_ANALYZER_DUMP): Likewise.
3329 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3330 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3331 * tree-pass.h (make_pass_analyzer): New decl.
3332 * tristate.cc: New file.
3333 * tristate.h: New file.
3335 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3337 PR target/93254
3338 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3339 alternatives 9 and 10.
3341 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3343 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3344 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3345 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3346 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3347 (selftest::hash_map_tests_c_tests): Call it.
3348 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3349 New static constant, using the value of = H::empty_zero_p.
3350 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3351 from default_hash_traits <Value>.
3352 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3353 from Traits.
3354 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3355 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3356 calls to mark_empty with !Descriptor::empty_zero_p.
3357 (hash_table::empty_slow): Conditionalize the memset call with a
3358 check that Descriptor::empty_zero_p; otherwise, loop through the
3359 entries calling mark_empty on them.
3360 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3361 (pointer_hash::empty_zero_p): Likewise.
3362 (pair_hash::empty_zero_p): Likewise.
3363 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3364 Likewise.
3365 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3366 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3367 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3368 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3369 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3370 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3371 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3372 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3373 * tree-vectorizer.h
3374 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3375 Likewise.
3377 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3379 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3380 fix typo on return value.
3382 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3384 PR ipa/69678
3385 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3386 target_prob.
3387 (cgraph_edge::make_speculative): Add param for setting speculative_id
3388 and target_prob.
3389 (cgraph_edge::speculative_call_info): Update comments and find reference
3390 by speculative_id for multiple indirect targets.
3391 (cgraph_edge::resolve_speculation): Decrease the speculations
3392 for indirect edge, drop it's speculative if not direct target
3393 left. Update comments.
3394 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3395 (cgraph_node::dump): Print num_speculative_call_targets.
3396 (cgraph_node::verify_node): Don't report error if speculative
3397 edge not include statement.
3398 (cgraph_edge::num_speculative_call_targets_p): New function.
3399 * cgraph.h (int common_target_id): Remove.
3400 (int common_target_probability): Remove.
3401 (num_speculative_call_targets): New variable.
3402 (make_speculative): Add param for setting speculative_id.
3403 (cgraph_edge::num_speculative_call_targets_p): New declare.
3404 (target_prob): New variable.
3405 (speculative_id): New variable.
3406 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3407 call summaries for multiple speculative call targets.
3408 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3409 * ipa-profile.c (struct speculative_call_target): New struct.
3410 (class speculative_call_summary): New class.
3411 (class speculative_call_summaries): New class.
3412 (call_sums): New variable.
3413 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3414 (ipa_profile_write_edge_summary): New function.
3415 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3416 (ipa_profile_dump_all_summaries): New function.
3417 (ipa_profile_read_edge_summary): New function.
3418 (ipa_profile_read_summary_section): New function.
3419 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3420 (ipa_profile): Generate num_speculative_call_targets from
3421 profile summaries.
3422 * ipa-ref.h (speculative_id): New variable.
3423 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3424 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3425 common_target_probability. Stream out speculative_id and
3426 num_speculative_call_targets.
3427 (input_edge): Likewise.
3428 * predict.c (dump_prediction): Remove edges count assert to be
3429 precise.
3430 * symtab.c (symtab_node::create_reference): Init speculative_id.
3431 (symtab_node::clone_references): Clone speculative_id.
3432 (symtab_node::clone_referring): Clone speculative_id.
3433 (symtab_node::clone_reference): Clone speculative_id.
3434 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3435 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3436 if indirect call contains multiple speculative targets.
3437 * value-prof.h (check_ic_target): Remove.
3438 * value-prof.c (gimple_value_profile_transformations):
3439 Use void function gimple_ic_transform.
3440 * value-prof.c (gimple_ic_transform): Handle topn case.
3441 Fix comment typos. Change it to a void function.
3443 2020-01-13 Andrew Pinski <apinski@marvell.com>
3445 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3446 (octeontx2t98): New define.
3447 (octeontx2t96): New define.
3448 (octeontx2t93): New define.
3449 (octeontx2f95): New define.
3450 (octeontx2f95n): New define.
3451 (octeontx2f95mm): New define.
3452 * config/aarch64/aarch64-tune.md: Regenerate.
3453 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3455 2020-01-13 Jason Merrill <jason@redhat.com>
3457 PR c++/33799 - destroy return value if local cleanup throws.
3458 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3460 2020-01-13 Martin Liska <mliska@suse.cz>
3462 * ipa-cp.c (get_max_overall_size): Use newly
3463 renamed param param_ipa_cp_unit_growth.
3464 * params.opt: Remove legacy param name.
3466 2020-01-13 Martin Sebor <msebor@redhat.com>
3468 PR tree-optimization/93213
3469 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3470 stores to be eliminated.
3472 2020-01-13 Martin Liska <mliska@suse.cz>
3474 * opts.c (print_help): Do not print CL_PARAM
3475 and CL_WARNING for CL_OPTIMIZATION.
3477 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3479 PR driver/92757
3480 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3481 depending on optimization settings.
3483 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3485 PR tree-optimization/90838
3486 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3487 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3488 argument rather than to initialize temporary for targets that
3489 don't use the mode argument at all. Initialize ctzval to avoid
3490 warning at -O0.
3492 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3494 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3495 * tree-core.h: Document it.
3496 * gimplify.c (gimplify_omp_workshare): Set it.
3497 * omp-low.c (lower_omp_target): Use it.
3498 * tree-pretty-print.c (dump_omp_clause): Print it.
3500 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3501 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3503 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3505 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3506 * common.opt (fdiagnostics-path-format=): New option.
3507 (diagnostic_path_format): New enum.
3508 (fdiagnostics-show-path-depths): New option.
3509 * coretypes.h (diagnostic_event_id_t): New forward decl.
3510 * diagnostic-color.c (color_dict): Add "path".
3511 * diagnostic-event-id.h: New file.
3512 * diagnostic-format-json.cc (json_from_expanded_location): Make
3513 non-static.
3514 (json_end_diagnostic): Call context->make_json_for_path if it
3515 exists and the diagnostic has a path.
3516 (diagnostic_output_format_init): Clear context->print_path.
3517 * diagnostic-path.h: New file.
3518 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3519 when printing a run of events in a diagnostic_path so that they
3520 all get the same color.
3521 (layout::m_diagnostic_path_p): New field.
3522 (layout::layout): Initialize it.
3523 (layout::print_any_labels): Don't colorize the label text for an
3524 event in a diagnostic_path.
3525 (gcc_rich_location::add_location_if_nearby): Add
3526 "restrict_to_current_line_spans" and "label" params. Pass the
3527 former to layout.maybe_add_location_range; pass the latter
3528 when calling add_range.
3529 * diagnostic.c: Include "diagnostic-path.h".
3530 (diagnostic_initialize): Initialize context->path_format and
3531 context->show_path_depths.
3532 (diagnostic_show_any_path): New function.
3533 (diagnostic_path::interprocedural_p): New function.
3534 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3535 (simple_diagnostic_path::num_events): New function.
3536 (simple_diagnostic_path::get_event): New function.
3537 (simple_diagnostic_path::add_event): New function.
3538 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3539 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3540 (debug): New overload taking a diagnostic_path *.
3541 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3542 * diagnostic.h (enum diagnostic_path_format): New enum.
3543 (json::value): New forward decl.
3544 (diagnostic_context::path_format): New field.
3545 (diagnostic_context::show_path_depths): New field.
3546 (diagnostic_context::print_path): New callback field.
3547 (diagnostic_context::make_json_for_path): New callback field.
3548 (diagnostic_show_any_path): New decl.
3549 (json_from_expanded_location): New decl.
3550 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3551 (-fdiagnostics-show-path-depths): New option.
3552 (-fdiagnostics-color): Add "path" to description of default
3553 GCC_COLORS; describe it.
3554 (-fdiagnostics-format=json): Document how diagnostic paths are
3555 represented in the JSON output format.
3556 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3557 Add optional params "restrict_to_current_line_spans" and "label".
3558 * opts.c (common_handle_option): Handle
3559 OPT_fdiagnostics_path_format_ and
3560 OPT_fdiagnostics_show_path_depths.
3561 * pretty-print.c: Include "diagnostic-event-id.h".
3562 (pp_format): Implement "%@" format code for printing
3563 diagnostic_event_id_t *.
3564 (selftest::test_pp_format): Add tests for "%@".
3565 * selftest-run-tests.c (selftest::run_tests): Call
3566 selftest::tree_diagnostic_path_cc_tests.
3567 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3568 * toplev.c (general_init): Initialize global_dc->path_format and
3569 global_dc->show_path_depths.
3570 * tree-diagnostic-path.cc: New file.
3571 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3572 non-static. Drop "diagnostic" param in favor of storing the
3573 original value of "where" and re-using it.
3574 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3575 maybe_unwind_expanded_macro_loc.
3576 (tree_diagnostics_defaults): Initialize context->print_path and
3577 context->make_json_for_path.
3578 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3579 decl.
3580 (default_tree_make_json_for_path): New decl.
3581 (maybe_unwind_expanded_macro_loc): New decl.
3583 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3585 PR tree-optimization/93210
3586 * fold-const.h (native_encode_initializer,
3587 can_native_interpret_type_p): Declare.
3588 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3589 simplify.
3590 (native_encode_initializer): New function, moved from dwarf2out.c.
3591 Adjust to native_encode_expr compatible arguments, including dry-run
3592 and partial extraction modes. Don't handle STRING_CST.
3593 (can_native_interpret_type_p): No longer static.
3594 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3595 offset / BITS_PER_UNIT fits into int and don't call it if
3596 can_native_interpret_type_p fails. If suboff is NULL and for
3597 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3598 native_encode_initializer.
3599 (fold_const_aggregate_ref_1): Formatting fix.
3600 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3601 (tree_add_const_value_attribute): Adjust caller.
3603 PR tree-optimization/90838
3604 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3605 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3606 CTZ_DEFINED_VALUE_AT_ZERO.
3608 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3610 PR inline-asm/93027
3611 * lra-constraints.c (match_reload): Permit input operands have the
3612 same mode as output while other input operands have a different
3613 mode.
3615 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3617 PR tree-optimization/90838
3618 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3619 (check_ctz_string): Likewise.
3620 (optimize_count_trailing_zeroes): Likewise.
3621 (simplify_count_trailing_zeroes): Likewise.
3622 (pass_forwprop::execute): Try ctz simplification.
3623 * match.pd: Add matching for ctz idioms.
3625 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3627 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3628 for target hook.
3629 (aarch64_invalid_unary_op): New function for target hook.
3630 (aarch64_invalid_binary_op): New function for target hook.
3632 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3634 * config.gcc: Add arm_bf16.h.
3635 * config/aarch64/aarch64-builtins.c
3636 (aarch64_simd_builtin_std_type): Add BFmode.
3637 (aarch64_init_simd_builtin_types): Define element types for vector
3638 types.
3639 (aarch64_init_bf16_types): New function.
3640 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3641 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3642 modes.
3643 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3644 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3645 patterns.
3646 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3647 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3648 * config/aarch64/aarch64.c
3649 (aarch64_classify_vector_mode): Add support for BF types.
3650 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3651 (aarch64_vq_mode): Add support for BF types.
3652 (aarch64_simd_container_mode): Add support for BF types.
3653 (aarch64_mangle_type): Add support for BF scalar type.
3654 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3655 * config/aarch64/arm_bf16.h: New file.
3656 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3657 * config/aarch64/iterators.md: Add BF types to mode attributes.
3658 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3660 2020-01-10 Jason Merrill <jason@redhat.com>
3662 PR c++/93173 - incorrect tree sharing.
3663 * gimplify.c (copy_if_shared): No longer static.
3664 * gimplify.h: Declare it.
3666 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3668 * doc/invoke.texi (-msve-vector-bits=): Document that
3669 -msve-vector-bits=128 now generates VL-specific code for
3670 little-endian targets.
3671 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3672 build_vector_type_for_mode to construct the data vector types.
3673 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3674 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3675 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3676 for 128-bit vectors.
3678 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3680 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3681 invocation.
3683 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3685 * config/aarch64/aarch64-builtins.c
3686 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3687 rather than checking the number of elements and the element mode.
3689 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3691 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3692 get_related_vectype_for_scalar_type rather than build_vector_type
3693 to create the index type for a conditional reduction.
3695 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3697 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3698 for any type of gather or scatter, including strided accesses.
3700 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3702 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3703 comment.
3705 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3707 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3708 get_dr_vinfo_offset
3709 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3710 parameter and its use to reset DR_OFFSET's.
3711 (vect_transform_loop): Remove orig_drs_init argument.
3712 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3713 member of dr_vec_info rather than the offset of the associated
3714 data_reference's innermost_loop_behavior.
3715 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3716 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3717 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3718 get_dr_vinfo_offset.
3719 (vectorizable_store): Likewise.
3720 (vectorizable_load): Likewise.
3722 2020-01-10 Richard Biener <rguenther@suse.de>
3724 * gimple-ssa-store-merging
3725 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3727 2020-01-10 Martin Liska <mliska@suse.cz>
3729 PR ipa/93217
3730 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3731 encapsulation that was there before r280040.
3733 2020-01-10 Richard Biener <rguenther@suse.de>
3735 PR middle-end/93199
3736 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3737 sequences to avoid walking them again for secondary opportunities.
3738 (pass_lower_eh_dispatch::execute): Instead actually insert
3739 them here.
3741 2020-01-10 Richard Biener <rguenther@suse.de>
3743 PR middle-end/93199
3744 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3745 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3746 avoid quadraticness.
3748 2020-01-10 Martin Jambor <mjambor@suse.cz>
3750 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3751 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3752 to get param_ipa_sra_max_replacements.
3753 (param_splitting_across_edge): Pass the caller to
3754 pull_accesses_from_callee.
3756 2020-01-10 Martin Jambor <mjambor@suse.cz>
3758 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3759 * ipa-cp.c (max_new_size): Removed.
3760 (orig_overall_size): New variable.
3761 (get_max_overall_size): New function.
3762 (estimate_local_effects): Use it. Adjust dump.
3763 (decide_about_value): Likewise.
3764 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3765 orig_overall_size. Adjust dump.
3766 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3768 2020-01-10 Martin Jambor <mjambor@suse.cz>
3770 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3771 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3772 instead of param_ipa_max_agg_items.
3773 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3774 optimization info for the callee.
3776 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3778 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3779 markers if debug_inline_points is false.
3781 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3783 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3784 extra_objs.
3785 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3786 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3787 aarch64-sve-builtins-sve2.h.
3788 (aarch64-sve-builtins-sve2.o): New rule.
3789 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3790 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3791 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3792 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3793 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3794 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3795 TARGET_SVE2_SM4.
3796 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3797 instructions that are handled here.
3798 (@cond_asrd<mode>): Generalize to...
3799 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3800 (*cond_asrd<mode>_2): Generalize to...
3801 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3802 (*cond_asrd<mode>_z): Generalize to...
3803 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3804 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3805 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3806 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3807 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3808 pattern.
3809 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3810 (@aarch64_scatter_stnt<mode>): Likewise.
3811 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3812 (@aarch64_mul_lane_<mode>): Likewise.
3813 (@aarch64_sve_suqadd<mode>_const): Likewise.
3814 (*<sur>h<addsub><mode>): Generalize to...
3815 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3816 new pattern.
3817 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3818 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3819 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3820 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3821 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3822 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3823 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3824 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3825 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3826 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3827 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3828 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3829 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3830 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3831 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3832 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3833 (@aarch64_sve2_xar<mode>): Likewise.
3834 (@aarch64_sve2_bcax<mode>): Likewise.
3835 (*aarch64_sve2_eor3<mode>): Rename to...
3836 (@aarch64_sve2_eor3<mode>): ...this.
3837 (@aarch64_sve2_bsl<mode>): New expander.
3838 (@aarch64_sve2_nbsl<mode>): Likewise.
3839 (@aarch64_sve2_bsl1n<mode>): Likewise.
3840 (@aarch64_sve2_bsl2n<mode>): Likewise.
3841 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3842 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3843 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3844 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3845 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3846 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3847 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3848 (<su>mull<bt><Vwide>): Generalize to...
3849 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3850 pattern.
3851 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3852 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3853 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3854 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3855 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3856 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3857 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3858 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3859 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3860 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3861 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3862 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3863 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3864 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3865 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3866 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3867 (<SHRNB:r>shrnb<mode>): Generalize to...
3868 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3869 new pattern.
3870 (<SHRNT:r>shrnt<mode>): Generalize to...
3871 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3872 new pattern.
3873 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3874 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3875 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3876 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3877 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3878 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3879 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3880 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3881 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3882 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3883 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3884 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3885 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3886 (@aarch64_sve2_cvtnt<mode>): Likewise.
3887 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3888 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3889 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3890 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3891 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3892 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3893 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3894 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3895 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3896 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3897 (@aarch64_sve2_pmul<mode>): Likewise.
3898 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3899 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3900 (@aarch64_sve2_tbl2<mode>): Likewise.
3901 (@aarch64_sve2_tbx<mode>): Likewise.
3902 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3903 (@aarch64_sve2_histcnt<mode>): Likewise.
3904 (@aarch64_sve2_histseg<mode>): Likewise.
3905 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3906 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3907 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3908 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3909 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3910 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3911 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3912 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3913 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3914 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3915 (SVE2_PMULL_PAIR_I): New mode iterators.
3916 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3917 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3918 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3919 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3920 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3921 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3922 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3923 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3924 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3925 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3926 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3927 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3928 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3929 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3930 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3931 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3932 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3933 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3934 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3935 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3936 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3937 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3938 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3939 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3940 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3941 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3942 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3943 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3944 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3945 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3946 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3947 further down file.
3948 (VNARROW, Ventype): New mode attributes.
3949 (Vewtype): Handle VNx2DI. Fix typo in comment.
3950 (VDOUBLE): New mode attribute.
3951 (sve_lane_con): Handle VNx8HI.
3952 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3953 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3954 (sve_int_op, sve_int_op_rev): Handle the above codes.
3955 (sve_pred_int_rhs2_operand): Likewise.
3956 (MULLBT, SHRNB, SHRNT): Delete.
3957 (SVE_INT_SHIFT_IMM): New int iterator.
3958 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3959 and UNSPEC_WHILEHS for TARGET_SVE2.
3960 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3961 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3962 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3963 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3964 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3965 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3966 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3967 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3968 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3969 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3970 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3971 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3972 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3973 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3974 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3975 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3976 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3977 (optab): Handle the new unspecs.
3978 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3979 and UNSPEC_RSHRNT.
3980 (lr): Handle the new unspecs.
3981 (bt): Delete.
3982 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3983 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3984 (sve_int_qsub_op): New int attributes.
3985 (sve_fp_op, rot): Handle the new unspecs.
3986 * config/aarch64/aarch64-sve-builtins.h
3987 (function_resolver::require_matching_pointer_type): Declare.
3988 (function_resolver::resolve_unary): Add an optional boolean argument.
3989 (function_resolver::finish_opt_n_resolution): Add an optional
3990 type_suffix_index argument.
3991 (gimple_folder::redirect_call): Declare.
3992 (gimple_expander::prepare_gather_address_operands): Add an optional
3993 bool parameter.
3994 * config/aarch64/aarch64-sve-builtins.cc: Include
3995 aarch64-sve-builtins-sve2.h.
3996 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3997 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3998 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3999 (TYPES_hsd_integer): Use TYPES_hsd_signed.
4000 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4001 (TYPES_s_unsigned): Likewise.
4002 (TYPES_s_integer): Use TYPES_s_unsigned.
4003 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4004 (TYPES_sd_integer): Use them.
4005 (TYPES_d_unsigned): New macro.
4006 (TYPES_d_integer): Use it.
4007 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4008 (TYPES_cvt_narrow): Likewise.
4009 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4010 (preds_mx): New variable.
4011 (function_builder::add_overloaded_function): Allow the new feature
4012 set to be more restrictive than the original one.
4013 (function_resolver::infer_pointer_type): Remove qualifiers from
4014 the pointer type before printing it.
4015 (function_resolver::require_matching_pointer_type): New function.
4016 (function_resolver::resolve_sv_displacement): Handle functions
4017 that don't support 32-bit vector indices or svint32_t vector offsets.
4018 (function_resolver::finish_opt_n_resolution): Take the inferred type
4019 as a separate argument.
4020 (function_resolver::resolve_unary): Optionally treat all forms in
4021 the same way as normal merging functions.
4022 (gimple_folder::redirect_call): New function.
4023 (function_expander::prepare_gather_address_operands): Add an argument
4024 that says whether scaled forms are available. If they aren't,
4025 handle scaling of vector indices and don't add the extension and
4026 scaling operands.
4027 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4028 fall back to using cond_* instead.
4029 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4030 Split out the member variables into...
4031 (rtx_code_function_base): ...this new base class.
4032 (rtx_code_function_rotated): Inherit rtx_code_function_base.
4033 (unspec_based_function): Split out the member variables into...
4034 (unspec_based_function_base): ...this new base class.
4035 (unspec_based_function_rotated): Inherit unspec_based_function_base.
4036 (unspec_based_function_exact_insn): New class.
4037 (unspec_based_add_function, unspec_based_add_lane_function)
4038 (unspec_based_lane_function, unspec_based_pred_function)
4039 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4040 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4041 (unspec_based_sub_function, unspec_based_sub_lane_function): New
4042 typedefs.
4043 (unspec_based_fused_function): New class.
4044 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4045 (unspec_based_fused_lane_function): New class.
4046 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4047 typedefs.
4048 (CODE_FOR_MODE1): New macro.
4049 (fixed_insn_function): New class.
4050 (while_comparison): Likewise.
4051 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4052 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4053 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4054 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4055 (load_gather_sv_restricted, shift_left_imm_long): Declare.
4056 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4057 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4058 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4059 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4060 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4061 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4062 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4063 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4064 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4065 Also add an initial argument for unary_convert_narrowt, regardless
4066 of the predication type.
4067 (build_32_64): Allow loads and stores to specify MODE_none.
4068 (build_sv_index64, build_sv_uint_offset): New functions.
4069 (long_type_suffix): New function.
4070 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4071 (binary_imm_long_base, load_gather_sv_base): Likewise.
4072 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4073 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4074 (unary_narrowb_base, unary_narrowt_base): Likewise.
4075 (binary_long_lane_def, binary_long_lane): New shape.
4076 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4077 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4078 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4079 (binary_to_uint_def, binary_to_uint): Likewise.
4080 (binary_wide_def, binary_wide): Likewise.
4081 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4082 (compare_def, compare): Likewise.
4083 (compare_ptr_def, compare_ptr): Likewise.
4084 (load_ext_gather_index_restricted_def,
4085 load_ext_gather_index_restricted): Likewise.
4086 (load_ext_gather_offset_restricted_def,
4087 load_ext_gather_offset_restricted): Likewise.
4088 (load_gather_sv_def): Inherit from load_gather_sv_base.
4089 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4090 (shift_left_imm_def, shift_left_imm): Likewise.
4091 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4092 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4093 (store_scatter_index_restricted_def,
4094 store_scatter_index_restricted): Likewise.
4095 (store_scatter_offset_restricted_def,
4096 store_scatter_offset_restricted): Likewise.
4097 (tbl_tuple_def, tbl_tuple): Likewise.
4098 (ternary_long_lane_def, ternary_long_lane): Likewise.
4099 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4100 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4101 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4102 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4103 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4104 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4105 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4106 (ternary_uint_def, ternary_uint): Likewise.
4107 (unary_convert): Fix typo in comment.
4108 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4109 (unary_long_def, unary_long): Likewise.
4110 (unary_narrowb_def, unary_narrowb): Likewise.
4111 (unary_narrowt_def, unary_narrowt): Likewise.
4112 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4113 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4114 (unary_to_int_def, unary_to_int): Likewise.
4115 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4116 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4117 (svasrd_impl): Delete.
4118 (svcadd_impl::expand): Handle integer operations too.
4119 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4120 new functions to derive the unspec numbers.
4121 (svmla_svmls_lane_impl): Replace with...
4122 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
4123 integer operations too.
4124 (svwhile_impl): Rename to...
4125 (svwhilelx_impl): ...this and inherit from while_comparison.
4126 (svasrd): Use unspec_based_function.
4127 (svmla_lane): Use svmla_lane_impl.
4128 (svmls_lane): Use svmls_lane_impl.
4129 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4130 (svwhilele, svwhilelt): Use svwhilelx_impl.
4131 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4132 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4133 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4134 * config/aarch64/aarch64-sve-builtins.def: Include
4135 aarch64-sve-builtins-sve2.def.
4137 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4139 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4140 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4141 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4142 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4143 immediates as well as vector ones.
4144 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4145 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4146 (aarch64_sve_qsub_immediate): Update calls accordingly.
4148 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4150 * config/aarch64/aarch64-sve2.md: Add banner comments.
4151 (<su>mulh<r>s<mode>3): Move further up file.
4152 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4153 (*aarch64_sve2_sra<mode>): Move further down file.
4154 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4156 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4158 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4159 and UNSPEC_WHILEWR.
4160 (while_optab_cmp): Handle them.
4161 * config/aarch64/aarch64-sve.md
4162 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4163 and add a "@" marker.
4164 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4165 instead of gen_aarch64_sve2_while_ptest.
4166 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4168 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4170 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4171 (UNSPEC_WHILELE): ...this.
4172 (UNSPEC_WHILE_LO): Rename to...
4173 (UNSPEC_WHILELO): ...this.
4174 (UNSPEC_WHILE_LS): Rename to...
4175 (UNSPEC_WHILELS): ...this.
4176 (UNSPEC_WHILE_LT): Rename to...
4177 (UNSPEC_WHILELT): ...this.
4178 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4179 (cmp_op, while_optab_cmp): Likewise.
4180 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4181 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4182 (svwhilelt): Likewise.
4184 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4186 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4187 (unary_to_uint): Define.
4188 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4189 (unary_count): Rename to...
4190 (unary_to_uint_def, unary_to_uint): ...this.
4191 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4193 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4195 * config/aarch64/aarch64-sve-builtins-functions.h
4196 (code_for_mode_function): New class.
4197 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4198 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4199 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4200 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4201 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4203 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4205 * config/aarch64/iterators.md (addsub): New code attribute.
4206 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4207 Re-express as...
4208 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4209 in the asm string and attributes. Fix indentation.
4210 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4211 Re-express as...
4212 (@aarch64_sve_<optab><mode>): ...this.
4213 * config/aarch64/aarch64-sve-builtins.h
4214 (function_expander::expand_signed_unpred_op): Delete.
4215 * config/aarch64/aarch64-sve-builtins.cc
4216 (function_expander::expand_signed_unpred_op): Likewise.
4217 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4218 try using code_for_aarch64_sve instead.
4219 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4220 (svqsub_impl): Likewise.
4221 (svqadd, svqsub): Use rtx_code_function instead.
4223 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4225 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4226 (HADDSUB, sur, addsub): Remove them.
4228 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4230 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4231 functions the same way, rather than singling out those that
4232 aren't mapped directly to optabs.
4234 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4236 * target.def (compatible_vector_types_p): New target hook.
4237 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4238 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4239 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4240 * doc/tm.texi: Regenerate.
4241 * gimple-expr.c: Include target.h.
4242 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4243 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4244 function.
4245 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4246 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4247 Use the original predicate if it already has a suitable type.
4249 2020-01-09 Martin Jambor <mjambor@suse.cz>
4251 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4252 resolve_speculation and redirect_call_stmt_to_callee static. Change
4253 return type of set_call_stmt to cgraph_edge *.
4254 * auto-profile.c (afdo_indirect_call): Adjust call to
4255 redirect_call_stmt_to_callee.
4256 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4257 make the this pointer explicit, adjust self-recursive calls and the
4258 call top make_direct. Return the resulting edge.
4259 (cgraph_edge::remove): Make this pointer explicit.
4260 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4261 (cgraph_edge::make_direct): Likewise, adjust call to
4262 resolve_speculation.
4263 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4264 call to set_call_stmt.
4265 (cgraph_update_edges_for_call_stmt_node): Update call to
4266 set_call_stmt and remove.
4267 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4268 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4269 (cgraph_node::create_edge_including_clones): Moved "first" definition
4270 of edge to the block where it was used. Adjusted calls to
4271 set_call_stmt.
4272 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4273 cgraph_edge::remove.
4274 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4275 make_direct and redirect_call_stmt_to_callee.
4276 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4277 resolve_speculation and make_direct.
4278 * ipa-inline-transform.c (inline_transform): Adjust call to
4279 redirect_call_stmt_to_callee.
4280 (check_speculations_1):: Adjust call to resolve_speculation.
4281 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4282 resolve-speculation.
4283 (inline_small_functions): Adjust call to resolve_speculation.
4284 (ipa_inline): Likewise.
4285 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4286 make_direct.
4287 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4288 safe with regards to edge removal, adjust calls to
4289 redirect_call_stmt_to_callee.
4290 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4291 and redirect_call_stmt_to_callee.
4292 * multiple_target.c (create_dispatcher_calls): Adjust call to
4293 redirect_call_stmt_to_callee
4294 (redirect_to_specific_clone): Likewise.
4295 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4296 Adjust calls to cgraph_edge::remove.
4297 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4298 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4299 (expand_call_inline): Adjust call to cgraph_edge::remove.
4301 2020-01-09 Martin Liska <mliska@suse.cz>
4303 * params.opt: Set Optimization for
4304 param_max_speculative_devirt_maydefs.
4306 2020-01-09 Martin Sebor <msebor@redhat.com>
4308 PR middle-end/93200
4309 PR fortran/92956
4310 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4312 2020-01-09 Martin Liska <mliska@suse.cz>
4314 * auto-profile.c (auto_profile): Use opt_for_fn
4315 for a parameter.
4316 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4317 (propagate_vals_across_arith_jfunc): Likewise.
4318 (hint_time_bonus): Likewise.
4319 (incorporate_penalties): Likewise.
4320 (good_cloning_opportunity_p): Likewise.
4321 (perform_estimation_of_a_value): Likewise.
4322 (estimate_local_effects): Likewise.
4323 (ipcp_propagate_stage): Likewise.
4324 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4325 (set_switch_stmt_execution_predicate): Likewise.
4326 (analyze_function_body): Likewise.
4327 * ipa-inline-analysis.c (offline_size): Likewise.
4328 * ipa-inline.c (early_inliner): Likewise.
4329 * ipa-prop.c (ipa_analyze_node): Likewise.
4330 (ipcp_transform_function): Likewise.
4331 * ipa-sra.c (process_scan_results): Likewise.
4332 (ipa_sra_summarize_function): Likewise.
4333 * params.opt: Rename ipcp-unit-growth to
4334 ipa-cp-unit-growth. Add Optimization for various
4335 IPA-related parameters.
4337 2020-01-09 Richard Biener <rguenther@suse.de>
4339 PR middle-end/93054
4340 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4342 2020-01-09 Richard Biener <rguenther@suse.de>
4344 PR tree-optimization/93040
4345 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4347 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4349 * common/config/avr/avr-common.c (avr_option_optimization_table)
4350 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4352 2020-01-09 Martin Liska <mliska@suse.cz>
4354 * cgraphclones.c (symbol_table::materialize_all_clones):
4355 Use cgraph_node::dump_name.
4357 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4359 PR inline-asm/93202
4360 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4361 output_operand_lossage instead of gcc_unreachable.
4362 * doc/md.texi (riscv f constraint): Fix typo.
4364 PR target/93141
4365 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4366 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4367 CONST_SCALAR_INT_P instead of CONST_INT_P.
4368 (*subv<mode>4_1): Rename to ...
4369 (subv<mode>4_1): ... this.
4370 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4371 define_insn_and_split patterns.
4372 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4373 patterns.
4375 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4377 * vec.c (class selftest::count_dtor): New class.
4378 (selftest::test_auto_delete_vec): New test.
4379 (selftest::vec_c_tests): Call it.
4380 * vec.h (class auto_delete_vec): New class template.
4381 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4383 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4385 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4387 2020-01-08 Jim Wilson <jimw@sifive.com>
4389 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4390 use of TLS_MODEL_LOCAL_EXEC when not pic.
4392 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4394 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4395 memory leak.
4397 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4399 PR target/93187
4400 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4401 *stack_protect_set_3 peephole2): Also check that the second
4402 insns source is general_operand.
4404 PR target/93174
4405 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4406 predicate for output operand instead of register_operand.
4407 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4408 memory destination and non-memory operands[2].
4410 2020-01-08 Martin Liska <mliska@suse.cz>
4412 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4413 ::dump_asm_name instead of (::name or ::asm_name).
4414 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4415 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4416 (analyze_functions): Likewise.
4417 (expand_all_functions): Likewise.
4418 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4419 (propagate_bits_across_jump_function): Likewise.
4420 (dump_profile_updates): Likewise.
4421 (ipcp_store_bits_results): Likewise.
4422 (ipcp_store_vr_results): Likewise.
4423 * ipa-devirt.c (dump_targets): Likewise.
4424 * ipa-fnsummary.c (analyze_function_body): Likewise.
4425 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4426 (process_hsa_functions): Likewise.
4427 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4428 (set_alias_uids): Likewise.
4429 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4430 * ipa-inline.c (recursive_inlining): Likewise.
4431 (inline_to_all_callers_1): Likewise.
4432 (ipa_inline): Likewise.
4433 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4434 (ipa_propagate_frequency): Likewise.
4435 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4436 (remove_described_reference): Likewise.
4437 * ipa-pure-const.c (worse_state): Likewise.
4438 (check_retval_uses): Likewise.
4439 (analyze_function): Likewise.
4440 (propagate_pure_const): Likewise.
4441 (propagate_nothrow): Likewise.
4442 (dump_malloc_lattice): Likewise.
4443 (propagate_malloc): Likewise.
4444 (pass_local_pure_const::execute): Likewise.
4445 * ipa-visibility.c (optimize_weakref): Likewise.
4446 (function_and_variable_visibility): Likewise.
4447 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4448 (ipa_discover_variable_flags): Likewise.
4449 * lto-streamer-out.c (output_function): Likewise.
4450 (output_constructor): Likewise.
4451 * tree-inline.c (copy_bb): Likewise.
4452 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4453 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4455 2020-01-08 Richard Biener <rguenther@suse.de>
4457 PR middle-end/93199
4458 * tree-eh.c (sink_clobbers): Update virtual operands for
4459 the first and last stmt only. Add a dry-run capability.
4460 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4461 after CFG manipulations and in RPO order to catch all
4462 secondary opportunities reliably.
4464 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4466 PR target/93182
4467 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4469 2019-01-08 Richard Biener <rguenther@suse.de>
4471 PR middle-end/93199
4472 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4473 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4474 virtual operand, also updating SSA use.
4475 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4476 Update stmt after resetting virtual operand.
4477 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4478 * gimple-iterator.c (gsi_remove): When not removing the stmt
4479 permanently do not delink immediate uses or mark the stmt modified.
4481 2020-01-08 Martin Liska <mliska@suse.cz>
4483 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4484 (ipa_call_context::estimate_size_and_time): Likewise.
4485 (inline_analyze_function): Likewise.
4487 2020-01-08 Martin Liska <mliska@suse.cz>
4489 * cgraph.c (cgraph_node::dump): Use systematically
4490 dump_asm_name.
4492 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4494 Add -nodevicespecs option for avr.
4496 PR target/93182
4497 * config/avr/avr.opt (-nodevicespecs): New driver option.
4498 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4499 "-specs=device-specs/..." if that option is not set.
4500 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4502 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4504 Implement 64-bit double functions for avr.
4506 PR target/92055
4507 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4508 --with-double-comparison.
4509 * doc/install.texi: Document them.
4510 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4511 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4512 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4513 * doc/invoke.texi (AVR Built-in Macros): Document them.
4514 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4515 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4516 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4518 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4520 PR target/93188
4521 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4522 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4523 when only building rm-profile multilibs.
4525 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4527 PR ipa/93084
4528 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4529 lattice for a value to check.
4530 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4531 finite propagation in self-recursive scc.
4533 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4535 * ipa-inline.c (caller_growth_limits): Restore the AND.
4537 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4539 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4540 (VEC_ALLREG_ALT): New iterator.
4541 (VEC_ALLREG_INT_MODE): New iterator.
4542 (VCMP_MODE): New iterator.
4543 (VCMP_MODE_INT): New iterator.
4544 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4545 (vec_cmp<u>v64qidi): New define_expand.
4546 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4547 (vec_cmpu<mode>di_exec): New define_expand.
4548 (vec_cmp<u>v64qidi_exec): New define_expand.
4549 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4550 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4551 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4552 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4553 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4554 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4555 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4556 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4557 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4558 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4559 this.
4560 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4561 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4563 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4565 * config/gcn/constraints.md (DA): Update description and match.
4566 (DB): Likewise.
4567 (Db): New constraint.
4568 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4569 parameter.
4570 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4571 Implement 'Db' mixed immediate type.
4572 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4573 (addcv64si3_dup<exec_vcc>): Delete.
4574 (subcv64si3<exec_vcc>): Rework constraints.
4575 (addv64di3): Rework constraints.
4576 (addv64di3_exec): Rework constraints.
4577 (subv64di3): Rework constraints.
4578 (addv64di3_dup): Delete.
4579 (addv64di3_dup_exec): Delete.
4580 (addv64di3_zext): Rework constraints.
4581 (addv64di3_zext_exec): Rework constraints.
4582 (addv64di3_zext_dup): Rework constraints.
4583 (addv64di3_zext_dup_exec): Rework constraints.
4584 (addv64di3_zext_dup2): Rework constraints.
4585 (addv64di3_zext_dup2_exec): Rework constraints.
4586 (addv64di3_sext_dup2): Rework constraints.
4587 (addv64di3_sext_dup2_exec): Rework constraints.
4589 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4591 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4592 existing target checks.
4594 2020-01-07 Richard Biener <rguenther@suse.de>
4596 * doc/install.texi: Bump minimal supported MPC version.
4598 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4600 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4601 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4602 * langhooks.c: Include stor-layout.h.
4603 (lhd_simulate_enum_decl): New function.
4604 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4605 handle_arm_sve_h for the LTO frontend.
4606 (register_vector_type): Cope with null returns from pushdecl.
4608 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4610 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4611 (aarch64_sve::nvectors_if_data_type): Replace with...
4612 (aarch64_sve::builtin_type_p): ...this.
4613 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4614 (find_vector_type): Delete.
4615 (add_sve_type_attribute): New function.
4616 (lookup_sve_type_attribute): Likewise.
4617 (register_builtin_types): Add an "SVE type" attribute to each type.
4618 (register_tuple_type): Likewise.
4619 (svbool_type_p, nvectors_if_data_type): Delete.
4620 (mangle_builtin_type): Use lookup_sve_type_attribute.
4621 (builtin_type_p): Likewise. Add an overload that returns the
4622 number of constituent vector and predicate registers.
4623 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4624 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4625 instead of aarch64_sve_argument_p.
4626 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4627 (aarch64_pass_by_reference): Likewise.
4628 (aarch64_function_value_1): Likewise.
4629 (aarch64_return_in_memory): Likewise.
4630 (aarch64_layout_arg): Likewise.
4632 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4634 PR tree-optimization/93156
4635 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4636 least significant bit is always clear.
4638 PR tree-optimization/93118
4639 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4640 simplifier with two intermediate conversions.
4642 2020-01-07 Martin Liska <mliska@suse.cz>
4644 * params.opt: Add Optimization for various parameters.
4646 2020-01-07 Martin Liska <mliska@suse.cz>
4648 PR ipa/83411
4649 * doc/extend.texi: Explain cloning for target_clone
4650 attribute.
4652 2020-01-07 Martin Liska <mliska@suse.cz>
4654 PR tree-optimization/92860
4655 * common.opt: Make in Optimization option
4656 as it is affected by -O0, which is an Optimization
4657 option.
4658 * tree-inline.c (tree_inlinable_function_p):
4659 Use opt_for_fn for warn_inline.
4660 (expand_call_inline): Likewise.
4662 2020-01-07 Martin Liska <mliska@suse.cz>
4664 PR tree-optimization/92860
4665 * common.opt: Make flag_ree as optimization
4666 attribute.
4668 2020-01-07 Martin Liska <mliska@suse.cz>
4670 PR optimization/92860
4671 * params.opt: Mark param_min_crossjump_insns with Optimization
4672 keyword.
4674 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4676 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4677 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4679 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4681 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4682 helper function to return the valid addressing formats for a given
4683 hard register and mode.
4684 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4686 * config/rs6000/constraints.md (Q constraint): Update
4687 documentation.
4688 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4689 documentation.
4691 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4692 Use 'Q' for doing vector extract from memory.
4693 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4694 memory.
4695 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4696 doing vector extract from memory.
4697 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4698 extract from memory.
4700 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4701 for the offset being 34-bits when -mcpu=future is used.
4703 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4705 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4706 instead of cmpib_comparison_operator in cmpib patterns.
4707 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4708 of cmpib_comparison_operator. Revise comment.
4710 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4712 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4713 in an IFN_DIV_POW2 node to be equal.
4715 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4717 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4718 (vect_check_scalar_mask): ...this.
4719 (vectorizable_store, vectorizable_load): Update call accordingly.
4720 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4721 argument in calls to conditional internal functions.
4723 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4725 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4726 '0' matching inputs.
4727 (subv64di3_exec): Likewise.
4729 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4731 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4732 * doc/md.texi (movstr): Likewise.
4734 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4736 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4737 clobber.
4739 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4741 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4742 Depend on...
4743 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4744 to a temporary file and use move-if-change to update the real
4745 file where necessary.
4747 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4749 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4750 rather than Upa for CPY /M.
4752 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4754 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4755 immediate.
4757 2020-01-06 Martin Liska <mliska@suse.cz>
4759 PR tree-optimization/92860
4760 * params.opt: Mark param_max_combine_insns with Optimization
4761 keyword.
4763 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4765 PR target/93141
4766 * config/i386/i386.md (SWIDWI): New mode iterator.
4767 (DWI, dwi): Add TImode variants.
4768 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4769 <general_hilo_operand> instead of <general_operand>. Use
4770 CONST_SCALAR_INT_P instead of CONST_INT_P.
4771 (*addv<mode>4_1): Rename to ...
4772 (addv<mode>4_1): ... this.
4773 (QWI): New mode attribute.
4774 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4775 define_insn_and_split patterns.
4776 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4777 patterns.
4778 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4779 <general_hilo_operand> instead of <general_operand>.
4780 (*addcarry<mode>_1): New define_insn.
4781 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4783 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4785 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4786 Use "call" instead of "set".
4788 2020-01-03 Martin Jambor <mjambor@suse.cz>
4790 PR ipa/92917
4791 * ipa-cp.c (print_all_lattices): Skip functions without info.
4793 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4795 PR target/93089
4796 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4797 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4798 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4799 for 'e' simd clones.
4801 PR target/93089
4802 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4803 entry.
4804 (mprefer-vector-width=): Add Save.
4805 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4806 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4807 (ix86_debug_options, ix86_function_specific_print): Adjust
4808 ix86_target_string callers.
4809 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4810 (ix86_valid_target_attribute_tree): Likewise.
4811 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4812 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4813 ix86_target_string caller.
4815 PR target/93110
4816 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4817 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4818 instead of gen_int_shift_amount + convert_modes.
4820 PR rtl-optimization/93088
4821 * loop-iv.c (find_single_def_src): Punt after looking through
4822 128 reg copies for regs with single definitions. Move definitions
4823 to first uses.
4825 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4827 * config/arm/arm-c.c (arm_cpu_builtins): Define
4828 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4829 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4830 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4831 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4832 * config/arm/arm-tables.opt: Regenerated.
4833 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4834 arm_arch_i8mm and arm_arch_bf16 when enabled.
4835 * config/arm/arm.h (TARGET_I8MM): New macro.
4836 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4837 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4838 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4839 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4840 (v8_6_a_simd_variants): New.
4841 (v8_*_a_simd_variants): Add i8mm and bf16.
4842 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4844 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4846 PR ipa/93087
4847 * predict.c (compute_function_frequency): Don't call
4848 warn_function_cold on functions that already have cold attribute.
4850 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4852 PR target/67834
4853 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4854 COMDAT group function labels in .data.rel.ro.local section.
4855 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4857 PR target/93111
4858 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4859 comparison_operator in B and S integer comparisons. Likewise, use
4860 ordered_comparison_operator instead of cmpib_comparison_operator in
4861 cmpib patterns.
4862 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4864 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4866 Update copyright years.
4868 * gcc.c (process_command): Update copyright notice dates.
4869 * gcov-dump.c (print_version): Ditto.
4870 * gcov.c (print_version): Ditto.
4871 * gcov-tool.c (print_version): Ditto.
4872 * gengtype.c (create_file): Ditto.
4873 * doc/cpp.texi: Bump @copying's copyright year.
4874 * doc/cppinternals.texi: Ditto.
4875 * doc/gcc.texi: Ditto.
4876 * doc/gccint.texi: Ditto.
4877 * doc/gcov.texi: Ditto.
4878 * doc/install.texi: Ditto.
4879 * doc/invoke.texi: Ditto.
4881 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4883 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4884 summary.
4886 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4888 PR tree-optimization/93098
4889 * match.pd (popcount): For shift amounts, use integer_onep
4890 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4891 tests. Make sure that precision is power of two larger than or equal
4892 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4893 instead of ULL suffixed constants. Formatting fixes.
4894 \f
4895 Copyright (C) 2020 Free Software Foundation, Inc.
4897 Copying and distribution of this file, with or without modification,
4898 are permitted in any medium without royalty provided the copyright
4899 notice and this notice are preserved.